JP3131041B2 - Quantizer - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a quantizer for quantizing data to be quantized into quantized representative values.

[0002]

2. Description of the Related Art Quantizers for quantizing data to be quantized are used for encoding signals such as image signals and audio signals.

FIG. 5 is a block diagram showing a configuration of a still picture signal coding apparatus having a quantizer which is a quantization apparatus.

An image signal is converted into a discrete cosine transform (hereinafter, DC).
(T-conversion).
The DCT processing unit 1 converts the image signal into DCT coefficients,
The DCT coefficient is given to the quantizer 2. Quantizer 2 is
A quantization index, which is a quantization representative value corresponding to the given DCT coefficient, is obtained, and the obtained quantization index is provided to the variable-length encoder 3. The variable-length encoder 3 performs variable-length encoding on the given quantization index and compresses it, and outputs encoded data that is the compressed data.

FIG. 6 is a block diagram showing the internal configuration of the quantizer 2. The DCT coefficient from the DCT converter 1 is provided to a divider 2a as a dividend input. Further, a quantizer coefficient, which is data for quantizing the DCT coefficient, is given to the divider 2a as a divisor input. The divider 2a divides the given DCT coefficient by the quantization coefficient, and supplies the result of the division to the post-processing unit 2b.

The post-processing unit 2b performs a rounding process such as rounding on the given division result, and obtains a quantization index as a result. Then, the post-processing unit 2 b gives the obtained quantization index to the variable-length encoder 3.

[0007] In the quantizer 2 having such a configuration, for example, by the processing of the divider 2a and the post-processing unit 2b, for example, FIG.
The quantization is performed using the quantization characteristics shown in FIG. FIG.
5 is a graph showing an example of a quantization characteristic of the quantizer 2, wherein a horizontal axis represents a DCT coefficient, and a vertical axis represents a quantization index, and shows a relationship between these. The quantization characteristic shown in FIG. 7 is called linear quantization determined by a quantization coefficient (QS), and is often used as a typical quantization characteristic of a quantization device.

[0008]

In the conventional quantization apparatus, it was necessary to use a divider as described above. However, since a divider such as a parallel divider has a large circuit size as is well known, There has been a problem that the integration of the device into an integrated circuit is hindered. Further, the conventional quantizer has a problem that it is difficult to speed up the processing operation of the quantizer because of the worst delay path in which data passes through the parallel divider.

SUMMARY OF THE INVENTION The present invention has been made to solve such a problem, and provides a quantization apparatus which can reduce the size of its circuit configuration and can speed up its processing operation. The purpose is to:

[0010]

According to the first aspect of the present invention, an image data obtained by discrete cosine transform of image data is obtained.
A quantizing device for quantizing a DCT coefficient to be used, comprising a storage unit, a multiplication unit, a shift unit, and a quantization representative value determination unit.

The storage means stores the reciprocal of the quantized coefficient for quantizing the DCT coefficient separately into mantissa data and exponent data. The multiplication means multiplies the mantissa data stored in the storage means by a DCT coefficient . The shift means shifts the result of the multiplication by the multiplication means based on the exponent part data stored in the storage means. Quantized representative value determining means, based on the shift result of the shift means
A quantization representative value corresponding to the DCT coefficient is determined.

The present invention according to claim 2 comprises storage means, multiplication means, shift means and quantization representative value determination means.

The storage means stores the reciprocal of the quantization coefficient for quantizing the DCT coefficient separately into mantissa data excluding the most significant bit and exponent data, and the mantissa data is stored in the most significant bit. Is normalized so as to be a predetermined value. The multiplication means adds a predetermined value to the mantissa data stored in the storage means, and multiplies the mantissa data to which the most significant bit is added by the DCT coefficient . The shift means shifts the result of the multiplication by the multiplication means based on the exponent part data stored in the storage means. The quantized representative value determining means determines a quantized representative value corresponding to the DCT coefficient based on the shift result of the shift means.

[0014]

According to the first aspect of the present invention, the reciprocal of the quantization coefficient stored in the storage means is stored separately for the mantissa part data and the exponent part data. The mantissa data and the quantized data are multiplied by the multiplication means, and the multiplication result is shifted by the shift means based on the exponent data stored in the storage means. That is, the data to be quantized is divided by the quantization coefficient by the multiplying means and the shifting means. With only the digit shift result of the digit shift means, the quantized data cannot be represented as a quantized representative value represented by an integer of a predetermined range. For example, a rounding process such as rounding and regulation of upper and lower limits is performed to determine a quantization representative value.

According to the second aspect of the present invention, the mantissa data is normalized such that the most significant bit becomes a predetermined value, and the storage means removes the most significant bit. The mantissa data is stored. As a result, the amount of data to be stored in the storage means is smaller than that of the first aspect. The multiplying means adds the predetermined value to the mantissa data stored in the storage means and multiplies the mantissa data by the data to be quantized. Perform multiplication.

[0016]

Next, an embodiment of the present invention will be described in detail with reference to the drawings.

FIG. 1 is a block diagram showing the configuration of the quantization device. In a storage unit 21 which is a storage unit including a RAM or a ROM, a reciprocal of a predetermined number of quantized coefficients for quantizing data to be quantized is divided into mantissa part data and exponent part data and tabulated. Is remembered. The address input terminal A of the storage unit 21 is supplied with quantization coefficient selection data for selecting a quantization coefficient.
The storage unit 21 outputs the mantissa part data of the reciprocal of the quantization coefficient corresponding to the given quantization coefficient selection data to the output terminal D1.
To the multiplier 22 and output the exponent data from the output terminal D2 to the shifter unit 23.

The multiplier 22 serving as a multiplying means is supplied with the quantized data as a multiplicand input and the mantissa data from the storage unit 21 as a multiplier input. The multiplier 22 multiplies the given quantized data by the mantissa data and supplies the multiplication result to the shifter 23.

The shifter unit 23 serving as a shift means applies the multiplication result data provided from the multiplier 22 to the data.
The shift processing is performed by the amount of the exponent part data provided from the storage unit 21, and the shift result is provided to the post-processing unit 24.

A post-processing unit 24 serving as a quantization representative value determining unit
Performs rounding processing such as rounding and clip processing (processing for regulating the upper limit and lower limit) on the data of the shift result given from the shifter unit 23 to obtain a quantized representative value represented by an integer value. Is determined, and the result is output.

Next, specific examples of the mantissa part data and the exponent part data of the reciprocal of the quantization coefficient stored in the storage unit 21 will be described. FIG. 2 is a data table diagram showing a first specific example of the reciprocal of the quantized coefficient stored in the storage unit 21. The quantized coefficient (in decimal notation) represented by integer values of 1 to 63 and , And the reciprocal mantissa part data (in binary notation) and the exponent part data (in decimal notation) corresponding to each quantized coefficient. In this data table, the result of dividing the mantissa data by the exponent data is the reciprocal of the quantization coefficient.

In FIG. 2, the mantissa part data of the reciprocal of the quantized coefficient is 11 bits which is a sufficient number of bits that does not cause a problem with the quantization error of the reciprocal, and the exponent data is 3 bits. is there. The data to be quantized in this case is a 12-bit binary number. Although the mantissa data in this case is not normalized,
When the mantissa data is not normalized, the possible value of the exponent data becomes smaller than the reciprocal of the quantized coefficient in which the mantissa data is normalized, as shown in FIG. The configuration of the unit 23 is simplified.

FIG. 3 is a data table diagram showing a second specific example of the reciprocal of the quantization coefficient stored in the storage unit 21. The data shown in FIG. 3 is normalized such that the mantissa data is within a predetermined numerical range (so that the most significant bit, which is the leftmost bit of the mantissa data, is 1), and the part thereof is It is different from the data shown in FIG.
When the mantissa data is normalized in this manner, the most significant bit of the mantissa data always becomes 1, and therefore the mantissa data stored in the storage unit 21 is as shown in FIG. In this case, 1 may always be fixedly input to the most significant bit of the multiplier input of the multiplier 22. By doing so, the data storage capacity of the storage unit 21 can be reduced by one bit.

As is well known, since a multiplier has a smaller circuit size than a divider which performs the same operation, the above-described quantization apparatus is configured by using the multiplier 22, so that a conventional multiplier is used. Thus, the circuit scale can be made smaller than that of the quantizer using the divider. In addition, the reciprocal of the quantized coefficient is divided into mantissa data and exponent data, and only the mantissa data is used for multiplication in the multiplier 22. The size can be reduced.

In the above quantization apparatus, the worst delay path is changed to a path including a multiplier different from the path including the conventional divider. In general, when the input word length is substantially the same, the division is performed. Since the delay time of the multiplier is smaller than that of the multiplier,
Processing speed of the quantizer can be easily increased.

[0026]

According to the first aspect of the present invention, quantization is performed by the circuit configuration using the multiplying means, and only the mantissa data of the reciprocal of the quantization coefficient is multiplied by the multiplying means. Therefore, the circuit can be configured with a multiplier having a smaller circuit size than a divider that performs equivalent arithmetic processing,
This makes it possible to easily realize an integrated circuit of the quantization device. Further, the multiplier used as the multiplication means is
In general, since the operation processing operation is significantly faster than that of a divider that performs the same operation, the processing of the quantization device can be speeded up.

According to the second aspect of the present invention, in addition to the effect of the first aspect of the present invention, since the most significant bit is removed from the mantissa data stored in the storage means,
The storage capacity of the storage means can be reduced.

[Brief description of the drawings]

FIG. 1 is a block diagram illustrating a configuration of a quantization device.

FIG. 2 is a data table diagram showing a first specific example of a reciprocal of a quantization coefficient stored in a storage unit.

FIG. 3 is a data table diagram showing a second specific example of a reciprocal of a quantization coefficient stored in a storage unit.

FIG. 4 is a data table diagram showing a third specific example of a reciprocal of a quantization coefficient stored in a storage unit.

FIG. 5 is a block diagram illustrating a configuration of a still image signal encoding device having a conventional quantization device.

FIG. 6 is a block diagram showing an internal configuration of a conventional quantization device.

FIG. 7 is a graph showing an example of a quantization characteristic of a quantization device.

[Explanation of symbols]

 DESCRIPTION OF SYMBOLS 21 Storage part 22 Multiplier 23 Shifter part 24 Post-processing part

──────────────────────────────────────────────────続 き Continued on the front page (58) Field surveyed (Int.Cl. ⁷ , DB name) H03M 7/30

Claims (2)

(57) [Claims] An image data is subjected to a discrete cosine transform.
A quantizer for quantizing DCT coefficients obtained by a memory means for storing the inverse of the quantization coefficient for quantizing the DCT coefficients are divided into a mantissa data and the exponential part data, said storage Mantissa data stored in the means and the DCT
Multiplication means for multiplying the number, the shift means for shift of the multiplication result of said multiplying means based on the stored exponent data in the storage means, on the basis of the shift result of said shift means A quantizing representative value determining means for determining a quantizing representative value corresponding to the DCT coefficient . 2. A method for performing discrete cosine transform on image data.
A quantizer for quantizing DCT coefficients obtained by the inverse of the quantization coefficient for quantizing the DCT coefficients, stores divided into a mantissa data and the exponential part data except for the most significant bit Storage means, wherein the mantissa data is normalized so that the most significant bit becomes a predetermined value, and the predetermined value is added to the mantissa data stored in the storage means, Multiplication means for multiplying the DCT coefficient by the mantissa data to which the most significant bit is added; shift means for shifting the multiplication result of the multiplication means based on the exponent data stored in the storage means; the shift means includes a shift results in further quantization representative value determining means for determining a quantization representative value corresponding to the DCT coefficients <br/> based, quantizer.