CN106327443A - Night image enhancement method based on improved genetic algorithm - Google Patents
Night image enhancement method based on improved genetic algorithm Download PDFInfo
- Publication number
- CN106327443A CN106327443A CN201610710468.9A CN201610710468A CN106327443A CN 106327443 A CN106327443 A CN 106327443A CN 201610710468 A CN201610710468 A CN 201610710468A CN 106327443 A CN106327443 A CN 106327443A
- Authority
- CN
- China
- Prior art keywords
- image
- brightness
- component
- genetic algorithm
- nighttime
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 230000002068 genetic effect Effects 0.000 title claims abstract description 54
- 238000000034 method Methods 0.000 title claims abstract description 53
- 210000000349 chromosome Anatomy 0.000 claims abstract description 19
- 230000035772 mutation Effects 0.000 claims abstract description 8
- 230000002708 enhancing effect Effects 0.000 claims description 27
- 230000008569 process Effects 0.000 claims description 17
- 238000006243 chemical reaction Methods 0.000 claims description 12
- 108090000623 proteins and genes Proteins 0.000 claims description 11
- 238000005516 engineering process Methods 0.000 claims description 10
- 238000005457 optimization Methods 0.000 claims description 10
- 238000011156 evaluation Methods 0.000 claims description 8
- 238000010353 genetic engineering Methods 0.000 claims description 8
- 230000008859 change Effects 0.000 claims description 7
- 239000000284 extract Substances 0.000 claims description 7
- 238000005728 strengthening Methods 0.000 claims description 7
- 230000000694 effects Effects 0.000 claims description 6
- 238000000354 decomposition reaction Methods 0.000 claims description 5
- 238000009826 distribution Methods 0.000 claims description 5
- 238000013441 quality evaluation Methods 0.000 claims description 4
- 230000004083 survival effect Effects 0.000 claims description 4
- 238000004458 analytical method Methods 0.000 claims description 3
- 238000005316 response function Methods 0.000 claims description 3
- 101150084750 1 gene Proteins 0.000 claims description 2
- 230000008649 adaptation response Effects 0.000 claims description 2
- 230000003044 adaptive effect Effects 0.000 claims description 2
- 230000004075 alteration Effects 0.000 claims description 2
- 238000004040 coloring Methods 0.000 claims description 2
- 238000005070 sampling Methods 0.000 claims description 2
- 210000004243 sweat Anatomy 0.000 claims description 2
- 238000006073 displacement reaction Methods 0.000 claims 1
- 238000012545 processing Methods 0.000 abstract description 6
- 230000007547 defect Effects 0.000 abstract description 4
- 125000001475 halogen functional group Chemical group 0.000 abstract 1
- 230000003287 optical effect Effects 0.000 abstract 1
- 238000004364 calculation method Methods 0.000 description 4
- 238000013507 mapping Methods 0.000 description 4
- 238000004519 manufacturing process Methods 0.000 description 2
- 230000009466 transformation Effects 0.000 description 2
- 238000009395 breeding Methods 0.000 description 1
- 230000001488 breeding effect Effects 0.000 description 1
- 239000012141 concentrate Substances 0.000 description 1
- 230000007850 degeneration Effects 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 238000001914 filtration Methods 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 238000012544 monitoring process Methods 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 238000010187 selection method Methods 0.000 description 1
- 230000000007 visual effect Effects 0.000 description 1
- 238000005303 weighing Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T5/00—Image enhancement or restoration
- G06T5/73—Deblurring; Sharpening
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/10—Image acquisition modality
- G06T2207/10024—Color image
Landscapes
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Image Processing (AREA)
Abstract
The invention firstly provides night image enhancement based on an improved genetic algorithm. The method is suitable for processing a night image with low contrast and low noise, and settles the problems such as confusion, ghost and halo in a traditional image enhancement algorithm. According to the night image enhancement method, an intersected mode is utilized for performing global searching, and linear combination between two chromosomes in the image is defined and enhanced through random selection, thereby ensuring direct searching of an intersecting operator in a parent generation and a filial generation. Particularly, the method comprises the steps of randomly generating chromosomes of which the lengths are brightness grade of an enhanced image; selecting the chromosomes from a current component by means of an appropriate function, and generating the filial generation by means of a genetic algorithm intersected mode and a mutation operator; eliminating the chromosomes which do not satisfy the condition by means of an eliminating algorithm; performing traversal of all chromosomes by means of a searching strategy of the genetic algorithm, if a searching stopping strategy is satisfied, outputting an optimal chromosome, and otherwise, repeating the second step of generating the filial generation until an optical solution is generated and output. Compared with an existing image contrast enhancement method, the night image enhancement method based on improved genetic algorithm is advantageous in that the method well settles image color un-matching after enhancement and defects of the genetic algorithm in image enhancement.
Description
Technical field
The present invention relates to computer vision field and signal processing field, particularly relate to image enhaucament, genetic algorithm and
The method of image procossing.
Background technology
Contrast enhanced technology is a key technology of image enhancement processing, such as some digital film, video monitoring, meter
Calculation machine visual processes etc..Image is often as some uncontrollable factors, such as non-professional Taking Pictures recording, image conversion process
The change of the loss of middle information, ambient brightness, the equipment etc. of acquisition image all can cause degeneration or the lost part information of image.For
Solving these problems, many researcheres propose the algorithm for image enhancement of low quality low-light levels.
Traditional image comparison strengthens algorithm and is broadly divided into two classes: spatial domain method, frequency domain method.Spatial domain method includes: rectangular histogram is equal
Weighing apparatus, smothing filtering, local gray level, edge extracting etc., traditional algorithm is suitable for specific image and general applicability is poor.Frequency domain method bag
Include: Fourier transform, dct transform etc..Method is computationally intensive, needs manual intervention.Based on analysis, the present invention increases image comparison
Strong technology is summarized as two classes: linear enhancement method and non-linear Enhancement Method.So-called linear enhancement method specifies justice Contrast enhanced
And improve contrast as far as possible, it is assumed that the gray-scale level scope of low-luminosity picture is [a, b], the gray-scale level scope line of enhanced image
Property expand to [c, d], then the linear mathematic(al) representation that strengthens is:
Fig. 1 (a) shows that a typical linear transformation is the curve of Contrast enhanced.Method is simple, and amount of calculation is less, but
When the gray-scale pixels of image concentrates on maximum tonal gradation or minimal gray grade, the result of linear transformation Contrast enhanced is also
Undesirable.Some researcheres propose nonlinear Contrast enhanced technology recently, and this kind of method is equivalent to that image carries out certain and becomes
Changing, to reach the purpose of image enhaucament, relate to the determination problem of running parameter, amount of calculation is bigger.Non-linear contrast strengthens
Algorithm specifically include that Gamma enhancing, the conversion of pixel grey scale technology, histogram equalization technology, filter technology, local gray level,
Edge extractings etc., Fig. 1 (b) shows the curve of the typical Contrast enhanced using non-linear conversion.
Image histogram is the statistical relationship in description digital picture between each gray level and its frequency of occurrences.Assume image
Tonal gradation scope is that the function of the number of pixels of gray level is as follows:
h(rk)=nk (2)
Wherein k=0,1 ..., L-1, rkIt is kth level gray scale, nkRepresent that in image, gray level is rkNumber of pixels.To returning
One rectangular histogram changed, by the sum of all pixels of kth gray level divided by the total number of pixels of image, it may be assumed that p (rk)=nk/ n obtains.Make straight
The purpose of side's figure is through the shape of observation figure, it is judged that production process is the most stable, it was predicted that the quality of production process.If one
The pixel of width image is occupied the most possible gray level and is evenly distributed, then the image shown has high-contrast and changeable ash
Degree tone.
Histogram equalization is the method utilizing image histogram to be adjusted contrast in image processing field.Nogata
Figure equalization is that the grey level histogram of original image is become in whole tonal ranges from certain the gray scale interval comparing concentration
Be uniformly distributed.Its objective is image is carried out Nonlinear extension, redistribute image pixel value, in making certain tonal range
Pixel quantity is roughly the same, then the histogram distribution of given image is changed over the distribution of " uniformly " distribution histogram.Exist
Problem is: after (1) conversion, the gray level of image reduces, and some details disappears；(2) some image, as rectangular histogram has peak, through place
The factitious undue enhancing of contrast after reason；(3) cause excessively strengthening or flicker effect due to gray scale stretching.The most various bases
Algorithm in histogram equalization amendment puts forward to make up these problems, and its target is close to the pattern of histogram equalization.In order to
Preferably protecting the edge details of image, some researcheres propose the framework of histogram modification, the mapping function T [n] of enhancing
It is expressed as follows:
Wherein b represents the tonal gradation being stretched to maximum black part, and w represents the gray scale etc. being stretched to maximum white portion
Level, g [n] is to represent sbAnd swBetween gamma function mapping relations.sbAnd swRepresent brightness light and shade stretching factor.In order to process one
A little noises and light and shade stretching problem, histogram equalization applies each stage at T [n].
Process based on gray scale layer based on Contrast enhanced algorithm majority linearly or nonlinearly, if color image enhancement
Directly use these algorithms will produce the inharmonic problem of color.Reason is: three components of the red, green, blue of coloured image are also
It not linear relationship, but have the strongest dependency between three components.The most usually need to carry out color of image space turning
Change.For image enhancement technique, genetic algorithm is usually present following defect in the process processed: 1. image carries out overall situation increasing
Strength is managed, and ignores the local message of image, and the enhancing result using genetic algorithm optimization is unsatisfactory；The most various parameters
The intervention needing user is set, is generally not capable of being automatically performed image enhancement processing；3. big due to image data amount, genetic algorithm is real
The existing required time is long.In consideration of it, the main object of the present invention is to solve color of image problem of disharmony and heredity calculation after enhancing
The method defect when strengthening image.In order to realize this target, we have proposed one based on Revised genetic algorithum to cromogram
As carrying out enhancement process.
Summary of the invention
Strengthening to solve color of image problem of disharmony and genetic algorithm after current techniques carries out color image enhancement
Defect during image.The present invention proposes one, based on Revised genetic algorithum, coloured image is carried out enhancement process.
In the framework that the present invention proposes, first RGB color is transformed into HSI color space, then extracts HSI face
Gray scale I component in the colour space.Owing to the present invention is to strengthen based on brightness, so further using after extracting I component
Retinex theoretical decomposition is inverse image R and luminance picture L.Contrast enhanced nighttime image brightness is complete by genetic algorithm optimization
Becoming, the nighttime image finally strengthened is by the brightness strengthened, inverse image and the reconstruct of nighttime image color.Propose based on improvement
Genetic algorithm strengthen nighttime image, its essence is real image to strengthen problem be converted into genetic algorithm parameter optimization problem.
Basic step includes:
Step A changes RGB color picture format into HSI color space image form；
Step B extracts I component from HSI color space, and using Retinex theoretical decomposition I component is inverse image and bright
Degree image；
Step C carries out Contrast enhanced based on genetic algorithm to the luminance picture at night extracted；
Step D uses the inverse image at night, highlights image and nighttime image color carries out image reconstruction.
Accompanying drawing explanation
Fig. 1 (a) is linear conversion method, Fig. 1 (b) non-linear conversion method.
Fig. 2 is the RGB nighttime image image to HSI color space.Wherein Fig. 2 (a) figure is brightness (gray scale) I component, Fig. 2
B () is colourity H component, Fig. 2 (c) saturation S component.
Fig. 3 (a) is the brightness mapping relations of image input and output, and Fig. 3 (b) is that the brightness of input picture and enhancing image is reflected
Penetrate relation.
Fig. 4 inventive algorithm is embodied as principle and block diagram.
Detailed description of the invention
The present invention is described further with specific embodiment below in conjunction with the accompanying drawings.
Implement step A: current picture superposition technology is mainly based upon gray level image and processes, if cromogram
Image intensifying directly uses gray level image to strengthen algorithm, and result is undesirable.If color space is changed, can be relevant
Three components are converted in incoherent chrominance space go, and enhanced result is preferable.Found by analysis, HSI color model
This demand can be better met.Assume that the low-light level RGB color image of input is determined by three tones of red, green, blue.From RGB
Color space conversion is to HSI color space, and each concrete component solves as follows, and H component can obtain by equation below:
Saturation component S is given by:
Last luminance component I is given by:
Wherein H represents colourity (Hue), and S represents saturation (Saturation), and I represents brightness value (Intensity).?
During enhancing, colourity and saturation keep constant, directly strengthen the luminance component of coloured image, it is ensured that do not have color inclined
Move, more can meet the requirement of human eye.Note: in the range of supposing that rgb value is normalized to [0,1] in the present invention, angle, θ according to
The axle in HSI space is weighed, in the range of the H component obtained can be normalized to [0,1] divided by 360 °.If the rgb value be given exists
In [0,1] is interval, then other two component S and I of HSI also can be in the range of normalizing be [0,1].Fig. 2 shows a RGB color figure
As being converted to the image conditions in HSI space.
Implement step B: in order to preferably protect image detail after enhancing, it is bright that the present invention uses genetic algorithm optimization to strengthen
Degree image, enhanced brightness contravariant again gains RGB color space.Gray level image I component is by the theoretical table further of Retinex
Being shown as luminance picture and inverse degree image, (x, y) represents low frequency component to luminance picture L in the picture, and (x's inverse image R y) exists
Image represents high fdrequency component, can be by changing luminance picture and reflected image in original image under conditions of constant color
Ratio reaches to strengthen the purpose of image.(x, y) uses gauss low frequency filter to assess luminance component to L, real in the present invention
Being the 2D discrete convolution of a band gaussian kernel on border, mathematical notation is as follows:
Wherein D (p, s) ≡ IP-Is, Ns is kernel function, represents the distance of pixel.
Implement step C: the present invention is directed to the brightness layer of image, it is proposed that a kind of Revised genetic algorithum parameters optimization is to increase
The algorithm of strong image.Basic process is: first carry out the conversion of color of image space, is transformed into HIS color from RGB color empty
Between；Then extract light intensity level, and brightness corresponding relation is encoded；Enhanced evaluation criterion is constructed for luminance component
Function, searches for optimal solution by genetic factor；Finally each component data is synthesized, it is achieved thereby that the enhancing of coloured image.Lose
Propagation algorithm flow process in image enhaucament is as follows:
C1 strengthens problem to nighttime image and describes, and finds the various corresponding pass of the brightness after color space conversion and pixel
System, establishes the chromosome in genetic algorithm；
C2 determines coded system, and the solution space of problem is mapped as the search volume of genes of individuals string and genetic algorithm；
C3 initializes population and scale, randomly generates a number of chromosome, forms initial population；
C4 sets up the parameter model of image enhaucament, it is possible to use the objective evaluation criteria of image enhaucament or function；
C5 carries out genetic manipulation, i.e. selects the image enhaucament model set up, intersects, mutation operator etc., produces new
Solve, be constantly iterated optimizing；
C6, according to end condition, repeats step C3 and step C4, until finding the optimal solution of objective evaluation function.
Each step to image enhaucament carries out analytic explanation below:
Establish chromosome coding: due to the robustness of genetic algorithm, it is the harshest to the requirement of coding.Genetic algorithm is not
It is that object of study is directly discussed, but by certain coding, unified for object imparting is arranged in certain sequence by special symbol
The string become.In order to strengthen the contrast of image, first the relation between nighttime image brightness and enhanced brightness of image is proposed,
Then utilize brightness degree to encode, in the present invention the brightness span of input picture is set to [0,255].Assume
The minimum brightness of image is Lmin, the high-high brightness of image is Lmax, minimum luminance value LminIt is mapped as 0 in the luminance frame strengthened,
Maximum brightness value LmaxIn the frame that highlights, be mapped as 255, the frame luminance dynamic range after the image enhaucament of i.e. 8 be [0,
255], shown in corresponding relation such as Fig. 3 (a).Wherein 0≤Lmin≤Lmax≤ 255, L (x, y) is input image lightness figure, and L ' (x, y)
For exporting the brightness of image after strengthening.
Chromosome is the underlying carrier of genetic manipulation, and its coding uses binary coding.Binary coding rule needs full
Foot: the encoding scheme designed by (1) should easily generate the short definition relevant to required problem away from and the gene block of low order.(2)
Designed encoding scheme should use minimum character set so that problem is naturally represented or describes.In the present invention, each
The increments of gene is set to Δ(i-1), go here and there long n=(Lmax-Lmin+ 1), it is expressed as Δn-1, Δn-2......Δ0, can pass through
Following expression formula calculates:
Loutput(i)=Loutput(i-1)+Δ(i-1)1≤i≤n(11)
Fig. 3 (b) shows frame of video at night and strengthens the brightness mapping relations of frame of video.Wherein i represents horizontal line brightness
Grade, LoutputRepresent the brightness degree of vertical line.Note when brightness value i is 0, the brightness L mapped outoutputI () value is also
0。
Analyzing according to above, the present invention provides enhancing brightness of image OL(i) and input image lightness corresponding pass between the two
It is as follows:
Therefore formula (12) its relation is corresponding input image lightness i and output brightness OL(i)。
Introduce fitness function: determine brightness corresponding relation, need to introduce image quality evaluation canonical function as suitable
Response function, and use genetic algorithm to carry out corresponding relation optimization.The present invention based on genetic algorithm at the image enhaucament of brightness layer,
Propose image quality evaluation standard, i.e. fitness function, then utilize the brightness corresponding relation that Genetic algorithm searching effect is optimum.
If (x, y) denotation coordination is that ((x, y) (x, image y) increases denotation coordination L ' for x, the brightness of original image y) to L
Brightness after Qiang, is first normalized.
M (x, y)=[L (x, y)-Lmin]/[Lmax-Lmin] (13)
Wherein Lmax, LminRepresent maximum and the minima of this luminance picture respectively, it is clear that have 0≤m (x, y)≤1, we
Definition non-linear transform function be m ' (x, y), 0≤i≤1, then can obtain
M ' (x, y)=OL[m (x, y)] (14)
Wherein 0≤m ' (x, y)≤1, then according to m ' (x, value y) can obtain export brightness of image L 'output(x, y).
Now utilizing genetic algorithm encoding and produce initial population, every chromosome comprises 1 gene section.Picture quality is commented
Price card is accurate, i.e. fitness function, and the strategic function of the evaluation image quality that the present invention proposes is as fitness function, and definition is such as formula
(15) shown in:
Wherein n=M × N, the width of M, N respectively enhancing image and height, chromosome i represents, if fitness function
The value of Function (i) is the biggest, then representing that the Luminance Distribution of image is the most uniform, the image effect of enhancing is the best.After strengthening
Luminance picture renormalization process, just obtain output strengthen image L ' (x, y), namely:
Genetic manipulation: genetic manipulation includes selecting, and intersects and the 3 kinds of basic operators that make a variation.Selection opertor effect is basis
Individual good and bad degree determines that it is eliminated the next generation or is replicated, and is used for guaranteeing convergence, and convergence exploitation can be more
Good balance mutation operation.Conventional selection mode includes: roulette selection, tournament selection, stable state replicate, sequence becomes with ratio
Change, league matches etc..Crossover operator is to choose for breeding in follow-on individuality, to two different individual same positions
Gene swaps, thus produces new individuality.Generally include single-point exchange, exchange, multiple spot exchange, uniform crossover etc. at 2.
Using the selection mode of roulette model in the present invention, its ultimate principle is that the ratio according to each chromosome adaptive value is come really
The select probability of this individuality fixed or survival probability.Therefore a roulette model can be set up to represent these probability.Select
Process is exactly rotation roulette (number of times is equal to population scale) several times, selects body one by one for new population every time.This choosing of wheel disc
The feature of selection method is exactly random sampling procedure.A pair parent is selected to have may producing of greater value effectively to intersect.
Mutation operation is to change the genic value on some gene location of the individual string in population, highlighted for guaranteeing
The Different Individual of degree.Being applied first to new structure individual, each random element representation integer is individual.In genetic manipulation,
If mutation probability strengthens, search is induced sweat and will not be absorbed in Local Minimum, but may destroy good individuality.
Stopping rule: according to the step of C4, it is proposed that the adaptive response function of genetic algorithm, initializes with one random
Gene information be given in the first generation.For stopping rule, the condition of stopping determines according to the brightness degree that image is final, than
Brightness degree such as the image of 8 bits is at most set to 256, it is also possible to stop with certain algebraically.In the present invention we
In conjunction with brightness degree and in constant 12 generation genetic algorithm stop be as the criterion, that index first reaches, and the most first stops.Arrange initial
Body number is 120 (i.e. brightness of image grades), and follow-on individual survival rate is set to 65%, and aberration rate is set to 0.01%.
Implement step D: after enhancing, the reconstruction of coloured image is the focus of image processing field one research.The present invention
Weight is carried out finally utilizing enhanced image luminance information derived above, the colouring information of nighttime image, half-tone information etc.
Structure, finally gives enhanced coloured image.
Claims (9)
1. the nighttime image enhancing method of a Revised genetic algorithum, it is characterised in that first RGB color is transformed into
HSI color space, then extracts the gray scale I component in HSI color space, owing to the present invention is to strengthen, so carrying based on brightness
Further using Retinex theoretical decomposition after taking I component is inverse image R and luminance picture L, and Contrast enhanced nighttime image is bright
Degree is completed by genetic algorithm optimization, and the nighttime image finally strengthened is by the brightness strengthened, inverse image and nighttime image face
Color reconstruct, proposition based on Revised genetic algorithum strengthen nighttime image, its essence is real image to be strengthened problem be converted into
Genetic algorithm parameter optimization problem, the basic step of this algorithm includes:
Step 101: change RGB color picture format into HSI color space image form；
Step 102: extract I component from HSI color space, using Retinex theoretical decomposition I component is inverse image and brightness
Image；
Step 103: the luminance picture at night extracted is carried out Contrast enhanced based on genetic algorithm；
Step 104: use the inverse image at night, highlights image and nighttime image color carries out image reconstruction.
The nighttime image enhancing method of a kind of Revised genetic algorithum the most according to claim 1, it is characterised in that described
Step 101 is mainly based upon gray level image in view of current picture superposition technology and processes, if coloured image increases
The tetanic use gray level image that connects strengthens algorithm, and result is undesirable；If color space is changed, can be relevant three
Component is converted in incoherent chrominance space go, and enhanced result is preferable, is found by analysis, and HSI color model can be more
Good meets this demand, assumes initially that the low-light level RGB color image of input is determined by three tones of red, green, blue, from RGB
Color space conversion is to HSI color space, and each concrete component solves as follows, and H component can obtain by equation below:
Saturation component S is given by:
Last luminance component I is given by:
Wherein H represents colourity (Hue), and S represents saturation (Saturation), and I represents brightness value (Intensity), is strengthening
During, colourity and saturation keep constant, directly strengthen the luminance component of coloured image, it is ensured that do not have color displacement, more
The requirement of human eye can be met, it is to note that in the range of supposing that rgb value is normalized to [0,1] in the present invention, angle, θ is according to HSI space
Axle weigh, in the range of the H component obtained can be normalized to [0,1] divided by 360 °, if the rgb value be given is in [0,1] district
In, then other two component S and I of HSI also can be in the range of normalizing be [0,1].
The nighttime image enhancing method of a kind of Revised genetic algorithum the most according to claim 1, it is characterised in that described
Step 102 extracts I component from HSI color space, and uses Retinex theoretical decomposition I component, in order to preferably protect increasing
Image detail after strong, the present invention uses genetic algorithm optimization to highlight image, and enhanced brightness contravariant again gains RGB color
Space, gray level image I component is further represented as luminance picture and inverse degree image, luminance picture L by Retinex theory
(x, y) represents low frequency component in the picture, and (x y) represents high fdrequency component, in the condition of constant color to inverse image R in the picture
Can reach to strengthen the purpose of image by change luminance picture and reflected image ratio in original image, in the present invention down
(x, y) uses gauss low frequency filter to assess luminance component to L, the 2D discrete convolution of an actually band gaussian kernel, mathematics
It is expressed as follows:
Wherein D (p, s) ≡ Ip-Is, Ns is kernel function, represents the distance of pixel.
The nighttime image enhancing method of a kind of Revised genetic algorithum the most according to claim 1, it is characterised in that step
The brightness layer of 103 pairs of images, it is proposed that a kind of Revised genetic algorithum parameters optimization is to strengthen the algorithm of image, basic process
It is: first carry out the conversion of color of image space, be transformed into HIS color space from RGB color；Then extract light intensity level,
And brightness corresponding relation is encoded；Construct enhanced evaluation criterion function for luminance component, searched by genetic factor
Rope optimal solution；Finally being synthesized by each component data, it is achieved thereby that the enhancing of coloured image, genetic algorithm is in image enhaucament
Flow process is as follows:
Step 201: nighttime image is strengthened problem and describes, finds the various corresponding pass of the brightness after color space conversion and pixel
System, establishes the chromosome in genetic algorithm；
Step 202: determine coded system, is mapped as the search volume of genes of individuals string and genetic algorithm by the solution space of problem；
Step 203: initialize population and scale, randomly generates a number of chromosome, forms initial population；
Step 204: set up the parameter model of image enhaucament, it is possible to use the objective evaluation criteria of image enhaucament or function；
Step 205: carry out genetic manipulation, i.e. selects the image enhaucament model set up, intersects, mutation operator etc., produces
New explanation, is constantly iterated optimizing；
Step 206: according to end condition, repeats step 203 and step 204, until finding the optimal solution of objective evaluation function.
The nighttime image enhancing method of a kind of Revised genetic algorithum the most according to claim 4, it is characterised in that described
Step 201, in order to strengthen the contrast of image, first proposes the relation between nighttime image brightness and enhanced brightness of image,
Then utilize brightness degree to encode, in the present invention the brightness span of input picture be set to [0,255], it is assumed that
The minimum brightness of image is Lmin, the high-high brightness of image is Lmax, minimum luminance value LminIt is mapped as 0 in the luminance frame strengthened,
Maximum brightness value LmaxIn the frame that highlights, be mapped as 255, the frame luminance dynamic range after the image enhaucament of i.e. 8 be [0,
255],
Chromosome is the underlying carrier of genetic manipulation, and its coding uses binary coding, and binary coding rule needs to meet:
(1) encoding scheme designed by should easily generate the short definition relevant to required problem away from and the gene block of low order, (2) are set
The encoding scheme of meter should use minimum character set so that problem is naturally represented or is described, in the present invention, each gene
Increments be set to Δ(i-1), go here and there long n=(Lmax-Lmin+ 1), it is expressed as Δn-1, Δn-2......Δ0, can be by as follows
Expression formula calculate:
Loutput(i)=Loutput(i-1)+Δ(i-1)1≤i≤n (8)
Analyzing according to above, the present invention provides enhancing brightness of image OLI () and input image lightness corresponding relation between the two are such as
Under:
Therefore formula (12) its relation is corresponding input image lightness i and output brightness OL(i)。
The nighttime image enhancing method of a kind of Revised genetic algorithum the most according to claim 4, it is characterised in that described
Step 204 at the image enhaucament of brightness layer, proposes image quality evaluation standard, i.e. fitness function, then based on genetic algorithm
Utilize the brightness corresponding relation that Genetic algorithm searching effect is optimum,
If (x, y) denotation coordination is that ((x, y) (x, after image enhaucament y) for denotation coordination for L ' for x, the brightness of original image y) to L
Brightness, be first normalized,
M (x, y)=[L (x, y)-Lmin]/[Lmax-Lmin] (10)
Wherein Lmax, LminRepresent maximum and the minima of this luminance picture respectively, it is clear that have 0≤m (x, y)≤1, we definition
Non-linear transform function be m ' (x, y), 0≤i≤1, then can obtain
M ' (x, y)=OL[m (x, y)] (11)
Wherein 0≤m ' (x, y)≤1, then according to m ' (x, value y) can obtain export brightness of image L 'output(x, y),
Now utilizing genetic algorithm encoding and produce initial population, every chromosome comprises 1 gene section, image quality evaluation mark
Standard, i.e. fitness function, the strategic function of the evaluation image quality that the present invention proposes is as fitness function, and definition is such as formula (15)
Shown in:
Wherein n=M × N, the width of M, N respectively enhancing image and height, chromosome i represents, if fitness function
The value of Function (i) is the biggest, then representing that the Luminance Distribution of image is the most uniform, the image effect of enhancing is the best, after strengthening
Luminance picture renormalization process, just obtain output strengthen image L ' (x, y), namely:
。
The nighttime image enhancing method of a kind of Revised genetic algorithum the most according to claim 4, it is characterised in that described
Step 205 uses the selection mode of roulette model as selection opertor, and its ultimate principle is according to each chromosome adaptive value
Ratio determine select probability or the survival probability of this individuality, therefore can set up a roulette model to represent that these are general
Rate, the process of selection is exactly rotation roulette (number of times is equal to population scale) several times, selects body one by one for new population, wheel every time
The feature of this system of selection of dish is exactly random sampling procedure, selects a pair parent to have may producing of greater value effectively to hand over
Fork,
Mutation operation is to change the genic value on some gene location of the individual string in population, is used for guaranteeing high brightness
Different Individual, is applied first to new structure individual, and each random element representation integer is individual, in genetic manipulation, if
Mutation probability strengthens, and search is induced sweat and will not be absorbed in Local Minimum, but may destroy good individuality.
The nighttime image enhancing method of a kind of Revised genetic algorithum the most according to claim 4, it is characterised in that described
Step 204 proposes the adaptive response function of genetic algorithm, initializes and is given in the first generation with a random gene information, for
Stopping rule, the condition of stopping determines according to the brightness degree that image is final, and the brightness degree of the image of such as 8 bits is most
It is set to 256, it is also possible to stop with certain algebraically, loses in we combine brightness degree and constant 12 generation in the present invention
Propagation algorithm stops being as the criterion, and that index first reaches, and the most first stops, and arranging initial individual amount is 120 (i.e. brightness of image etc.
Level), follow-on individual survival rate is set to 65%, and aberration rate is set to 0.01%.
The nighttime image enhancing method of a kind of Revised genetic algorithum the most according to claim 4, it is characterised in that described
Step 104 uses enhanced image luminance information, the colouring information of nighttime image, half-tone information etc. to be reconstructed, it is thus achieved that increase
Image after Qiang.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610710468.9A CN106327443A (en) | 2016-08-24 | 2016-08-24 | Night image enhancement method based on improved genetic algorithm |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610710468.9A CN106327443A (en) | 2016-08-24 | 2016-08-24 | Night image enhancement method based on improved genetic algorithm |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN106327443A true CN106327443A (en) | 2017-01-11 |
Family
ID=57742310
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201610710468.9A Pending CN106327443A (en) | 2016-08-24 | 2016-08-24 | Night image enhancement method based on improved genetic algorithm |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN106327443A (en) |
Cited By (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106997744A (en) * | 2017-03-15 | 2017-08-01 | 广东欧珀移动通信有限公司 | The control method and control device of screen intensity |
| CN108509487A (en) * | 2018-02-08 | 2018-09-07 | 杨睿嘉 | Image search method, equipment and the storage medium of cortex model are provided based on pulse |
| CN109003243A (en) * | 2018-07-20 | 2018-12-14 | 广州市普汉科技有限公司 | A kind of anaglyph image processing method for laser engraving |
| CN109242795A (en) * | 2018-08-30 | 2019-01-18 | 福建师范大学 | A kind of brightness enhancement of low-light level human tissue cell two-photon micro-image |
| CN109359531A (en) * | 2018-09-12 | 2019-02-19 | 仲恺农业工程学院 | Fruit harvesting area automatic positioning method facing natural scene |
| CN109886901A (en) * | 2019-03-21 | 2019-06-14 | 电子科技大学 | A kind of nighttime image enhancing method decomposed based on multi-path |
| CN109977814A (en) * | 2019-03-13 | 2019-07-05 | 武汉大学 | A kind of AdaBoost pedestrian detection method based on unification LBP |
| CN113012051A (en) * | 2019-12-20 | 2021-06-22 | 江西晶润光学有限公司 | Image processing method, image processing device, electronic equipment and computer readable storage medium |
| CN114282581A (en) * | 2021-01-29 | 2022-04-05 | 北京有竹居网络技术有限公司 | Training sample obtaining method and device based on data enhancement and electronic equipment |
| CN118447026A (en) * | 2024-07-08 | 2024-08-06 | 天津德通电气有限公司 | Mineral step processing data processing method based on improved image processing |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1731451A (en) * | 2005-08-22 | 2006-02-08 | 上海广电（集团）有限公司中央研究院 | Method of image color enhancement |
| CN101951523A (en) * | 2010-09-21 | 2011-01-19 | 北京工业大学 | Adaptive colour image processing method and system |
| CN103903229A (en) * | 2014-03-13 | 2014-07-02 | 中安消技术有限公司 | Night image enhancement method and device |
| CN104123698A (en) * | 2013-04-25 | 2014-10-29 | 北京信路威科技股份有限公司 | Night image enhancement method applied to intelligent integrated camera |
-
2016
- 2016-08-24 CN CN201610710468.9A patent/CN106327443A/en active Pending
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1731451A (en) * | 2005-08-22 | 2006-02-08 | 上海广电（集团）有限公司中央研究院 | Method of image color enhancement |
| CN101951523A (en) * | 2010-09-21 | 2011-01-19 | 北京工业大学 | Adaptive colour image processing method and system |
| CN104123698A (en) * | 2013-04-25 | 2014-10-29 | 北京信路威科技股份有限公司 | Night image enhancement method applied to intelligent integrated camera |
| CN103903229A (en) * | 2014-03-13 | 2014-07-02 | 中安消技术有限公司 | Night image enhancement method and device |
Non-Patent Citations (3)
| Title |
|---|
| YUNBO RAO 等: "Illumination-based nighttime video contrast enhancement using genetic algorithm", 《MULTIMED TOOLS APPL (2014) 》 * |
| 饶云波: "夜间视频增强的关键技术研究", 《中国博士学位论文全文数据库 信息科技辑》 * |
| 黄楠: "遗传算法在图像增强中的应用研究", 《计算机仿真》 * |
Cited By (16)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106997744A (en) * | 2017-03-15 | 2017-08-01 | 广东欧珀移动通信有限公司 | The control method and control device of screen intensity |
| CN108509487A (en) * | 2018-02-08 | 2018-09-07 | 杨睿嘉 | Image search method, equipment and the storage medium of cortex model are provided based on pulse |
| CN108509487B (en) * | 2018-02-08 | 2022-09-30 | 杨睿嘉 | Image retrieval method, device and storage medium based on pulse-issued cortex model |
| CN109003243A (en) * | 2018-07-20 | 2018-12-14 | 广州市普汉科技有限公司 | A kind of anaglyph image processing method for laser engraving |
| CN109003243B (en) * | 2018-07-20 | 2021-10-29 | 广州市普汉科技有限公司 | Relief effect image processing method for laser engraving |
| CN109242795B (en) * | 2018-08-30 | 2021-09-07 | 福建师范大学 | Brightness enhancement method of low-brightness human tissue cell two-photon microscopic image |
| CN109242795A (en) * | 2018-08-30 | 2019-01-18 | 福建师范大学 | A kind of brightness enhancement of low-light level human tissue cell two-photon micro-image |
| CN109359531A (en) * | 2018-09-12 | 2019-02-19 | 仲恺农业工程学院 | Fruit harvesting area automatic positioning method facing natural scene |
| CN109359531B (en) * | 2018-09-12 | 2021-12-14 | 仲恺农业工程学院 | Fruit harvesting area automatic positioning method facing natural scene |
| CN109977814A (en) * | 2019-03-13 | 2019-07-05 | 武汉大学 | A kind of AdaBoost pedestrian detection method based on unification LBP |
| CN109886901A (en) * | 2019-03-21 | 2019-06-14 | 电子科技大学 | A kind of nighttime image enhancing method decomposed based on multi-path |
| CN113012051A (en) * | 2019-12-20 | 2021-06-22 | 江西晶润光学有限公司 | Image processing method, image processing device, electronic equipment and computer readable storage medium |
| CN114282581A (en) * | 2021-01-29 | 2022-04-05 | 北京有竹居网络技术有限公司 | Training sample obtaining method and device based on data enhancement and electronic equipment |
| CN114282581B (en) * | 2021-01-29 | 2023-10-13 | 北京有竹居网络技术有限公司 | Training sample acquisition method and device based on data enhancement and electronic equipment |
| CN118447026A (en) * | 2024-07-08 | 2024-08-06 | 天津德通电气有限公司 | Mineral step processing data processing method based on improved image processing |
| CN118447026B (en) * | 2024-07-08 | 2024-09-20 | 天津德通电气有限公司 | Mineral step processing data processing method based on improved image processing |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN106327443A (en) | Night image enhancement method based on improved genetic algorithm | |
| CN109559276B (en) | Image super-resolution reconstruction method based on quality evaluation and feature statistics | |
| CN109934154B (en) | Remote sensing image change detection method and detection device | |
| CN108681991A (en) | Based on the high dynamic range negative tone mapping method and system for generating confrontation network | |
| CN101151883B (en) | Display image processing method, image processing method and device, and integrated circuit containing the image processing device | |
| Rahman et al. | Tripartite sub-image histogram equalization for slightly low contrast gray-tone image enhancement | |
| CN109712165A (en) | A kind of similar foreground picture image set dividing method based on convolutional neural networks | |
| Ozdemir et al. | Fuzzy algorithms for combined quantization and dithering | |
| CN105118067A (en) | Image segmentation method based on Gaussian smoothing filter | |
| CN110120034B (en) | Image quality evaluation method related to visual perception | |
| CN103971340A (en) | High-bit-width digital image dynamic range compression and detail enhancement method | |
| CN106971379A (en) | A kind of underwater picture Enhancement Method merged based on stratified calculation | |
| CN111768326B (en) | High-capacity data protection method based on GAN (gas-insulated gate bipolar transistor) amplified image foreground object | |
| CN105913408A (en) | Multi-focusing-image fusion method based on bilateral filter | |
| Li et al. | An adaptive enhancement method for low illumination color images | |
| CN109816646A (en) | A kind of non-reference picture quality appraisement method based on degeneration decision logic | |
| CN115908602A (en) | Style migration method for converting landscape photos into Chinese landscape paintings | |
| Subramani et al. | Cuckoo search optimization‐based image color and detail enhancement for contrast distorted images | |
| CN1992782B (en) | Image processing device and image processing method | |
| CN105844640A (en) | Color image quality evaluation method based on gradient | |
| CN117935079B (en) | Remote sensing image fusion method, system and readable storage medium | |
| CN106204476B (en) | A kind of low-luminance color image enchancing method | |
| Oza et al. | Automatic image colorization using ensemble of deep convolutional neural networks | |
| Yuan et al. | Color image quality assessment with multi deep convolutional networks | |
| CN115439361B (en) | Underwater image enhancement method based on self-countermeasure generation countermeasure network |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| WD01 | Invention patent application deemed withdrawn after publication | ||
| WD01 | Invention patent application deemed withdrawn after publication |
Application publication date: 20170111 |