JP2008118693A - Digital signal conversion method and digital signal conversion device - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a digital signal conversion method and a digital signal conversion device which can efficiently perform processing for resolution conversion etc., and have small deterioration of signals. <P>SOLUTION: An input digital signal of a first format (DV video signal) is restored to a variable-length code by having its framing canceled by a de-framing section 11, then decoded by a variable-length decoding (VLD) section 12, inversely quantized by an inverse quantizing (IQ) section 13, and resolution-converted by a resolution conversion section 16. After that, the video signal having the resolution converted is inversely weighted by the first format and also weighted by a second format by a weighting (IW*W) section 18, then quantized by a quantizing (Q) section 19, coded by variable-length coding by a variable-length coding (VLC) section 20, and output as a digital signal of a second format (MPEG video signal). The weighting section 18 may be disposed in front of the resolution conversion section 16. <P>COPYRIGHT: (C)2008,JPO&INPIT

Description

  The present invention relates to a conversion process of a digital signal compressed and encoded using orthogonal transform such as discrete cosine transform (DCT), and more particularly, a digital signal conversion method for converting resolution between compressed video signals having different formats and The present invention relates to a digital signal converter.

  Conventionally, discrete cosine transform (DCT), which is a kind of orthogonal transform coding, is used as an encoding method for efficiently compressing and encoding still image data, moving image data, and the like. When handling such orthogonally-transformed digital signals, it may be necessary to change the resolution or conversion base.

  For example, a second digital signal whose resolution is so-called MPEG1 format is 360 × 240 pixels from a first orthogonally-transformed digital signal whose resolution is 720 × 480 pixels, which is one of home digital video formats. In the case of conversion to the orthogonally transformed digital signal, after performing inverse orthogonal transformation on the first signal to restore the signal in the spatial domain, necessary conversion processing such as interpolation and decimation is performed. And orthogonal transform is performed again to convert the signal into the second signal.

  In this way, the orthogonally transformed digital signal is often inversely transformed and returned to the original signal, and then a required transformation operation is performed, and then orthogonal transformation is performed again in many cases.

  FIG. 28 shows a configuration example of a conventional digital signal processing apparatus for performing resolution conversion as described above on a DCT-converted digital signal.

In this conventional digital signal converter, a so-called “DV format” video signal (hereinafter referred to as a DV video signal), which is one of the formats of a home digital video signal, is input as a digital signal of the first format. , So-called MPEG
A video signal (hereinafter referred to as an MPEG video signal) conforming to the format of (Moving Picture Experts Group) is output as a digital signal of the second format.

  The deframing unit 51 is for solving the framing of the DV video signal. In the deframing unit 51, a DV video signal that has been framed according to a so-called DV format is returned to a variable length code.

  A variable length decoding (VLD) unit 52 performs variable length decoding on the video signal returned to the variable length code by the deframing unit 51. Compressed data in the DV format is compressed at a fixed rate so that the amount of data is about 1/5 of the original signal, and is variable-length encoded to increase data compression efficiency. The variable length decoding unit 52 performs decoding according to the variable length encoding.

  The inverse quantization (IQ) unit 53 inversely quantizes the video signal decoded by the variable length decoding unit 52.

  The inverse weighting (IW) unit 54 performs inverse weighting, which is an inverse operation of weighting performed on the video signal inversely quantized by the inverse quantization unit 53.

  Here, the weighting refers to making the DCT coefficient value smaller for the high frequency component of the video signal by utilizing the property that the human visual characteristic is not very sensitive to the high frequency distortion. As a result, the number of high frequency coefficients having a value of 0 increases, and the efficiency of variable length coding can be improved. As a result, the amount of calculation of the DCT conversion may be reduced.

  An inverse discrete cosine transform (IDCT) unit 55 performs inverse DCT (inverse discrete cosine transform) on the video signal inversely weighted by the inverse weighting unit 54 to return the DCT coefficient to spatial domain data, that is, pixel data.

  Then, the resolution conversion unit 56 performs a required resolution conversion on the video signal converted back to the pixel data by the inverse discrete cosine conversion unit 55.

  Next, a discrete cosine transform (DCT) unit 57 performs a discrete cosine transform (DCT) on the video signal whose resolution has been converted by the resolution conversion unit 56 and converts the video signal again into an orthogonal transform coefficient (DCT coefficient).

  The weighting (W) unit 58 performs weighting on the video signal after resolution conversion converted into the DCT coefficient. This weighting is as described above.

  The quantization (Q) unit 59 quantizes the video signal weighted by the weighting unit 58.

  Then, the variable length coding (VLC) unit 60 performs variable length coding on the video signal quantized by the quantization unit 59 and outputs it as an MPEG video signal.

  The above-mentioned “MPEG” is ISO / IEC JTC1 / SC29 (International Organization for Standardization / International Electrotechnical Commission, Joint Technical Committee 1 / Sub Committee 29: International Organization for Standardization / International Electrotechnical Commission Joint Technical Committee 1 / Specialized This is an abbreviation for the Moving Picture Image Coding Experts Group of the group 29), which includes ISO11172 as the MPEG1 standard and ISO13818 as the MPEG2 standard. In these international standards, ISO11172-1 and ISO13818-1 for multimedia multiplexing items, ISO11172-2 and ISO13818-2 for video items, and ISO11172-3 and ISO13818-3 for audio items, respectively. It has been standardized.

  In ISO11172-2 or ISO13818-2 as an image compression coding standard, image signals are compressed and coded using picture time and space direction correlation in picture (frame or field) units. The use of the correlation in the spatial direction is realized by using DCT coding.

  In addition to this, orthogonal transform such as DCT is widely used for various image information compression encoding such as JPEG (Joint Photographic Coding Experts group).

  In general, orthogonal transform enables compression encoding with high compression efficiency and excellent reproducibility by transforming an original signal in the time domain or space domain into a domain subjected to orthogonal transform such as a frequency domain.

  The above-mentioned “DV format” is for compressing the data amount of the digital video signal to about 1/5 and recording the component on the magnetic tape. The home digital video device or the commercial digital video device is used. It is used for a part of. This DV format realizes efficient compression of video signals by combining discrete cosine transform (DCT) and variable length coding (VLC).

  Patent documents 1 to 4 are known as conventional techniques.

Japanese Patent Application Laid-Open No. 08-066563 JP 09-219861 A Japanese Patent Laid-Open No. 05-111005 Japanese Patent Application Laid-Open No. 07-203443

  By the way, since orthogonal transformation such as discrete cosine transformation (DCT) and its inverse transformation usually require a large amount of calculation, there is a problem that the above-described video signal resolution conversion cannot be performed efficiently. . There is also a problem that the signal deteriorates because errors are accumulated as the amount of calculation increases.

  The present invention has been made to solve such a problem, and by reducing the calculation processing amount of the data amount of a signal subjected to processing such as resolution conversion in order to convert to a different format, the resolution It is an object of the present invention to provide a digital signal conversion method and a digital signal conversion apparatus that can efficiently perform conversion processing such as conversion and that cause little signal deterioration.

  In order to solve the above-described problems, a digital signal conversion method according to the present invention is a digital signal conversion method for converting a digital signal of a first format into a digital signal of a second format. A decoding step for decoding a digital signal in the format; a signal conversion step for converting the decoded digital signal in the first format into a digital signal in the second format; and encoding the digital signal in the second format And a weighting processing step for collectively performing inverse weighting on the decoded digital signal in the first format and weighting on the digital signal in the second format. .

  Further, in the present invention proposed to solve the above problems, in the digital signal converter for converting the digital signal of the first format into the digital signal of the second format, the digital signal of the first format is converted. Decoding means for decoding; signal converting means for converting the decoded digital signal of the first format into a digital signal of the second format; encoding means for encoding the second digital signal; There is provided weighting processing means for collectively performing inverse weighting on the digital signal in the first format and weighting on the digital signal in the second format.

  According to the present invention, the amount of calculation is reduced as compared with the case where inverse weighting and weighting, which are inverse operations of weighting applied to the input digital signal of the first format, are performed separately, resolution conversion, etc. Therefore, it is possible to provide a digital signal conversion processing method and a digital signal conversion processing apparatus that can efficiently perform the conversion process of FIG.

  Hereinafter, preferred embodiments of the present invention will be described with reference to the drawings.

  In the following, the configuration of the digital signal conversion apparatus according to the present invention will be described first, and then the digital signal conversion method according to the present invention will be described with reference to the configuration.

  FIG. 1 shows an example of the configuration of the main part of the digital signal converter according to the first embodiment of the present invention. The signal conversion is exemplified by resolution conversion, but the present invention is not limited to this. Of course, the present invention can be applied to various signal conversions such as filter processing and the like.

  In this digital signal converter, the video signal (hereinafter referred to as DV signal) of the so-called “DV format” described above is input as the first digital signal, and the video signal (hereinafter referred to as the MPEG (Moving Picture Experts Group) format). MPEG video signal) is output as a second digital signal.

  The deframing unit 11 is for solving the framing of the DV video signal. In the deframing unit 11, a DV video signal that has been framed according to a predetermined format (so-called DV format) is returned to a variable-length code.

  The variable length decoding (VLD) unit 12 performs variable length decoding on the video signal returned to the variable length code by the deframing unit 11.

  The inverse quantization (IQ) unit 13 inversely quantizes the video signal variable-length decoded by the variable-length decoding unit 12.

  The inverse weighting (IW) unit 14 performs inverse weighting, which is an inverse operation of weighting applied to the video signal inversely quantized by the inverse quantization unit 14.

  When resolution conversion is performed as an example of signal conversion processing, the resolution conversion unit 16 performs necessary resolution conversion in the orthogonal transform region (frequency region) on the video signal inversely weighted by the inverse weighting unit 14. Applied.

  The weighting (W) unit 18 weights the video signal after resolution conversion.

  The quantization (Q) unit 19 quantizes the video signal weighted by the weighting unit 18.

  Then, the variable length coding (VLC) unit 20 performs variable length coding on the video signal quantized by the quantization unit 19 and outputs it as an MPEG video signal.

  The configuration of each part of the digital signal converter according to the present invention illustrated in FIG. 1 described above can be the same as that of each part of the conventional digital signal converter illustrated in FIG.

  However, the digital signal converter according to the present invention is different from the conventional digital signal converter in that an inverse cosine transform (IDCT) unit and a cosine transform (DCT) unit are not arranged before and after the resolution conversion unit 16. Yes.

  That is, the conventional digital signal conversion apparatus performs a required conversion operation after inverse orthogonal transformation is performed on the orthogonal transformation coefficient of the input digital signal of the first format to return the data to the spatial domain (on the frequency axis). Therefore, an operation for returning to the orthogonal transformation coefficient by performing the orthogonal transformation again is performed.

  On the other hand, the digital signal conversion apparatus according to the present invention performs a required conversion operation on the orthogonal transform coefficient of the input digital signal of the first format in the orthogonal transform coefficient domain (frequency domain), and performs resolution conversion or the like. It is characterized in that the inverse orthogonal transforming means and the orthogonal transforming means are not provided before and after the means for performing the conversion process.

  Next, the principle of resolution conversion processing in the resolution conversion unit 16 will be described with reference to FIGS.

In FIG. 2, an input orthogonal transformation matrix generation unit 1 generates an inverse matrix Ts _(k) ⁻¹ of an orthogonal transformation matrix Ts _(k) representing an orthogonal transformation performed in advance on the input digital signal 5, and a transformation matrix. It is sent to the generation unit 3. The output orthogonal transformation matrix generation unit 2 generates an orthogonal transformation matrix Td _(L) corresponding to the inverse transformation matrix Td _(L) ⁻¹ indicating the inverse orthogonal transformation performed on the output digital signal, and the transformation matrix generation unit It is sent to 3. The conversion matrix generation unit 3 generates a conversion matrix D for performing conversion processing such as resolution conversion in the frequency domain, and sends it to the signal conversion unit 4. The signal conversion unit 4 converts an input digital signal 5 converted into, for example, a frequency domain by orthogonal transformation into an output digital signal 6 by converting the input digital signal 5 in a region subjected to orthogonal transformation such as a frequency domain. .

That is, as illustrated in FIG. 3, the original time domain (or space domain) signal (original signal A) is transformed into, for example, the frequency domain by the orthogonal transformation matrix Ts _(k) and the frequency signal B ₁ (above and equivalent) to the input digital signal 5, which was a reduction, for example N / L by the signal conversion unit 4 (or enlarged) to the frequency signal B _{2 (corresponding} to the output digital signal 6), the frequency signal B ₂ The signal C in the time domain is obtained by performing inverse orthogonal transform using the inverse transform matrix Td _(L) ^-1 .

  Here, in the example shown in FIG. 3, the one-dimensional original signal A is orthogonally transformed for each transform block of length k, and m blocks adjacent to the obtained transform block in the frequency domain, that is, the length L This shows a case where (= k × m) continuous frequency signals are converted into one block of length N (where N <L), that is, the whole is reduced to N / L.

In the following description, a matrix (orthogonal transformation matrix) in which orthogonal transformation basis vectors < e ₁ , e ₂ ,..., E _n > of length n are arranged in each row is T _(n) , and its inverse transformation matrix is T _{(n )} ^Write as ^-1 . Note that x represents a vector representation of x. At this time, all the matrices are n-order square matrices. As an example, a one-dimensional DCT transformation matrix T ₍₈₎ when n = 8 is shown in the following equation (1).

In FIG. 3, when the size of the orthogonal transform block, that is, the length of the base, is k for the input digital signal 5 that has already been orthogonally transformed into the frequency domain by the orthogonal transformation matrix Ts _(k) , the input orthogonal transformation is performed. The matrix generation unit 1 generates an inverse orthogonal transformation matrix Ts _(k) ^−1, and the output orthogonal transformation matrix generation unit 2 generates an orthogonal transformation matrix Td _(L) whose base length is L (= k × m _). Is generated.

At this time, the inverse orthogonal transformation matrix Ts _(k) ⁻¹ generated by the input orthogonal transformation matrix generation unit 1 corresponds to the orthogonal transformation processing (reverse processing) when the input digital signal 5 is generated, and the output orthogonal transformation. The orthogonal transformation matrix Td _(L) generated by the matrix generation unit 2 is an inverse orthogonal transformation process when the output digital signal transformed by the signal transformation unit 14 is decoded, that is, when it is transformed into the time domain. It is assumed that both of these orthogonal transformation matrix generation units 1 and 2 can generate a base vector having an arbitrary length.

The orthogonal transformation matrix generation units 1 and 2 may be the same orthogonal transformation matrix generation unit. In this case, the orthogonal transformation matrices Ts _(k) and Td _(L) are only the base lengths. Different orthogonal transformation matrices of the same type. The orthogonal transformation matrix generation unit exists for each different orthogonal transformation method.

Next, in the transformation matrix generation unit 3, the inverse orthogonal transformation matrix Ts _(k) ⁻¹ generated by the input orthogonal transformation matrix generation unit 1 is placed diagonally as shown in the following equation (2). An L-th order square matrix A is created by arranging m. When the base length of the output digital signal 6 is N, N low-frequency base vectors of the orthogonal transformation matrix Td _(L) are extracted, and a matrix B composed of N rows and L columns is created.

Here, e ₁ , e ₂ ,..., E _N are obtained by extracting N low-frequency components when Td _(L) is represented by a basis vector as follows.

And
D = α · B · A (5)
To create a matrix D with N rows and L columns. This matrix D becomes a conversion matrix for converting the resolution to the reduction ratio (or enlargement ratio) N / L. Α is a scalar value or vector value, and is a coefficient for level correction or the like.

Divided in the signal conversion unit 4 of FIG. 2, as shown in FIG. 3, the block m input digital signal B ₁ of the frequency domain together, the L size of the metablock (1 metablock = m blocks) To do. If the length of the input digital signal B ₁ is not a multiple of L, due like compensate for signal, for example, such as by the dummy data of 0 or the like filling (stuffing), to be a multiple of L. Let the metablock thus formed be Mi (i = 0, 1, 2,...).

  The principle of the resolution conversion processing described above is described in detail in PCT / JP98 / 02653 filed on June 16, 1998 by the present applicant.

  Next, the digital signal conversion method according to the first embodiment will be described with reference to the configuration of the digital signal conversion apparatus described above.

  4A to 4C schematically show processing when a DV video signal is converted into an MPEG video signal by digital signal conversion according to the embodiment of the present invention. This processing is mainly performed by the resolution conversion unit 16 in the digital signal processing apparatus according to the embodiment of the present invention shown in FIG.

  In the following description, a one-dimensional DCT coefficient block will be described as an example, but the same processing applies to a two-dimensional DCT coefficient.

  First, as shown in FIG. 4A, the low-frequency side DCT coefficients are obtained from adjacent blocks (i) and (i + 1) each composed of eight DCT coefficients of the digital signal of the first format. Take out four at a time. That is, out of the eight DCT coefficients a0, a1, a2, a3,..., A7 of the block (i), only the four DCT coefficients a0, a1, a2, a3 on the low frequency side are extracted, and the DCT coefficients A partial block in which the number of is halved is created. Similarly, out of the eight DCT coefficients b0, b1, b2, b3,..., B7 of the block (i + 1), only the four low-frequency DCT coefficients b0, b1, b2, b3 are extracted, and the DCT A partial block in which the number of coefficients is halved is created. Here, the DCT coefficient on the low frequency side is extracted based on the property that when the video signal is frequency-converted, the energy is concentrated on the low frequency of DC and AC.

  Then, each of the partial blocks each including four DCT coefficients is subjected to 4-point inverse discrete cosine transform (4-point IDCT) to obtain reduced pixel data. These are shown as pixel data p0, p1, p2, p3 and pixel data p4, p5, p6, p7 in FIG. 4B, respectively.

  Next, the partial blocks composed of the reduced pixel data that have been subjected to inverse discrete cosine transform are combined to generate a block having the same size as the original block. That is, the pixel data p0, p1, p2, and p3 and the pixel data p4, p5, p6, and p7 are combined to generate a new block of eight pixel data.

  Then, an 8-point discrete cosine transform (8-point DCT) is applied to the new block consisting of the above eight pixel data, and as shown in FIG. 4C, eight DCT coefficients c0, c1, c2, c3,. .., One block (j) consisting of c7 is generated.

  According to the procedure as described above, the number of orthogonal transform coefficients (DCT coefficients) per predetermined block unit can be reduced to half and converted into video signals of different formats. For example, when it is desired to thin out the number of DCT coefficients to ¼, it can be realized by performing the above-described process twice.

  The resolution conversion process described above can be applied, for example, when converting from the DV format to the MPEG1 format.

  Here, the relationship between the DV format and the MPEG format and the format conversion between them will be described with reference to FIG.

  That is, as shown in FIG. 5, when the video signal is in the NTSC format, the DV format has a resolution of 720 pixels × 480 pixels, and the ratio of the sampling frequency of the luminance signal and the sampling frequency of the two color difference signals is 4: 1. The MPEG1 format is a compressed video signal having a resolution of 360 pixels × 240 pixels and a ratio of the sampling frequency of the luminance signal and the sampling frequency of the two color difference signals to 4: 2: 0. Therefore, in this case, the number of DCT coefficients in the horizontal and vertical directions of the luminance (Y) signal is halved and the DCT in the vertical direction of the color difference (C) signal is obtained by the resolution conversion processing according to the present invention described above. The number of coefficients may be reduced to ¼.

  Note that 4: 2: 0 is represented by representing one value because the odd lines and the even lines are alternately 4: 2: 0 and 4: 0: 2.

  When the video signal is in the PAL format, the DV format is a compressed video in which the resolution is 720 pixels × 576 pixels and the ratio between the sampling frequency of the luminance signal and the sampling frequency of the two color difference signals is 4: 2: 0. The MPEG1 format is a compressed video signal having a resolution of 360 pixels × 288 pixels and a ratio of the sampling frequency of the luminance signal and the sampling frequency of the two color difference signals to 4: 2: 0. Therefore, in this case, the number of DCT coefficients in the horizontal and vertical directions of the Y signal is halved and the number of DCT coefficients in the horizontal and vertical directions of the C signal is obtained by the resolution conversion processing according to the present invention described above. Can be halved.

  Further, the resolution conversion process described above can be applied in the same manner, for example, when converting from the DV format to the MPEG2 format.

  When the video signal is in the NTSC format, the MPEG2 format is a compressed video signal with a resolution of 720 pixels × 480 pixels and a ratio between the sampling frequency of the luminance signal and the sampling frequency of the two color difference signals of 4: 2: 0. is there. Therefore, in this case, conversion processing is not performed on the Y signal, the number of DCT coefficients in the vertical direction of the C signal is halved, and the number of DCT coefficients in the horizontal direction of the C signal is doubled. do it. This enlargement method will be described later.

  When the video signal is in the PAL format, the MPEG2 format is a compressed video with a resolution of 720 pixels × 576 pixels and a ratio between the sampling frequency of the luminance signal and the sampling frequency of the two color difference signals of 4: 2: 0. Signal. Therefore, in this case, it is not necessary to perform conversion processing for both the Y signal and the C signal.

  FIG. 6 shows a basic calculation procedure for the resolution conversion process described above.

  That is, the four DCT coefficients a0, a1, a2, a3 and the four DCT coefficients b0, b1, b2, b3 respectively extracted from two adjacent blocks of the input digital signal of the first format are connected. The block consisting of eight DCT coefficients created in this way includes two inverse discrete cosine transform matrices (IDCT4) each given as a (4 × 4) matrix on the diagonal, and the other components are zero ( An 8 × 8) matrix is multiplied.

  These products are further multiplied by a discrete cosine transform matrix (DCT8) given as an (8 × 8) matrix, and eight DCT coefficients c0, c1, c2, c3,. . , C7, a new block is obtained.

  Here, in the digital signal conversion method according to the present invention, since the resolution conversion processing is performed in the DCT domain (frequency domain), the inverse DCT and DCT before and after that are no longer necessary, and the above two A product of an (8 × 8) matrix diagonally including a (4 × 4) inverse discrete cosine transform matrix (DCT4) and the above-described (8 × 8) discrete cosine transform matrix is obtained in advance as a transform matrix D. Therefore, the amount of calculation can be effectively reduced.

  Next, a process for converting the above-described DV video signal, which is a digital signal of the first format, into an MPEG1 video signal, which is a digital signal of the second format, will be described in more detail.

  The DV format includes a “still mode” and a “motion mode” that can be switched in accordance with the image motion detection result. These modes are discriminated by motion detection before DCT of each (8 × 8) matrix in the video segment, for example, and DCT is performed in one of the modes according to the result. Various methods are conceivable for the motion detection described above, and specifically, there is a method of comparing the sum of absolute values of differences between fields with a predetermined threshold.

  The “still mode” is a basic mode of the DV format, and (8 × 8) DCT is performed on (8 × 8) pixels in the block.

  The (8 × 8) block is composed of one DC component and 63 AC components.

  Further, the “motion mode” is used in order to avoid a reduction in compression efficiency due to energy dispersion due to interlace scanning when DCT is performed while the subject is moving. In this motion mode, the (8 × 8) block is divided into a (4 × 8) block in the first field and a (4 × 8) block in the second field, and the pixel data of each (4 × 8) block. On the other hand, by applying (4 × 8) DCT, it is possible to suppress an increase in high-frequency components in the vertical direction and prevent a reduction in compression rate.

  Each (4 × 8) block is composed of one DC component and 31 AC components.

  As described above, in the DV format, the block configuration differs between the still mode and the motion mode. Therefore, in order to perform the subsequent processing in the same manner, each block (4 × After the DCT of 8), the sum and difference are obtained between the coefficients of the same order in each block to form an (8 × 8) block. By this processing, the motion mode block can be regarded as being composed of one DC component and 63 AC components, similarly to the still mode block.

  By the way, when converting a DV format video signal to an MPEG1 format video signal, the MPEG1 format only handles a video signal of 30 frames / second, and there is no concept of a field, so it is necessary to separate only one field. There is.

  FIG. 7A schematically shows a process of separating fields when converting a DCT coefficient in the “motion mode (2 × 4 × 8 DCT mode)” in the DV format into a DCT coefficient in the MPEG1 format.

  The upper (4 × 8) block 31a in the upper half of the (8 × 8) DCT coefficient block 31 is the sum (A + B) of the first field coefficient and the second field coefficient, and the above (8 × 8) DCT. The lower half (4 × 8) block 31b of the coefficient block 31 is a difference (A−B) between the coefficients of the two fields.

  Therefore, by adding the upper half (4 × 8) block 31a and the lower half (4 × 8) block 31b of the (8 × 8) DCT coefficient block 31, the sum is halved. A (4 × 8) block 35a consisting only of DCT coefficients of the first field (A) can be obtained. Similarly, if the (4 × 8) block 31a and the lower half (4 × 8) block 31b are subtracted and the difference is halved, only the discrete cosine coefficient of the second field (B) is formed ( A 4 × 8) block 35b can be obtained. That is, by the above processing, an (8 × 8) block 35 in which fields are separated can be obtained.

  Then, the above-described resolution conversion processing is performed on any one of these fields, for example, the DCT coefficient of the first field.

  FIG. 7B schematically shows processing for separating fields in the “still mode (8 × 8 DCT mode)”.

  In the (8 × 8) DCT coefficient block 32, the DCT coefficient of the first field (A) and the DCT coefficient of the second field (B) are mixed. Therefore, this can be field-separated by the processing described below to obtain a (4 × 8) block 35a consisting only of the first field (A). Similarly, the (4 × 8) block 31a and the lower half ( If the 4 × 8) block 31b is subtracted, it is necessary to perform a conversion process to obtain a (4 × 8) block 35b consisting only of the second field (B).

  FIG. 8 shows the procedure of the field separation process in the “still mode”.

  First, an input composed of eight DCT coefficients d0, d1, d2, d3,..., D7 is multiplied by an eighth-order inverse discrete cosine transform matrix (IDCT8), and returned to pixel data.

  Next, by multiplying the (8 × 8) matrix for field separation, the upper and lower sides of the (8 × 8) block are divided into the first field and the second field of the (4 × 8) block, respectively.

  Then, an (8 × 8) matrix diagonally including two discrete cosine transform matrices (DCT4) each given as a (4 × 4) matrix is further multiplied.

  As a result, eight DCT coefficients including four DCT coefficients e0, e1, e2, e3 in the first field and four DCT coefficients f0, f1, f2, f3 in the second field are obtained.

  Then, the above-described resolution conversion processing is performed on any one of these fields, for example, the DCT coefficient of the first field.

  Here, in the digital signal conversion method according to the present invention, since the resolution conversion is performed in the DCT domain (frequency domain), the inverse DCT and DCT before and after that become unnecessary, and in addition, By calculating in advance a product of an (8 × 8) matrix diagonally including the (4 × 4) inverse discrete cosine transform matrix (IDCT4) and the above (8 × 8) discrete cosine transform matrix. The amount can be effectively reduced.

  The resolution conversion process described above is for the case of reducing an image, and hereinafter, the resolution conversion process for the case of enlarging an image will be described as a second embodiment.

  9A to 9C schematically show a state in which a DV video signal is converted into an MPEG2 video signal by the digital signal conversion method according to the present invention.

  In the following description, a one-dimensional DCT coefficient will be described as an example, but a two-dimensional DCT coefficient can be similarly processed.

  First, an 8-point inverse discrete cosine transform (8-point IDCT) is performed on a block (u) composed of eight orthogonal coefficients (DCT coefficients g0 to g7) shown in FIG. 9A to obtain eight pixel data (h0). Return to ~ h7).

  Next, a block composed of 8 pixel data is divided into two, and two partial blocks each composed of 4 pixel data are generated.

  Next, a 4-point DCT (4-point DCT) is applied to the two partial blocks each consisting of four DCT coefficients, and two partial blocks (i0 to i0) each consisting of four DCT coefficients. i3 and j0-j3) are generated.

  Then, as shown in FIG. 9C, each of the two partial blocks composed of the four pixel data is filled with 0 as four DCT coefficients on the high frequency side, and each of the blocks is composed of eight DCT coefficients. Generate (v) and block (v + 1).

  With the above procedure, resolution conversion between compressed video signals having different formats is performed in the orthogonal transform region.

  FIG. 10 shows the procedure of the conversion process at this time.

  First, an input composed of eight DCT coefficients g0, g1, g2, g3,..., G7 is multiplied by an eighth-order inverse discrete cosine transform (IDCT) matrix, and returned to eight pixel data.

  Next, a block composed of 8 pixel data is divided into two, and two partial blocks each composed of 4 pixel data are generated.

  Next, for each of the two partial blocks each consisting of four DCT coefficients, a 4-point discrete cosine transform matrix given as a (4 × 4) matrix and a 0 matrix given as a (4 × 4) matrix are By multiplying the upper and lower (4 × 8) matrices, two partial blocks (i0 to i7 and j0 to j7) including eight DCT coefficients are generated.

  By processing in this way, DCT coefficients of two blocks can be obtained from one block, so that the resolution can be expanded in the frequency domain.

  In the case of NTSC, in order to convert the DV format to the MPEG2 format, it is not necessary to convert the luminance signal Y in the horizontal and vertical directions as shown in FIG. In the vertical direction of the color difference signal C, it is necessary to reduce it to ½. Therefore, the enlargement process described above is used for converting the resolution of the color difference signal C in the horizontal direction when converting from the DV format to the MPEG2 format.

  FIG. 11 shows an example of the configuration of the main part of a digital signal conversion apparatus according to the third embodiment of the present invention. The same reference numbers are assigned to the same components as those in the first embodiment. The difference from FIG. 1 is that the weighting unit 18 and the inverse weighting unit 14 are combined into a weighting processing unit 21.

  In other words, the weighting processing (IW * W) unit 21 performs inverse weighting, which is a reverse operation of weighting applied to the DV video signal, which is an input digital signal of the first format, and the second format of the output second format. Weighting for the MPEG video signal which is a digital signal is also performed.

  According to such a configuration, the inverse weighting process for the input video signal of the first format and the weighting process for the output video signal of the second format can be performed together. The amount of calculation can be reduced as compared with the case where the weighting process and the weighting process are performed separately.

  In the digital signal conversion apparatus according to the third embodiment illustrated in FIG. 11, the weighting processing unit 21 is arranged after the resolution conversion unit 16, but the weighting processing unit is arranged before the resolution conversion unit 16. You may make it do.

  FIG. 12 shows a digital signal conversion apparatus according to the fourth embodiment of the present invention, in which the weighting processing unit 22 is arranged in front of the resolution conversion unit 16. The configuration of each part of the digital signal processing apparatus shown in this figure can be the same as that of each part of the digital signal conversion apparatus shown in FIG.

  Here, the weighting process in which the inverse weighting for the digital signal of the first format and the weighting for the second digital signal are combined, and the above weighting process is performed before or after the orthogonal transformation such as discrete cosine transform (DCT). What can be done is that these arithmetic operations are linear operations.

  Hereinafter, an embodiment of a digital signal conversion method and apparatus according to a fifth embodiment of the present invention will be described with reference to FIG.

  As shown in FIG. 13, the digital video signal conversion apparatus includes a decoding unit 8 that decodes the DV video signal, and a resolution conversion unit 16 that performs resolution conversion processing for format conversion on the decoded output from the decoding unit 8. A determination unit 7 that determines whether or not to perform forward interframe differential encoding for each predetermined block unit of the conversion output from the resolution conversion unit 16 according to the motion mode / still mode information, and the determination unit 7 is provided with an encoding unit 9 that encodes the conversion output from the resolution conversion unit 16 based on the determination result from 7 and outputs the MPEG video signal.

  In the following, a digital video signal conversion apparatus constituted by each of these units will be described. Of course, each component performs processing of each step of the digital signal conversion method according to the present invention.

  In the DV video signal input to the digital video signal converter, a mode flag (for example, 1 bit) which is information indicating the still mode / motion mode is added in advance to each DCT block.

  In this digital video signal conversion apparatus, based on the mode flag, the determination unit 7 determines whether or not to perform forward interframe difference encoding for each predetermined block unit of the conversion output from the resolution conversion unit 16. Details of this operation will be described later.

  The deframing unit 11 extracts a mode flag indicating the still mode / motion mode and supplies the mode flag to the determination unit 7.

  The deshuffling unit 15 solves the shuffling performed in order to equalize the amount of information in the video segment which is a unit of fixed length on the DV encoding side.

  The determination unit 7 includes an adder 27 and an I (I picture) / P (P picture) determination unit & determination unit 28. The adder 27 adds a reference DCT coefficient stored in a later-described frame memory (FM) unit 24 as a negative value to the resolution conversion output. A mode flag indicating the still mode / motion mode from the deframing unit 11 is also supplied to the I / P determination & determination unit 28 to which the addition output from the adder 27 is supplied.

  Details of the operation of the I / P determination & determination unit 28 will be described. The conversion output from the resolution conversion unit 16 is in units of 8 × 8 DCT coefficients. The 8 × 8 DCT coefficient blocks are divided into four for the luminance signal and two for the chrominance signal, and a total of six DCT coefficient blocks constitute one predetermined block. This predetermined block is called a macro block.

  By the way, it is assumed that the P picture simply takes a difference from the previous frame. In the case of a still image, if the difference is taken, the amount of information decreases. On the other hand, in the case of a moving image, if the difference is taken, the amount of information increases. For this reason, if the mode flag indicating the still mode / motion mode is viewed and it is determined that it is moving, the amount of information increases. Efficient coding can be performed by adopting P as a P picture.

  For example, when all the mode flags sent from the deframing for the six DCT coefficient blocks indicate the motion mode, the I / P determination & determination unit 28 sets the macro block as an I picture. Further, for example, when only one flag indicating the motion mode can be detected in the six DCT blocks, the macro block is set as a P picture.

  Further, if a motion flag is added to four or more DCT blocks among the six DCT blocks, the macro block may be an I picture. Further, the macroblock may be a P picture when flags indicating the still mode are added to all of the six DCT blocks.

  Each DCT coefficient in units of macroblocks determined as an I / P picture by the I / P determination & determination unit 28 is supplied to the encoding unit 9.

  The encoding unit 9 includes a weighting (W) unit 18, a quantization (Q) unit 19, an inverse quantization (IQ) unit 26, an inverse weighting (IW) unit 25, an FM unit 24, and a variable length code. (VLC) unit 20, buffer memory 23, and rate control unit 29.

  The weighting (W) unit 18 weights the DCT coefficient that is the conversion output from the conversion unit 16 via the determination unit 7.

  The quantization (Q) unit 19 quantizes the DCT coefficients weighted by the weighting (W) unit 18. Then, the variable length coding (VLC) unit 20 performs variable length coding on the DCT coefficient quantized by the quantization unit 19 and supplies it to the buffer memory 23 as MPEG encoded data.

  The buffer memory 23 keeps the transfer rate of the MPEG encoded data constant and outputs it as a bit stream. The rate control unit 29 controls the increase / decrease of the information generation amount in the quantization (Q) unit 19, that is, the quantization step, based on the change information of the increase / decrease of the buffer capacity in the buffer memory 23.

  The inverse quantization (IQ) unit 26 inversely quantizes the quantized DCT coefficient from the quantization (Q) unit 19 and supplies the quantized DCT coefficient to the inverse weighting (IW) unit 25. The inverse weighting (IW) unit 25 performs inverse weighting, which is an inverse operation of weighting, on the DCT coefficient from the inverse quantization (IQ) unit 26. The DCT coefficient inversely weighted by the inverse weighting (IW) unit 25 is stored in the FM unit 24 as a reference DCT coefficient.

  As described above, in the digital video signal conversion apparatus shown in FIG. 13, the determination unit 7 performs the I / P determination & response according to the mode flag indicating the motion mode / still mode sent from the deframing unit 11. Since the determination unit 28 is used to determine the I or P picture for each macroblock, it is possible to convert a DV signal originally consisting only of the I picture into an MPEG picture using the I picture or P picture. It is possible to take advantage of the improvement in compression rate, which is a characteristic of the signal.

  Next, a digital signal conversion method and apparatus according to a sixth embodiment of the present invention will be described.

  The digital video signal conversion apparatus according to the sixth embodiment is a digital video signal conversion apparatus having a configuration in which the determination unit 7 shown in FIG. 13 is replaced with a determination unit 30 shown in FIG.

  That is, a decoding unit 8 that partially decodes the DV signal to obtain an orthogonal transform domain signal, for example, a DCT coefficient, and a conversion that performs a signal conversion process for format conversion on the DCT coefficient from the decoding unit 8 A determination to determine whether or not to perform forward interframe difference encoding for each predetermined block unit of the conversion output from the conversion unit 16 according to the maximum absolute value of the interframe difference of the conversion output Unit 30 and an encoding unit 9 that encodes the conversion output from the conversion unit 16 based on the determination result from the determination unit 30 and outputs the MPEG video signal.

  The determination unit 30 looks at the maximum value of the absolute value of the AC coefficient when the difference between the converted DCT coefficient, which is the conversion output from the conversion unit 16, and the reference DCT coefficient from the FM unit 24 is taken. A predetermined threshold value is compared, and an I / P picture is assigned to each macroblock based on the comparison result.

  The determination unit 30 includes a difference calculation unit 31, a maximum value detection unit 32, a comparison unit 33, and an I / P determination unit 35.

  The difference calculating unit 31 calculates a difference between the converted DCT coefficient from the converting unit 16 and the reference DCT coefficient from the FM unit 24. The difference output from the difference calculation unit 31 is supplied to the maximum value detection unit 32 and also to the I / P determination unit 35.

  The maximum value detector 32 detects the maximum absolute value of the AC coefficient of the difference output. Basically, when the amount of information converted to the DCT coefficient is large, the AC coefficient becomes large, while when the amount of information is small, the AC coefficient becomes small.

  The comparison unit 33 compares the maximum absolute value from the maximum value detection unit 32 with a predetermined threshold supplied from the terminal 34. If the predetermined threshold is appropriately selected, it can be determined whether the amount of information converted into the DCT coefficient is large or small based on the magnitude of the maximum absolute value of the AC coefficient.

  The I / P determination unit 35 uses the comparison result from the comparison unit 33 to determine whether the difference in DCT coefficients from the difference calculation unit 31, that is, the difference in information amount is large or small, When it is determined that the difference is large, a macro block composed of the converted DCT coefficient block from the conversion unit 16 is allocated to the I picture, and when it is determined that the difference is small, the macro block from the difference calculation unit 31 is allocated to the P picture.

  That is, if the absolute value of the maximum value is larger than the threshold value, it is determined that the information amount of the difference is large, and the macroblock is set as an I picture. If the absolute value of the maximum value is smaller than the threshold value, it is determined that the information amount of the difference is small, and the macroblock is set as a P picture.

  As a result, the digital video signal conversion apparatus according to the sixth embodiment can also convert a DV signal originally composed of only an I picture into an MPEG picture using an I picture or a P picture. It is possible to take advantage of the improvement of the compression ratio, which is a feature of.

  In the digital video signal conversion apparatus shown in FIGS. 13 and 14, the NTSC DV signal and the MPEG1 video signal are input and output, but may be applied to each PAL signal.

  Further, the resolution conversion process described above can be applied in the same manner, for example, when converting from the DV format to the MPEG2 format.

  Further, as the resolution conversion processing by the conversion unit 16, an example in which resolution conversion is performed mainly in the direction of reduction has been described, but enlargement is also possible. That is, generally, by adding a high frequency component to an input digital signal in the frequency domain, the resolution can be enlarged at an arbitrary magnification.

  For example, when an MPEG2 video signal is applied to a digital broadcasting service, the signal is classified by profile (function) / level (resolution). For example, a main profile / high level used in a digital HDTV in the United States. For example, when the DV signal is converted into a video signal of (MP @ HL), enlargement of the resolution can be applied.

  Further, the processing of the sixth embodiment may be performed by software.

Next, a digital signal conversion method and apparatus according to the seventh embodiment of the present invention will be described with reference to FIG. The same reference numbers are assigned to the same components as those in the above-described embodiment.
The rate control unit 40 performs data amount control in the quantization unit 19 based on the quantizer number (Q_NO) and the class number (Class) from the deframing unit 11.

  FIG. 16 is a diagram in which a quantization scale is set for each macroblock (MB) of each frame when a DV video signal is converted into an MPEG video signal by the digital signal conversion method in the seventh embodiment. The procedure is shown.

  In step S1, first, a quantization number (Q_NO) and a class number (Class) are acquired for each macroblock. This quantization number (Q_NO) is indicated by a value from 0 to 15, and is common to all six DCT blocks in the macroblock. The class number (Class) is indicated by a value from 0 to 3, and is given for every six DCT blocks.

    Next, in step S2, a quantization parameter (q_param) is calculated for each DCT block by the following procedure.

Quantization table q_table [4] = {9, 6, 3, 0}
Quantization parameter q_param = Q_NO + q_table [class]

  That is, the quantization table has four values (9, 6, 3, 0), and each value corresponds to class numbers 0, 1, 2, and 3. For example, when the class number is 2 and the quantizer number is 8, the quantization table value 3 corresponding to the class number 2 and the quantizer number 8 are added, and the quantization parameter becomes 11.

  Next, in step S3, the average of the quantization parameters (q_param) of the six DCT blocks in the macroblock is calculated.

  In step S4, the quantization scale (quantizer_scale) of the MPEG macroblock is obtained by the following procedure, and the process ends.

Quantization table q_table [25]
= {32, 16, 16, 16, 16, 8, 8, 8, 8, 4,
4, 4, 2, 2, 2, 2, 2, 2, 2, 2,
2,2,2,2}
quantizer_scale = q_table [q_pram]

  That is, the quantization table has 25 values (32 to 2), and each value corresponds to the quantization parameter calculated as described above. That is, the quantization table corresponding to the quantization parameter value 0 is 32, the quantization table corresponding to the quantization parameter value 1 is 16, and the quantization table corresponding to the quantization parameter value 5 is 8. For example, when the average value of the quantization parameters obtained as described above is 10, a value of 4 corresponding to the quantization parameter value 10 is the quantization scale value. Through the above procedure, the MPEG quantization scale (quantizer_scale) depending on the target rate is calculated in each frame based on the quantization parameter (q_param) for each macroblock. The correspondence relationship between the class number and the quantization table and the relationship between the quantization parameter and the quantization table are determined empirically.

  In the digital signal conversion apparatus according to the present invention illustrated in FIG. 15, the above processing is performed by the rate control unit 40 based on the quantization number (Q_NO) and class number (Class) sent from the deframing unit 11. It is what is said.

  FIG. 17 shows a basic procedure for applying feedback to the next frame using the quantization scale set by the above-described procedure.

  In step S11, first, the number of target bits per frame at the bit rate set by the above-described procedure is set.

  Next, in step S12, the total number of generated bits per frame is integrated.

  Next, in step S13, a difference (diff) between the number of target bits and the total number of generated bits is calculated.

  In step S14, the quantization scale is adjusted based on the calculation result.

  The calculation in each of the above steps is expressed as follows.

diff = cont * diff (cont: constant)
q_param = q_param ± f (diff)
quantizer_scale = q_table [q_param]

  That is, normalization is performed by multiplying the difference value diff obtained in step S13 by the constant cont. The normalized difference value is multiplied by an empirically determined function, and the result obtained by adding and subtracting the quantization parameter is used as the quantization parameter. A value corresponding to the quantization parameter value is selected from the above-described quantization table having 25 values, and used as the quantization scale of the next frame.

  Through the above procedure, a new quantization scale (quantizer_scale) is calculated based on the adjusted quantization parameter (q_param), and feedback is performed between frames using it for the next frame.

  Next, as an eighth embodiment, a digital signal conversion method and digital signal conversion apparatus according to the present invention will be described. In the above-described embodiment, an example in which the DV format is converted into the MPEG format has been shown. In the following embodiment, an example in which the MPEG format is converted into the DV format will be described.

  First, a conventional apparatus for converting from MPEG format to DV format will be described with reference to FIG.

  The digital video signal conversion apparatus shown in FIG. 18 includes an MPEG decoder 70 that decodes MPEG2 video data and a DV encoder 80 that outputs DV video data.

  In the MPEG decoder 70, a parser 71 to which the bit stream of the MPEG2 video data is supplied detects the header of the bit stream of the quantized DCT coefficient that has been framed according to the MPEG2 format, and performs variable length coding. The quantized DCT coefficients are supplied to a variable length decoding (VLD) unit 72, and a motion vector (mv) is extracted and supplied to a motion compensation (MC) unit 77.

  The variable length decoding (VLD) unit 72 performs variable length decoding on the quantized DCT coefficient that has been subjected to variable length coding, and supplies the quantized DCT coefficient to the inverse quantization (IQ) unit 73.

  The inverse quantization unit 73 multiplies the quantized DCT coefficient decoded by the variable length decoding unit 72 by the quantization step used on the encoding side to perform an inverse quantization process to obtain a DCT coefficient, A discrete cosine transform (IDCT) unit 74 is supplied.

  The inverse discrete cosine transform unit 74 performs inverse DCT on the DCT coefficient from the inverse quantization unit 73 to return the DCT coefficient to spatial domain data, that is, pixel data. Specifically, pixel values (luminance Y, color difference Cr, Cb) are calculated for each 8 × 8 pixel block by inverse DCT. However, the pixel value here is a value of the actual pixel value itself in the I picture, but is a difference value between corresponding pixel values in the P picture and the B picture.

  The motion compensation unit 77 generates a motion compensation output using the image information stored in the two frame memories FM of the frame memory (FM) unit 76 and the motion vector mv extracted by the parser 71, and this motion compensation output Is supplied to the adder 75.

  The adder 75 adds the motion compensation output to the difference value from the inverse discrete cosine transform unit 74 and supplies the decoded image data to the discrete cosine transform (DCT) unit 81 and the frame memory unit 76 of the DV encoder 80.

  In the DV encoder 80, the discrete cosine transform unit 81 performs DCT processing on the decoded image data, transforms it again into orthogonal transform domain data, that is, DCT coefficients, and supplies it to the quantization (Q) unit 82.

  The quantization unit 82 quantizes the DCT coefficient using a matrix table that takes visual characteristics into consideration, and supplies it to the variable length coding (VLC) unit 83 as an I picture in the DV format.

  The variable length encoding unit 83 performs variable length encoding processing on the DV format I picture, compresses the I picture, and supplies the compressed I picture to the framing unit 84.

  The framing unit 84 frames the DV format data that has been subjected to the variable-length encoding process, and outputs it as a bit stream of DV video data.

  By the way, since orthogonal transformation such as discrete cosine transformation (DCT) and its inverse transformation usually require a large amount of calculation, there is a problem that the format conversion of video data as described above cannot be performed efficiently. . There is also a problem that the signal deteriorates because errors are accumulated as the amount of calculation increases.

  Therefore, a digital video signal converter for solving this problem will be described as an eighth embodiment with reference to FIG.

  The digital signal conversion apparatus shown in FIG. 19 receives an MPEG video signal conforming to the above-described MPEG format as a first digital signal and outputs a DV signal as a second digital signal.

  The parser 111 extracts image motion information such as a motion vector mv and a quantization scale with reference to a header of an MPEG video signal which is a digital signal of the first format input as a bit stream.

  The motion vector mv is sent to the motion compensation (MC) unit 115 for motion compensation. The quantization scale (quantizer_scale) is sent to the evaluation unit 123 described later.

  A variable length decoding (VLD) unit 112 performs variable length decoding on the bit stream of the MPEG video signal from which necessary information is extracted by the parser 111.

  The inverse quantization (IQ) unit 113 performs inverse quantization on the MPEG video signal decoded by the variable length decoding unit 112.

  Then, the MPEG video signal inversely quantized by the inverse quantization unit 113 is input to the addition unit 125. The result of motion compensation for the motion vector mv from the parser 111 is also input to the adder 125 from the motion compensation unit 115.

  The output from the adder 125 is sent to the signal converter 116 described later, and is also input to the motion compensator 115 via the frame memory 114. In the signal conversion unit 116, the video signal input through the addition unit 125 is subjected to necessary signal conversion processing such as resolution conversion in the orthogonal transform region (frequency region).

  Then, the video signal that has undergone the required signal conversion processing by the signal conversion unit 116 is shuffled by the shuffling unit 117 and sent to the buffer 118 and the classifying unit 122.

  The video signal sent to the buffer 118 is sent to the quantization (Q) unit 119, where it is quantized, subjected to variable length coding by the variable length coding (VLC) unit 120, and further framed by the framing unit 121. It is output as a bit stream of a DV video signal.

  On the other hand, the classification unit 122 classifies the video signals shuffled by the shuffling unit 117 and sends the result to the evaluation unit 123 as class information.

  The evaluation unit 123 determines the quantization number in the quantization unit 119 based on the class information from the classification unit 122 and the quantization scale (quantizer_scale) from the parser 111.

  According to such a configuration, the data amount of the DV video signal output as the video signal of the second format is based on the data amount information included in the MPEG video signal input as the video signal of the first format. Therefore, the processing for determining the data amount of the video signal of the second format generated by performing the signal conversion can be simplified.

  In the seventh and eighth embodiments described above, for example, one of the first format digital signal and the second format digital signal is an MPEG1 video signal, and the other is an MPEG2 video signal. It can also be applied in some cases.

  Next, a digital signal conversion method and a digital signal conversion apparatus according to the present invention will be described as a ninth embodiment with reference to FIG.

  This is a digital video signal converter for converting MPEG video data in accordance with the MPEG2 format into DV video data in accordance with the DV format, both of which assume PAL format data.

  When the video signal is a PAL system, the MPEG2 format and the DV format are compressed with a resolution of 720 pixels × 576 pixels and a ratio of the sampling frequency of the luminance signal and the sampling frequency of the two color difference signals to 4: 2: 0. Since it is a video signal, it is not necessary to perform resolution conversion processing for both the Y signal and the C signal.

  In FIG. 20, an MPEG decoder 100 includes a parser 111, a variable length decoding (VLD) unit 112, an inverse quantization (IQ) unit 113, an adder 125, an inverse discrete cosine (IDCT) unit 131, A frame memory (FM) unit 132, a motion compensation (MC) unit 115, and a discrete cosine transform (DCT) unit 130. Here, the frame memory FM unit 132 is configured to be used as two prediction memories.

  Among these, although details will be described later, the inverse discrete cosine transform unit 131 performs inverse discrete cosine transform processing on the I picture and P picture partially decoded by the variable length decoding unit 112 and the inverse quantization unit 113. The motion compensation unit 115 generates a motion compensation output based on the inverse discrete cosine transform output. The discrete cosine transform unit 130 performs discrete cosine transform on the motion compensation output. The adder 125 adds the motion compensation output from the discrete cosine transform unit 130 to the P picture and B picture partially decoded by the variable length decoding unit 112 and the inverse quantization unit 113.

  The overall operation will be described below. First, the parser 111 refers to the header of the MPEG2 video data input as a bit stream, returns the quantized DCT coefficient that has been framed in accordance with the MPEG2 format to a variable length code, and a variable length decoding unit 112. And a motion vector (mv) is extracted and supplied to the motion compensation unit 115.

  The variable length decoding unit 112 performs variable length decoding on the quantized DCT coefficient returned to the variable length code, and supplies it to the inverse quantization unit 113.

  The inverse quantization unit 113 multiplies the quantized DCT coefficient decoded by the variable length decoding unit 112 by the quantization step used on the encoding side to perform an inverse quantization process to obtain a DCT coefficient, This is supplied to the adder 125. The DCT coefficients obtained by the variable length decoding unit 112 and the inverse quantization unit 113 are supplied to the adder 125 as an output that is not subjected to inverse discrete cosine transform and returned to the pixel data, that is, partially decoded data. Is done.

  The adder 125 is also supplied with a motion compensation output from the motion compensation unit 115 orthogonally transformed by the discrete cosine transform unit 130. The adder 125 adds the motion compensation output to the partially decoded data in the orthogonal transform domain, and supplies the addition output to the DV encoder 110 and also to the inverse discrete cosine transform unit 131.

  The inverse discrete cosine transform unit 131 performs an inverse discrete cosine transform process on the I picture and the P picture in the addition output to obtain spatial domain data. The data in this spatial region becomes reference image data used for motion compensation. The reference image data for motion compensation is stored in the frame memory unit 132.

  Then, the motion compensation unit 115 generates a motion compensation output by using the reference image data stored in the frame memory unit 132 and the motion vector mv extracted by the parser 111, and outputs the motion compensation output to the discrete cosine transform unit 130. Supply.

  The discrete cosine transform unit 130 returns the motion compensated output processed in the spatial domain to the orthogonal transform domain again as described above and then supplies it to the adder 125.

  The adder 125 adds the DCT coefficient of the motion compensation output from the discrete cosine transform unit 130 to the DCT coefficient of the differential signal of P and B pictures partially decoded from the inverse quantization unit 113. The added output from the adder 125 is supplied to the DV encoder 110 and the inverse discrete cosine transform unit 131 as partially decoded data in the orthogonal transform domain.

  Since the partially decoded I picture from the inverse quantization unit 113 is an intra-frame encoded image signal, no motion compensation addition process is required, and it is supplied to the inverse discrete cosine transform unit 131 as it is. At the same time, it is also supplied to the DV encoder 110.

  The DV encoder 110 includes a quantization (Q) unit 141, a variable length coding (VLC) unit 142, and a framing unit 143.

  The quantization unit 141 quantizes the decoded output of the I-picture, P-picture, and B-picture as they are from the MPEG decoder 100, that is, the DCT coefficient, and supplies it to the variable-length encoding unit 142.

  The variable length coding unit 142 performs variable length coding processing on the quantized DCT coefficient and supplies the result to the framing unit 143. The framing unit 143 frames the compressed encoded data from the variable length encoding unit 142 and outputs it as a bit stream of DV video data.

  As described above, when the MPEG2 video data to be converted is an I picture, the MPEG decoder 100 partially decodes the MPEG2 video data to the orthogonal transform region by the variable length decoding unit 112 and the inverse quantization unit 113, and the DV encoder 110 The quantization unit 141 and the variable length coding unit 142 partially code. At the same time, an inverse discrete cosine transform unit 131 performs inverse discrete cosine transform on the I picture and stores it in the frame memory unit 132 in order to obtain a P / B picture reference image.

  When the P picture and the B picture are to be transformed, as described above, only the process for generating the motion compensation output is performed in the spatial domain using the inverse discrete cosine transform unit 131, and the variable length decoding unit 112 and the inverse quantization unit are performed. In addition to the differential signals that are partly decoded P pictures and B pictures in step 113, the parts constituting the frame are performed in the discrete cosine transform region by the discrete cosine transform unit 130. Then, partial encoding is performed by the DV encoder 110.

  In particular, in the case of a P picture, the macro block at the position indicated by the motion vector mv is brought from the I picture subjected to inverse discrete cosine transform by the inverse discrete cosine transform unit 131 by motion compensation processing by the motion compensation unit 115. Discrete cosine transform processing is performed on the macroblock by the discrete cosine transform unit 130, and the DCT coefficient of the P picture, which is a differential signal, is added using the adder 125 in the discrete cosine transform region. This is based on the fact that the result obtained by performing discrete cosine transform on the addition result in the spatial domain is equivalent to the result obtained by adding the discrete cosine transform. Then, the result is partially encoded by the DV encoder 110. At the same time, the inverse discrete cosine transform unit 131 performs inverse discrete cosine transform on the addition output from the adder 125 and stores it in the frame memory unit 132 for reference to the next B picture.

  In the case of a B picture, the macroblock at the position indicated by the motion vector mv is brought from the P picture that has been subjected to inverse discrete cosine transform by the inverse discrete cosine transform unit 131. Then, the discrete cosine transform unit 130 performs a discrete cosine transform on the macroblock, and adds a B picture DCT coefficient that is a difference signal in the discrete cosine transform region. Here, in the case of bidirectional communication, the average is obtained from two reference frames.

  The result is partially encoded by the DV encoder 110. Since the B picture does not become a reference frame, the inverse discrete cosine transform unit 131 does not perform the inverse discrete cosine transform.

  According to the ninth embodiment as described above, in order to decode an I picture, both inverse discrete cosine transform (IDCT) and discrete cosine transform (DCT) processing are conventionally required. In the digital video signal conversion apparatus of the form, the IDCT is only required for reference.

  In addition, to decode a P picture, DCT and reference IDCT processing are required. However, in order to decode a B picture, both DCT and IDCT are conventionally required, and only DCT is used for reference. IDCT is not required.

For example, in the case of general MPEG2 data, for example, the number of GOPs N = 15 and the picture interval M = 3 for forward prediction, there are 1 I picture, 4 P pictures, and 10 B pictures. . Assuming that the calculation amount of DCT and IDCT is almost the same, the MPEG2 data per 15 frames is conventionally weighted,
2 × DCT × (1/15) + 2 × DCT × (4/15)
+ 2 × DCT × (10/15) = 2 × DCT
On the other hand, in the digital video signal converter shown in FIG.
1 × DCT × (1/15) + 2 × DCT × (4/15)
+ 1 * DCT * (10/15) = 1.2666 * DCT
Thus, the amount of calculation can be greatly reduced. DCT in this equation indicates a calculation amount.

  That is, the digital video signal conversion apparatus shown in FIG. 20 can greatly reduce the amount of data calculation processing for converting the format from MPEG2 video data to DV video data.

  Next, another embodiment of the digital video signal converter according to the tenth embodiment will be described with reference to FIG.

  The tenth embodiment is also a digital video signal converter for converting MPEG video data according to the MPEG2 format into DV video data according to the DV format. The MPEG2 video data has a high resolution, for example, 1440 pixels × A compressed video signal of 1080 pixels is assumed.

  For example, when an MPEG2 video signal is applied to a digital broadcasting service, the signal is classified by profile (function) / level (resolution). For example, a main profile / high level used in a digital HDTV in the United States. The video signal of (MP @ HL) has a high resolution as described above, and this is a case where it is converted into the DV video data.

  Therefore, the digital video signal conversion apparatus shown in FIG. 21 includes a signal conversion unit 140 for performing the conversion process between the MPEG decoder 100 and the DV encoder 110 shown in FIG.

  The signal conversion unit 140 includes an inverse orthogonal transformation matrix corresponding to the orthogonal transformation matrix used in the DCT encoding applied to the MPEG encoded data, and an IDCT code for obtaining a signal conversion output signal in the time domain. Resolution conversion processing is performed on the DCT coefficients in the DCT conversion region from the MPEG decoder 100 using a transformation matrix generated based on the orthogonal transformation matrix corresponding to the inverse orthogonal transformation matrix used for conversion.

  The DCT coefficient that is the resolution conversion output from the signal conversion unit 140 is supplied to the DV encoder 110.

  The DV encoder 110 performs quantization and variable length coding on the DCT coefficient of the resolution conversion output, and after framing, outputs the bit stream of the DV video data.

  As described above, this digital video signal conversion apparatus converts the resolution of the video signal of the main profile / high level (MP @ HL) in the MPEG2 video signal by the signal conversion unit 140 and then encodes it by the DV encoder. DV video data.

  At this time, as in the digital video signal conversion apparatus shown in FIG. 20, the I picture conventionally requires both IDCT and DCT processing, but the digital video signal of the tenth embodiment is used. The conversion device only performs IDCT for reference.

  For P picture, IDCT is applied for DCT and reference, which is the same as before, but for B picture, IDCT for reference is unnecessary only for DCT, compared to DCT and IDCT. It is.

  That is, the digital video signal conversion apparatus shown in FIG. 21 can also greatly reduce the amount of data calculation processing for converting the format from high-resolution MPEG2 video data to DV video data.

  Note that, as the resolution conversion processing by the signal conversion unit 140, an example of performing resolution conversion mainly in the direction of reduction has been described, but enlargement is also possible. That is, generally, by adding a high frequency component to an input digital signal in the frequency domain, the resolution can be enlarged at an arbitrary magnification. For example, the MPEG1 video data is converted into the DV video data.

  Further, the above processing may be performed by software.

  By the way, in the above-mentioned MPEG format and DV format compression methods, in order to efficiently compress and encode still image data and moving image data, a hybrid compression encoding method combining orthogonal transform encoding and predictive encoding is used. It is used.

  By the way, when the input information signal compression-encoded by the hybrid compression encoding method is subjected to resolution conversion processing, when performing orthogonal transformation again and predictive encoding with motion compensation again, re-predictive encoding processing The motion vector must also be estimated in the process for performing.

  If re-predictive coding is performed at exactly the same resolution without performing resolution conversion processing, the motion vector at the time of predictive coding may be used. However, since conversion distortion changes when the resolution is converted, the corresponding amount This is because the motion vector used in the re-predictive encoding process also changes.

  Thus, in the re-predictive encoding step, it is necessary to estimate a motion vector, but the amount of calculation for estimating the motion vector has become very large.

  The digital signal conversion apparatus according to the eleventh embodiment solves this problem. The digital signal conversion method and apparatus according to the eleventh embodiment can be applied to an input information signal compression-coded by hybrid compression coding that combines orthogonal transform coding and predictive coding in the time domain or orthogonal transform domain. For example, signal conversion processing such as resolution conversion is performed, and the signal is returned to the orthogonal transform region again or recompressed and encoded in the orthogonal transform region.

  As a specific example of the hybrid compression coding, H.264 recommended by ITU-T (International Telecommunication Union-Telecommunication Standardization Sector). 261 and H.264. H.263, and encoding standards such as MPEG and DV.

  H. 261 is a video coding standard for low bit rates, and was developed mainly for video conferences and videophones using ISDN. H. H.263 for the GSTN videophone system. This is an encoding method improved from H.261.

  The eleventh embodiment will be described below with reference to FIG. In this embodiment, MPEG encoded data in accordance with the MPEG format is input, a digital video signal is output as resolution-converted MPEG encoded data after performing resolution conversion processing on the MPEG encoded data as signal conversion processing. It is a conversion device.

  As shown in FIG. 22, this digital video signal conversion apparatus performs decoding using a motion compensation MC on a bit stream of MPEG encoded data that has been compression encoded with motion vector (mv) detection. 210, a resolution conversion unit 160 that performs resolution conversion processing on the decoded output from the decoding unit 210, and a motion vector mv added to the MPEG encoded data in the converted output image from the resolution conversion unit 160 And an encoding unit 220 that performs a compression encoding process with motion detection based thereon and outputs a bit stream of video encoded data whose resolution has been converted.

  In the following, a digital video signal conversion apparatus constituted by each of these units will be described. Of course, each component performs processing of each step of the digital signal conversion method according to the present invention.

  The decoding unit 210 includes a variable length decoding (VLD) unit 112, an inverse quantization (IQ) unit 113, an inverse discrete cosine transform (IDCT) unit 150, an adder 151, a motion compensation (MC) unit 152, And a frame memory (FM) unit 153. Here, the FM unit 153 includes a frame memory FM used as two prediction memories.

  The VLD unit 112 decodes the MPEG encoded data, that is, the encoded data in which the motion vector as the additional information and the quantized DCT coefficient are variable-length encoded according to the variable-length encoding, and the motion vector mv Extract. The IQ unit 113 multiplies the quantized DCT coefficient decoded by the VLD unit 112 by the quantization step used on the encoding side, and performs an inverse quantization process to obtain a DCT coefficient.

  The IDCT unit 150 performs inverse DCT on the DCT coefficient from the IQ unit 113 to return the DCT coefficient to data in the spatial domain, that is, pixel data. Specifically, pixel values (luminance Y, color difference Cr, Cb) are calculated for each 8 × 8 pixel block by inverse DCT. However, the pixel value here is a value of the actual pixel value itself in the I picture, but is a difference value between corresponding pixel values in the P picture and the B picture.

  The MC unit 152 performs a motion compensation process on the image information stored in the two FMs of the FM unit 153 using the motion vector mv extracted by the VLD unit 112, and supplies the motion compensation output to the adder 151. .

  The adder 151 adds the motion compensation output from the MC unit 152 to the difference value from the IDCT unit 150, and outputs a decoded image signal. The resolution conversion unit 160 performs a required resolution conversion process on the decoded image signal. The conversion output from the resolution conversion unit 160 is supplied to the encoding unit 220.

  The encoding unit 220 includes a scale conversion unit 171, a motion estimation ME unit 172, an adder 173, a DCT unit 175, a rate control unit 183, a quantization (Q) unit 176, and a variable length encoding (VLC). ) Section 177, buffer memory 178, IQ section 179, IDCT section 180, adder 181, FM section 182 and MC section 174.

  The scale conversion unit 171 performs scale conversion on the motion vector mv extracted by the VLD unit 112 according to the resolution conversion rate used by the resolution conversion unit 160. For example, when the resolution conversion rate in the resolution conversion unit 160 is ½, the scale conversion is performed to ½ of the motion vector mv.

  The ME unit 172 uses the scale conversion information from the scale conversion unit 171 to search for a narrow range of the conversion output from the resolution conversion unit 160, thereby estimating an optimal motion vector at the converted resolution.

  The motion vector estimated by the ME unit 172 is used at the time of motion compensation by the MC unit 174. The converted output image from the resolution converting unit 160 used when the ME unit 172 estimates the motion vector is supplied to the adder 173.

  The adder 173 takes a difference between a reference image (to be described later) and a conversion output from the resolution conversion unit 160 and supplies the difference to the DCT unit 175.

  The DCT unit 175 performs a discrete cosine transform on the difference between the reference image obtained by motion compensation in the MC unit 174 and the converted output image with an 8 × 8 block size. Since the I picture is intra-frame (frame) coding, the DCT calculation is performed as it is without taking the difference between the frames.

  The quantization (Q) unit 176 quantizes the DCT coefficient from the DCT unit 175 using a matrix table in consideration of visual characteristics. The VLC unit 177 compresses the quantized DCT coefficient from the Q unit 176 by variable length coding.

  The buffer memory 178 is a memory for making the transfer rate of the encoded data compressed by the variable length encoding by the VLC unit 177 constant. Video encoded data whose resolution has been converted is output from the buffer memory 178 as a bit stream at a constant transfer rate.

  The rate control unit 183 controls the increase / decrease of the information generation amount in the Q unit 176, that is, the quantization step, according to the change information of the increase / decrease of the buffer capacity in the buffer memory 178.

  IQ section 179 constitutes a local decoding section together with IDCT section 180, dequantizes the quantized DCT coefficients from Q section 176, and supplies the DCT coefficients to IDCT section 180. The IDCT unit 180 performs inverse DCT conversion on the DCT coefficient from the IQ unit 179, returns it to pixel data, and supplies it to the adder 181.

  The adder 181 adds the motion compensation output from the MC unit 174 to the pixel data that is the inverse DCT output from the IDCT unit 180. Image information that is an addition output from the adder 181 is supplied to the FM unit 182. The MC section 174 performs motion compensation processing on the image information stored in the FM section 182.

  The MC unit 174 performs motion compensation processing on the image information stored in the FM unit 182 using the optimal motion vector estimated by the ME unit 172, and adds a motion compensation output serving as a reference image to the adder 173. To supply.

  As described above, the adder 173 takes the difference between the conversion output image from the resolution conversion unit 160 and the reference image and supplies the difference to the DCT unit 175.

  The DCT unit 175, the Q unit 176, the VLC unit 177, and the buffer memory 178 operate as described above, and finally, the video encoded data whose resolution is converted from the digital video signal conversion device is bit-streamed at a constant transfer rate. Is output as

  In this digital video signal conversion apparatus, when the motion vector is estimated by the ME unit 172 of the encoding unit 220, the motion associated with the macroblock of the original compressed video signal is not estimated from a state where there is no information. The vector is scale-converted by the scale conversion unit 171 in accordance with the resolution conversion rate in the resolution conversion unit 160, and a narrow range of the converted output image from the resolution conversion unit 160 is obtained based on the scale conversion information from the scale conversion unit 171. A motion vector for motion compensation is estimated by searching. For this reason, since the amount of calculation in the ME unit 172 can be significantly reduced, it is possible to reduce the size of the apparatus and shorten the conversion processing time.

  Next, a twelfth embodiment will be described. This embodiment is also a digital video signal conversion apparatus that performs resolution conversion processing on an MPEG video signal and outputs the result.

  As shown in FIG. 23, this digital video signal conversion apparatus performs orthogonal transform coding by performing only predictive decoding processing using MC on MPEG coded data subjected to the above hybrid coding. A decoding unit 211 that obtains decoded data in the orthogonal transformation region as it is, a resolution conversion unit 260 that performs resolution conversion processing on the decoded data in the orthogonal transformation region from the decoding unit 211, and a conversion output from the resolution conversion unit 260 And an encoding unit 221 that performs compression encoding processing with motion compensation prediction using motion detection based on motion vector information of the MPEG encoded data.

  In the following, a digital video signal conversion apparatus constituted by each of these units will be described, but it goes without saying that each of the components implements the process of each step of the digital signal conversion method according to the present invention.

  Compared with the apparatus shown in FIG. 22, this digital video signal conversion apparatus eliminates the need for the IDCT section 150 in the decoding section 210 and the DCT section 175 and the IDCT section 180 in the encoding section 220. That is, the digital video signal conversion apparatus performs resolution conversion processing on the decoded data in the DCT region and encodes the converted output.

  An orthogonal transformation such as DCT and its inverse transformation generally require a large amount of calculation. For this reason, there is a possibility that the resolution conversion as described above cannot be performed efficiently. Further, since errors are accumulated as the amount of calculation increases, the signal may be deteriorated.

  Therefore, the digital video signal conversion apparatus shown in FIG. 23 omits the IDCT unit 150, the DCT unit 175, and the IDCT unit 150 in FIG. 22, and further changes the function of the resolution conversion unit 160.

  Further, in the DCT region, in order to calculate the later-described definition from the converted DCT coefficient from the resolution conversion unit 160 and to estimate the motion vector using this definition, a definition is used instead of the scale conversion unit 171 shown in FIG. The degree calculation unit 200 is used.

  23 adds the motion compensation output from the MC unit 252 to the DCT coefficient obtained by dequantizing the quantized DCT coefficient decoded by the VLD unit 212 by the IQ unit 213 by the adder 251. The added output (DCT coefficient) is supplied.

  The resolution converter 260 includes an inverse orthogonal transform matrix corresponding to the orthogonal transform matrix used in the DCT encoding applied to the MPEG encoded data, and an IDCT code for obtaining a signal conversion output signal in the time domain. Resolution conversion processing is performed on the DCT coefficients in the DCT conversion region from the decoding unit 211 using a transformation matrix generated based on the orthogonal transformation matrix corresponding to the inverse orthogonal transformation matrix used for conversion.

  The DCT coefficient that is the resolution conversion output from the resolution conversion unit 260 is supplied to the definition calculation unit 200. The definition calculation unit 200 calculates the definition (activity) of the space in units of macroblocks from the luminance component of the DCT coefficient from the resolution conversion unit 260. Specifically, the feature of the image is calculated using the maximum AC value of the DCT coefficient. For example, if there are few high frequency components, it will show that it is a flat image.

  The ME unit 272 estimates an optimal motion vector at the converted resolution based on the definition calculated by the definition calculation unit 200. That is, the ME unit 272 converts the motion vector mv extracted by the VLD 212 based on the definition calculated by the definition calculation unit 200, estimates the motion vector mv, and sends the estimated motion vector mv to the ME unit 272. Supply. Here, the ME unit 272 estimates the motion vector while remaining in the orthogonal transformation region. The ME in this orthogonal transformation region will be described later.

  The resolution conversion DCT coefficient from the resolution conversion unit 260 is supplied to the adder 273 via the definition calculation unit 200 and the ME unit 272.

  The adder 273 takes a difference between a reference DCT coefficient (to be described later) and a converted DCT coefficient from the resolution conversion unit 260 and supplies the difference to a quantization (Q) unit 276.

  The Q unit 276 quantizes the difference value (DCT coefficient) and supplies the quantized DCT coefficient to the VLC unit 277 and the IQ unit 279.

  Further, the rate control unit 283 controls the increase / decrease of the information generation amount in the Q unit 276, that is, the quantization step, based on the definition information from the definition calculation unit 200 and the change information of the increase / decrease of the buffer capacity in the buffer memory 278. .

  The VLC unit 277 compresses and encodes the quantized DCT coefficient from the Q unit 276 by variable length coding, and supplies it to the buffer memory 278. The buffer memory 278 makes the transfer rate of the encoded data compressed by the variable length encoding by the VLC unit 277 constant, and outputs the video encoded data subjected to resolution conversion as a bit stream at a constant transfer rate.

  The IQ unit 279 inversely quantizes the quantized DCT coefficient from the Q unit 276 and supplies the DCT coefficient to the adder 281. The adder 281 adds the motion compensation output from the MC unit 274 to the DCT coefficient that is the inverse Q output from the IQ unit 279. The DCT coefficient information that is the addition output from the adder 281 is supplied to the FM unit 282. The MC unit 274 performs motion compensation processing on the DCT coefficient information stored in the FM unit 282.

  The MC unit 274 performs motion compensation processing on the DCT coefficient information stored in the FM unit 282 using the optimal motion vector estimated by the ME unit 272, and adds the motion compensation output that becomes the reference DCT coefficient. To the container 281.

  As described above, the adder 273 takes the difference between the converted DCT coefficient from the resolution converter 260 and the reference DCT coefficient and supplies the difference to the Q unit 276.

  The Q unit 276, the VLC unit 277, and the buffer memory 278 operate as described above, and finally, video encoded data whose resolution is converted from the digital video signal conversion device is output at a constant transfer rate.

  Here, the MC unit 274 uses the optimal motion vector estimated by the ME unit 272 and the reference DCT coefficient stored in the FM 282, and performs motion compensation in the orthogonal transform region as in the ME unit 272. .

ME and MC in the orthogonal transformation region will be described with reference to FIGS. In FIG. 24, a solid line represents a macroblock of image A to be compressed, and a dotted line represents a macroblock of image B for reference. When the image A to be compressed using the motion vector and the reference image B are overlapped as shown in FIG. 24, the boundary line of the macroblock may not match. In the case of FIG. 24, the macroblock B ′ that is currently compressed straddles the _four macroblocks B ₁ , B ₂ , B ₃ , and B ₄ of the reference image B. Therefore, there is no macroblock of the reference image B that corresponds to the macroblock B ′, and the DCT coefficient of the reference image B where the macroblock B ′ is located can be obtained. Absent. Therefore, it is necessary to obtain the DCT coefficients of the reference image B in the portion where the macroblock B ′ is located by converting the DCT coefficients of the four macroblocks of the reference image B over which the macroblock B ′ straddles. is there.

FIG. 25 schematically shows the procedure of this conversion process. Although the lower left part of the macro block B ₁ of the reference image B overlaps with the macro block B ′, since the upper right part overlaps with the macro block B ′, the DCT coefficient of the macro block B ₁ The macro block B ₁₃ is generated by conversion described later. 'Is a portion that overlaps the macroblock B' lower right portion of the macroblock B ₂ similarly reference image B macroblock B since the portion where the upper left portion overlaps when viewed from the macroblock B ₂ The macroblock B ₂₄ is generated by transforming the DCT coefficients of the later-described DCT coefficients. Generating macro blocks _{B 31} and _{B 42} by performing similar processing to the macroblock _{B 3} and _{B 4.} By combining the four macro blocks B ₁₃ , B ₂₄ , B ₃₁ , and B ₄₂ generated in this way, the DCT coefficient of the reference image B in the portion where the macro block B ′ is located can be obtained.

  That is, it can be expressed as the following equations (6) and (7).

B ′ = B ₁₃ + B ₂₄ + B ₃₁ + B ₄₂ (6)
DCT (B ′) = DCT (B ₁₃ ) + DCT (B ₂₄ )
+ DCT (B ₃₁ ) + DCT (B ₄₂ ) (7)

Next, conversion of the DCT coefficient of the macroblock will be described with reference to FIG. FIG. 26 shows a mathematical model for obtaining a partial B ₄₂ by calculation from an original block, such as B ₄ in the spatial domain. Specifically, extracts the top left of B _4, interpolated by 0, are moved to the lower right. B ₄₂ obtained from the calculation of the following equation (8) from the block B ₄ is shown.

Here, I _h and I _w are identification codes of respective matrices of size h × h and w × w composed of columns and rows of h and w extracted from the block B ₄ . As shown in FIG. 26, the pre-matrix H ₁ synthesized first with B ₄ takes out the first h columns and converts it to the bottom, and H ₂ synthesized later with B ₄ takes out the first w rows. And convert to the right.

Based the above formula _(8), the DCT of _{B 42} can be directly calculated from the DCT of _{B 4} by the following equation (9).

DCT (B ₄₂ ) = DCT (H ₁ ) × DCT (B ₄ ) × DCT (H ₂ ) (9)

When this is applied to all the sub-blocks and summed up, the DCT coefficients of the new block B ′ can be obtained directly from the DCTs of the original blocks B _{1 to} B ₄ as shown in the following equation (10).

Here, the DCT of H _i1 and H _i2 may be calculated in advance and stored in the memory to configure the table memory. In this way, ME and MC are possible even in the orthogonal transform region.

  In the encoding unit 221, when the motion vector is estimated by the ME unit 272, the motion vector attached to the macro block of the original compressed video signal is not estimated from a state where there is no information at all. A narrow range is searched for and estimated according to the definition calculated by the definition calculation unit 200 from the converted output of the unit 260.

  As described above, in the decoding unit 211 of the digital video signal conversion device according to another embodiment, MPEG coding subjected to hybrid coding including predictive coding with motion detection and orthogonal transform coding is performed. Predictive decoding processing with motion compensation for data, that is, IQ after VLD, motion compensation is performed there to obtain decoded data as it is in the DCT region, and resolution conversion is performed on the decoded data in the DCT region, so that orthogonal transformation is performed. Therefore, resolution conversion can be performed directly in the area, decoding into time domain or space domain (inverse orthogonal transformation) is not required, calculation is simplified, and high quality conversion with small calculation error can be performed. Further, in the encoding unit 221, when the ME unit 272 estimates a motion vector, it is not estimated from a state where there is no information at all. Instead, the motion vector attached to the macroblock of the original compressed video signal is subjected to resolution conversion. A motion vector is estimated by searching in a narrow range according to the definition calculated from the output. For this reason, since the amount of calculation in the ME unit 272 can be significantly reduced, the apparatus can be downsized and the conversion process can be shortened.

  Next, a thirteenth embodiment will be described. This example is also a digital video signal conversion device that performs signal conversion processing such as resolution conversion processing on MPEG encoded data and outputs video encoded data.

  As shown in FIG. 27, the digital video signal conversion apparatus includes a decoding unit 340 that performs partial decoding processing on the MPEG encoded data that has been subjected to the hybrid encoding to obtain data in the orthogonal transform region. A conversion unit 343 that performs resolution conversion processing on the orthogonal transform region data from the decoding unit 340, and a motion vector based on the motion vector information of the MPEG encoded data is added to the conversion output from the conversion unit 343. And an encoding unit 350 that performs compression encoding processing.

  The decoding unit 340 includes a VLD unit 341 and an IQ unit 342. The VLD unit 341 and IQ unit 342 have the same configuration as the VLD unit 112 and IQ unit 113 shown in FIG. 21 and operate in the same manner. The decoding unit 340 is characterized in that MC is not performed.

  That is, the P picture and the B picture are not subjected to MC, and the conversion unit 343 performs resolution conversion on the DCT coefficient serving as difference information. The converted DCT coefficient obtained by the resolution conversion is quantized by the Q unit 345 whose rate is controlled by the rate control unit 348, variable-length decoded by the VLC unit 346, and then output at a constant rate by the buffer memory 347. Is done.

  At this time, the motion vector conversion unit 344 of the encoding unit 350 rescals the motion vector mv extracted by the VLD unit 341 according to the resolution conversion rate and supplies the rescaled motion vector mv to the VLC unit 346.

  The VLC unit 346 adds the rescaled motion vector mv to the quantized DCT coefficient from the Q unit 345, performs variable length encoding processing, and supplies the encoded data to the buffer memory 347.

  In this way, the digital video signal conversion apparatus shown in FIG. 27 does not perform MC in the decoding unit 340 and the encoding unit 350, so that the calculation can be simplified and the hardware burden can be reduced.

  Rate conversion may be performed by each of the digital video signal conversion devices described above. In other words, it may be applied when converting the transfer rate from 4 Mbps to 2 Mbps with the resolution unchanged.

  In each of the above embodiments, the device configuration has been described. However, each of the above devices may be configured by using the digital signal conversion method according to the present invention as software.

  According to the embodiment of the present invention described above, decoding with motion compensation is performed on an input information signal that has been compression-encoded with motion detection, signal conversion processing is performed on the decoded signal, and The converted signal is subjected to compression coding processing with motion detection based on the motion vector information of the input information signal. When applying the resolution conversion process as the signal conversion process, the conversion signal is subjected to a compression encoding process with motion compensation based on information obtained by scaling the motion vector information according to the resolution conversion process. . In particular, since motion vector information necessary for compression encoding is scaled according to the resolution conversion rate and searched and estimated in a narrow range, the amount of calculation at the time of motion vector estimation can be greatly reduced, and the device Downsizing and conversion processing time can be shortened.

  In the embodiment of the present invention, a partial decoding process is performed on an input information signal that has been subjected to compression coding including predictive coding with motion detection and orthogonal transform coding, thereby performing orthogonal transform regions. The decoded signal of the orthogonal transform domain is subjected to signal conversion processing, and the converted signal is subjected to compression coding with motion compensation prediction using motion detection based on the motion vector information of the input information signal. Apply processing. At this time, when applying the resolution conversion process as the signal conversion process, the compression code is accompanied by motion compensation based on the information obtained by converting the motion vector information according to the definition obtained from the resolution conversion process. Since the conversion process is performed on the converted signal, motion vector information required for compression encoding can be searched and estimated in a narrow range, and the amount of calculation can be greatly reduced, so that the apparatus can be downsized and the conversion processing time can be shortened. Can be achieved. In addition, since signal transformation processing can be performed in the orthogonal transformation domain, inverse orthogonal transformation processing is unnecessary, decoding into the time domain and space domain (inverse orthogonal transformation) is unnecessary, calculation is simplified, and high quality with small calculation errors Can be converted.

  In the embodiment of the present invention, a partial decoding process is performed on an input information signal that has been subjected to compression coding including predictive coding with motion detection and orthogonal transform coding, thereby performing orthogonal transform regions. The decoded signal of the orthogonal transform domain is subjected to signal conversion processing, and motion vector information converted based on the motion vector information of the input information signal is added to the converted signal to perform compression encoding processing. Apply. For this reason, when the resolution conversion process is applied as the signal conversion process, a compression encoding process to which information obtained by scaling the motion vector information is added to the converted signal according to the resolution conversion process. .

  That is, since motion vector information added at the time of compression encoding can be searched and estimated within a narrow range, the amount of calculation at the time of motion vector estimation can be greatly reduced. In addition, since the signal conversion process can be performed in the orthogonal transform region, the inverse orthogonal transform process can be omitted. In addition, since no motion compensation processing is used at the time of decoding and encoding, the amount of calculation can be further reduced.

  It should be noted that the present invention is not limited to the above-described embodiments, and various modifications can be made without departing from the scope of the present invention.

It is a block diagram which shows the structural example of the digital signal converter which concerns on the 1st Embodiment of this invention. It is a figure for demonstrating the principle of the resolution conversion in an orthogonal transformation area | region. It is a figure for demonstrating the principle of the resolution conversion in an orthogonal transformation area | region. 4A to 4C are diagrams schematically showing a state in which a DV video signal is converted into an MPEG video signal by digital signal conversion according to the first embodiment of the present invention. It is a figure for demonstrating the relationship between DV format and MPEG format. It is a figure for demonstrating the basic calculation procedure for the resolution conversion process. 7A and 7B are diagrams for explaining the “still mode” and “motion mode” of the DV format. It is a figure for demonstrating the procedure of the conversion process in "still mode." FIG. 9A to FIG. 9C are block diagrams showing configuration examples of the digital signal conversion apparatus according to the second embodiment of the present invention. It is a figure for demonstrating the procedure of the conversion process in the case of enlarging an image. It is a block diagram which shows the structural example of the digital signal converter which concerns on the 3rd Embodiment of this invention. It is a block diagram which shows the structural example of the digital signal converter which concerns on the 4th Embodiment of this invention. It is a block diagram which shows the structural example of the digital signal converter which concerns on the 5th Embodiment of this invention. It is a block diagram which shows the structural example of the digital signal converter which concerns on the 6th Embodiment of this invention. It is a block diagram which shows the structural example of the digital signal converter which concerns on the 7th Embodiment of this invention. In the 7th Embodiment of this invention, when a DV video signal is converted into an MPEG video signal, it is a flowchart which shows the basic procedure by which a quantization scale is set for every macroblock (MB) of each frame. is there. In the 7th Embodiment of this invention, it is a flowchart which shows the basic procedure which applies a feedback with respect to the following flame | frame using the set quantization scale. It is a block diagram which shows the structural example of the digital signal converter in the past which converts an MPEG video signal into a DV video signal. It is a block diagram which shows the structural example of the digital signal converter which concerns on the 8th Embodiment of this invention. It is a block diagram which shows the structural example of the digital signal converter which concerns on the 9th Embodiment of this invention. It is a block diagram which shows the structural example of the digital signal converter which concerns on the 10th Embodiment of this invention. It is a block diagram which shows the structural example of the digital signal converter which concerns on the 11th Embodiment of this invention. It is a block diagram which shows the structural example of the digital signal converter which concerns on the 12th Embodiment of this invention. In the 12th Embodiment of this invention, it is a figure for demonstrating the motion compensation and motion estimation process in an orthogonal transformation area | region, and shows a mode that the macroblock B straddled the several macroblock of the image for a reference | standard FIG. In the twelfth embodiment of the present invention, it is a diagram for describing motion compensation and motion estimation processing in an orthogonal transform region, and is a diagram illustrating conversion processing of a reference macroblock. In the twelfth embodiment of the present invention, it is a diagram for explaining motion compensation and motion estimation processing in an orthogonal transform region, and is a diagram showing a conversion procedure of a reference macroblock. It is a block diagram which shows the structural example of the digital signal converter which concerns on the 13th Embodiment of this invention. It is a block diagram which shows the structural example of the conventional digital signal converter.

Explanation of symbols

  8,210 decoding unit, 9,220 encoding unit, 11 deframing unit, 12,112 variable length decoding unit, 13, 26,113 inverse quantization unit, 14,25 inverse weighting unit, 16 resolution conversion unit, 18 weighting Unit, 19, 141 quantization unit, 20, 142 variable length coding unit, 21 rate conversion unit, 24 frame memory unit, 29 rate control unit, 100 MPEG decoder, 110 DVD encoder, 115, 152, 174 motion compensation unit, 130 discrete cosine transform unit, 131 inverse discrete cosine transform unit, 143 framing unit, 160 resolution transform unit, 171 motion vector transform unit, 172 motion estimation unit

Claims (14)

In a digital signal conversion method for converting a digital signal in a first format into a digital signal in a second format,
A decoding step of decoding the digital signal of the first format;
A signal converting step of converting the decoded digital signal of the first format into a digital signal of the second format;
An encoding step of encoding the digital signal of the second format;
A digital signal conversion method comprising: a weighting process step for collectively performing inverse weighting on the decoded digital signal in the first format and weighting on the digital signal in the second format. The digital signal conversion method according to claim 1, wherein the digital signal of the first format is a digital signal subjected to orthogonal transformation, and the weighting processing step is performed in an orthogonal transformation region. The digital signal of the first format is a digital signal subjected to orthogonal transformation, and the weighting processing step is performed in a spatial domain after the orthogonal transformation of the digital signal subjected to the orthogonal transformation. Digital signal conversion method. The digital signal of the first format is a video signal compressed and encoded at a predetermined fixed rate by discrete cosine transform, and the digital signal of the second format is a video signal compressed and encoded at a variable rate by discrete cosine transform. The digital signal conversion method according to claim 1, wherein the signal is a signal. The digital signal conversion method according to claim 1, wherein the weighting step is performed after the inverse quantization of the digital signal of the first format and before the signal conversion step. 2. The digital signal conversion method according to claim 1, wherein the weighting processing step is performed after the signal conversion step and before the quantization of the second format. The signal conversion step uses the data amount information included in the digital signal of the first format to control the data amount of the digital signal of the second format in an orthogonal transform region. The digital signal conversion method according to 1. In a digital signal converter for converting a digital signal of a first format into a digital signal of a second format,
Decoding means for decoding the digital signal of the first format;
Signal converting means for converting the decoded digital signal of the first format into the digital signal of the second format;
Encoding means for encoding the second digital signal;
A digital signal conversion apparatus comprising: weighting processing means for collectively performing inverse weighting on the digital signal in the first format and weighting on the digital signal in the second format. The digital signal of the first format is a digital signal subjected to orthogonal transformation, and the weighting processing means performs inverse weighting for the digital signal of the first format and weighting for the digital signal of the second format in the orthogonal transformation region. 9. The digital signal converter according to claim 8, wherein the digital signal converter is performed. The digital signal of the first format is a digital signal subjected to orthogonal transformation, and the weighting processing means performs inverse weighting for the digital signal of the first format and weighting for the digital signal of the second format in the spatial domain. 9. The digital signal converter according to claim 8. The digital signal of the first format is a video signal compressed and encoded at a predetermined fixed rate by discrete cosine transform, and the digital signal of the second format is a video signal compressed and encoded at a variable rate by discrete cosine transform. 9. The digital signal converter according to claim 8, wherein the digital signal converter is a signal. The weighting processing means is arranged upstream of the signal converting means, and performs inverse weighting on the digital signal in the first format and weighting on the digital signal in the second format. 9. The digital signal converter according to claim 8, wherein the digital signal converter is performed after the inverse quantization of. The weighting processing means is disposed downstream of the signal converting means, and performs inverse weighting on the digital signal in the first format and weighting on the digital signal in the second format. The digital signal conversion apparatus according to claim 8, wherein the digital signal conversion is performed prior to quantization. The signal conversion means controls the data amount of the digital signal of the second format in an orthogonal transform region by using data amount information included in the digital signal of the first format. 9. The digital signal converter according to 8.

Images