CN108305234B - Double-histogram equalization method based on optimization model - Google Patents
Double-histogram equalization method based on optimization model Download PDFInfo
- Publication number
- CN108305234B CN108305234B CN201810044231.0A CN201810044231A CN108305234B CN 108305234 B CN108305234 B CN 108305234B CN 201810044231 A CN201810044231 A CN 201810044231A CN 108305234 B CN108305234 B CN 108305234B
- Authority
- CN
- China
- Prior art keywords
- histogram
- lookup table
- image
- lut
- temporary
- 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.)
- Active
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T5/00—Image enhancement or restoration
- G06T5/40—Image enhancement or restoration by the use of histogram techniques
Abstract
The invention provides a double histogram equalization method based on an optimization model, which comprises the following steps: firstly, dividing an initial histogram of an input image into two sub-histograms; then, according to the proposed optimization model, determining the optimal dynamic range of the sub-histogram by adopting a traversal optimization method; finally, carrying out double-histogram equalization to obtain a result image; the invention provides an image enhancement technology with wide adaptation range, which can adaptively improve the definition, brightness and contrast of images in various environments.
Description
Technical Field
The invention relates to the field of video image enhancement, in particular to a double histogram equalization method based on model control.
Background
The histogram of an image is a statistical graph obtained by counting image pixel values, contains distribution characteristics of image brightness, and is often used as a tool for image enhancement, thereby generating a plurality of correlation algorithms, and histogram equalization is the most traditional method among the methods, and the image is optimized according to the distribution characteristics of the histogram, so that the histogram tends to be uniformly distributed, and the contrast of the image is enhanced, so that the image is clearer. However, the traditional histogram equalization algorithm has the problems of over-strong saturation, detail loss and the like, so that a great number of improved algorithms are proposed by later people and can be mainly divided into two types.
One type is called local histogram equalization, introduces the spatial characteristics of the image, divides the image into a plurality of overlapped or non-overlapped areas, and respectively carries out independent histogram equalization, thereby improving the defects of the traditional histogram equalization, being capable of obtaining better processing results, but increasing the complexity of calculation and being difficult to be applied to actual production.
The other type is called multi-histogram equalization, distribution characteristics of the histograms are further analyzed, a histogram equalization method based on a threshold value is provided, the histogram is divided into two sub-histograms (double histogram equalization) or a plurality of sub-histograms (multi-histogram equalization) to be equalized respectively, and therefore the result of brightness maintenance or histogram feature maintenance is obtained. The method is low in calculation complexity and suitable for video real-time processing and hardware integration, but the algorithm still has many defects and is narrow in application range.
Disclosure of Invention
The invention mainly aims to overcome the defects in the prior art and provides an optimized model-based double histogram equalization method which can adaptively enhance images in various shooting environments and improve the definition, brightness and contrast of the images.
The invention adopts the following technical scheme:
a dual histogram equalization method based on optimized model is characterized by inputting digital image with bit B and level L2 ^ B, obtaining brightness image, counting initial histogram of the brightness image, dividing point S according to preset datadsDividing the initial histogram into a left histogram and a right histogram, and respectively carrying out normalization processing and accumulation and operation to obtain a left accumulation array CL(j) And right accumulation array CR(k) J is equal to or less than data dividing point SdsK is greater than the data dividing point SdsThe gray level of (a); segmenting points S from a predetermined dynamic rangersInitial value L ofmin+1 Start, LminFor minimum gray level, S is performedrsTraversing operation:
1) according to the current SrsValue, combined with left accumulation array CL(j) And right accumulation array CR(k) Calculating to obtain an integer lookup table, looking up the initial histogram to obtain a temporary histogram, and dividing the point S according to the dynamic rangersDividing the histogram into a temporary left histogram and a temporary right histogram;
2) respectively calculating corresponding weighted average probability densities of the temporary left histogram and the temporary right histogram, and performing weighted summation on the two obtained weighted average probability densities according to the set weight to obtain a total weighted average probability density sum;
3) storing the calculated sum of the weighted average probability densities into the S-th arrayrsIn each element, the length of the array is L, let Srs＝Srs+1, if Srs＜Lmax，LmaxReturning to step 1) for maximum gray level, if S isrs≥LmaxStep 4) is entered;
4) s corresponding to the minimum value of the array in the step 3) by the total weighted average probability densityrsAnd dividing the initial histogram as an optimal division point, calculating a lookup table to obtain an output lookup table, and performing lookup operation on each numerical value of the initial brightness image to obtain a final result image.
And if the digital image is a multi-channel image, taking a brightness channel of the digital image as the brightness image, and if the digital image is a single-channel image, directly taking the brightness channel as the brightness image.
The data dividing point SdsBy a maximum between-class variance algorithm or a one-dimensional maximum entropy algorithm or using the mean, median or inflection point of the initial histogram of the luminance image.
In the step 1), the left accumulation array C is combinedL(k) And right accumulation array CR(k) Calculating to obtain an integer lookup table, specifically: firstly, according to left accumulation array CL(j) And right accumulation array CR(k) Respectively calculating the left lookup table LutL(j) And right lookup table LutR(k) Is of the formula
LutL(j)＝(Srs-Lmin)*CL(j)+Lmin
LutR(k)＝(Lmax-Srs)*CR(k)+Srs
Wherein L ismin≤j≤Sds，Sds<k≤Lmax(ii) a Then merging the left lookup table LutL(j) And right lookup table LutR(k) And obtaining the integer lookup table.
In the step 2), the average probability density is calculated in the following manner: the pre-emphasis index r is defined,
r is more than or equal to 0 and less than or equal to 1, and a temporary left histogram is calculatedThe r-th power of each value in the graph is then summed and divided by SrsObtaining the weighted average probability density of the temporary left histogram; calculating the r power of each value in the temporary right histogram, then summing and dividing by L-SrsAnd obtaining the weighted average probability density of the temporary right histogram.
In the step 2), the weight is the ratio of the dynamic ranges of the left histogram and the right histogram in the step 1).
In the step 4), the calculation lookup table specifically includes: according to left accumulation array CL(j) And right accumulation array CR(k) Respectively calculating the left lookup table LutL(j) And right lookup table LutR(k) Is of the formula
LutL(j)＝(Srs-Lmin)*CL(j)+Lmin
LutR(k)＝(Lmax-Srs)*CR(k)+Srs
Wherein L ismin≤j≤Sds，Sds<k≤Lmax(ii) a Then merge the left lookup table LutL(j) And right lookup table LutR(k) And obtaining the output lookup table.
In the step 4), if the initially input digital image is a color image, the final result is restored to the color image.
As can be seen from the above description of the present invention, compared with the prior art, the present invention has the following advantages:
the invention inherits the advantages of multi-histogram equalization, simultaneously corrects the defects of the multi-histogram equalization and increases the practicability of the multi-histogram algorithm. In addition, the invention provides a new optimization model, so the method has strict mathematical theory support and has better robustness and visual effect.
Drawings
FIG. 1 is a flow chart of the present invention;
FIG. 2(a) is an input grayscale map;
FIG. 2(b) is a graph of the enhancement results of FIG. 2 (a);
FIG. 3(a) shows HSV spatial luminance channels for an input color low illumination image;
FIG. 3(b) is a graph of the enhancement results of FIG. 3 (a);
FIG. 4(a) shows HSV spatial luminance channels for input color HDMA images;
fig. 4(b) is a graph of the enhancement result of fig. 4 (a).
Detailed Description
The invention is further described below by means of specific embodiments.
Fig. 1, a dual histogram equalization method based on an optimized model according to the present invention, includes the following steps:
and inputting a digital image with the bit number of B and the stage number of L2 ^ B, wherein if the digital image is a multi-channel image, the brightness channel of the digital image is taken as a brightness image, and if the digital image is a single-channel image, the digital image is directly taken as a brightness image. Counting the initial histogram of the brightness image, and segmenting points S according to preset datadsDividing the initial histogram into a left histogram and a right histogram, and respectively carrying out normalization processing and accumulation and operation to obtain a left accumulation array CL(j) And right accumulation array CR(k) J is equal to or less than data dividing point SdsK is greater than the data dividing point SdsThe gray scale of (2). Wherein, the data dividing point SdsBy a maximum between-class variance algorithm or a one-dimensional maximum entropy algorithm or using the mean, median or inflection point of the initial histogram of the luminance image.
1) Segmenting points S from a predetermined dynamic rangersInitial value L ofmin+1 Start, LminFor minimum gray level, S is performedrsTraversing operation: according to the current SrsValue, combined with left accumulation array CL(j) And right accumulation array CR(k) Calculating to obtain an integer lookup table, looking up the initial histogram to obtain a temporary histogram, and dividing the point S according to the dynamic rangersInto a temporary left histogram and a temporary right histogram. The method specifically comprises the following steps: firstly, according to left accumulation array CL(j) And right accumulation array CR(k) Respectively calculating the left lookup table LutL(j) And right lookup table LutR(k) The formula is as follows:
LutL(j)＝(Srs-Lmin)*CL(j)+Lmin
LutR(k)＝(Lmax-Srs)*CR(k)+Srs
wherein SLmin≤j≤Sds，Sds<k≤Lmax(ii) a Then merging the left lookup table LutL(j) And right lookup table LutR(k) And obtaining the integer lookup table.
2) Respectively calculating corresponding weighted average probability densities of the temporary left histogram and the temporary right histogram in the following calculation mode: defining a pre-emphasis index r, wherein r is more than or equal to 0 and less than or equal to 1, calculating the r power of each numerical value in the temporary left histogram, summing, and dividing by SrsObtaining the weighted average probability density of the temporary left histogram; calculating the r power of each value in the temporary right histogram, then summing and dividing by L-SrsAnd obtaining the weighted average probability density of the temporary right histogram. And according to the set weight, carrying out weighted summation on the two obtained weighted average probability densities to obtain a total weighted average probability density sum. The weight is taken as the ratio of the dynamic ranges of the left histogram and the right histogram in step 1).
3) Storing the calculated sum of the weighted average probability densities into an array of length LrsIn each element, let Srs＝Srs+1, if Srs＜LmaxGo back to step 1), if Srs≥LmaxStep 4) is entered.
4) S corresponding to the minimum value of the array in the step 3) by the total weighted average probability densityrsAnd dividing the initial histogram as an optimal division point, calculating a lookup table to obtain an output lookup table, and performing lookup operation on each numerical value of the initial brightness image to obtain a final result image. The calculation lookup table is similar to the method in the step 2), and specifically comprises the following steps: according to left accumulation array CL(j) And right accumulation array CR(k) Respectively calculating the left lookup table LutL(j) And right lookup table LutR(k) Is of the formula
LutL(j)＝(Srs-Lmin)*CL(j)+Lmin
LutR(k)＝(Lmax-Srs)*CR(k)+Srs
Wherein L ismin≤j≤Sds，Sds<k≤Lmax(ii) a Then merge the left lookup table LutL(j) And right lookup table LutR(k) And obtaining the output lookup table.
In addition, if the digital image initially input is a color image, the final result is restored to the color image.
Examples of the applications
Assuming an RGB three-channel color digital image with an input bit B of 8 and a number L of 256, the processing steps according to the flowchart should be as follows:
because the input is a multi-channel image, a luminance image thereof should be obtained for input, for example, the image is converted into an HSV color space, and then a luminance channel V thereof is taken as a luminance image, at which time, the maximum gray level L of the image ismax255, minimum gray level LminThe initial histogram of the luminance image is counted as 0.
Setting histogram data dividing point SdsThe setting mode of the segmentation point is not limited, but the point is obtained by adopting an OTSU algorithm by default, the initial histogram is divided into a left histogram and a right histogram by taking the gray level equal to the data segmentation point as a boundary, the gray level of the segmentation point can be included in the left histogram, and then normalization operation is carried out, namely, each value of the left histogram and the right histogram is divided by the total number of pixels of the brightness image to obtain a left normalized histogram and a right normalized histogram.
Respectively accumulating and operating the left normalized histogram and the right normalized histogram, wherein the specific method comprises the following steps: left cumulative sum array CL(j) Wherein j is a gray level of not more than a data division point, CL(j) Should be the left normalized histogram from L in gray scaleminThe result of the accumulation to gray level j; right cumulative sum array CR(k) In (d), k is a gray level greater than the data division point, CR(k) Should be right normalized histogram in gray scale SdsTo the result of the gray level k accumulation.
1) Setting dynamic Range cut Point SrsOf (2) is initiatedValue of L by defaultmin+1, to SrsGo through the traversal according to CL(j) And CR(k) Calculating a left lookup table and a right lookup table, wherein the specific formula is as follows:
LutL(j)＝(Srs-Lmin)*CL(j)+Lmin(Lmin≤j≤Sds)----(1)
LutR(k)＝(Lmax-Srs)*CR(k)+Srs(Sds<k≤Lmax)----(2)
and splicing the left lookup table and the right lookup table according to the order of left and right, and then removing the decimal part of each element in the tables to obtain a complete integer lookup table, wherein the lookup table represents the mapping relation of pixel values before and after image processing.
The method comprises the following steps of looking up a table of an input image histogram to obtain a temporary histogram, and specifically comprises the following steps: according to the obtained mapping relation of the complete lookup table, the amplitude values of all gray levels of the initial histogram are moved to the gray level mapped by the complete lookup table, if more than one amplitude value is moved to the same gray level, the amplitude values are added to obtain the final amplitude value of the gray level, after the calculation is finished, the temporary histogram is divided by the total number of pixels of the brightness image to obtain a normalized temporary histogram, and the gray level S is used for obtaining the normalized temporary histogramrsAnd dividing the histogram into a normalized temporary left histogram and a normalized temporary right histogram.
2) Defining a data pre-emphasis index r (r is more than or equal to 0 and less than or equal to 1), setting the default value to be 0.5, calculating the r power of each numerical value in the normalized temporary left histogram, then summing and dividing by SrsObtaining a left weighted average probability density; calculating the r power of each value in the normalized temporary right histogram, then summing and dividing by L-SrsAnd obtaining the right weighted average probability density.
Based on the segmented left histogram dynamic range (i.e., S)dsSubtracting the minimum gray level of the left histogram) and dividing by the dynamic range of the initial histogram (namely subtracting the minimum gray level from the maximum gray level) to obtain a left weight; then according to the dynamic range of the divided right histogram (namely, subtracting S from the maximum gray level of the right histogram)ds) Dividing the dynamic range of the initial histogram by the dynamic range of the initial histogram to obtain a right weight; multiplying the left weight by the left weightAnd (4) the average probability density, the right weight is multiplied by the right weighted average probability density, and the two are added to obtain the sum of the total weighted average probability density.
3) Storing the calculated sum of the weighted average probability densities into an array of length LrsIn each element, let Srs＝Srs+1, if Srs＜LmaxGo to step 1), if Srs≥LmaxStep 4) is entered.
4) S corresponding to minimum value of total weighted average probability density sumrsAnd (3) as the optimal dynamic range division point, calculating the left lookup table and the right lookup table by using the formulas (1) and (2) again, splicing the left lookup table and the right lookup table according to the sequence of left and right, and removing the decimal part of each element in the table to obtain an output lookup table.
Performing table lookup operation on each pixel value of the initial brightness image by using an output lookup table, and mapping each pixel value of the brightness image to a new value through the lookup table to obtain a final result image; if the original input is a color image, the result image is restored to be a color image, for example, the V-channel image of the HSV space should be replaced with the result image.
For example, fig. 2(a), 3(a) and 4(a) are original drawings, and fig. 2(b), 3(b) and 4(b) are luminance channels of the algorithm processing results proposed by the present invention; in actual operation, fig. 2(a) is a single-channel diagram, which can be directly input as a luminance diagram, and fig. 3(a) and 4(a) are luminance diagrams of RGB color images, and the luminance component of HSV color space thereof is taken as a luminance diagram for processing; setting data pre-emphasis index r to 0.5, number of levels L to 256, and minimum gray level Lmin0, maximum gray level Lmax255; the total weighted probability density array calculated according to the steps 1) to 4) is a concave function, only one minimum value, namely the minimum value, is taken, and S corresponding to the minimum value is takenrsAs the final dynamic range division point, reestablishing the lookup table according to the step 5); finally, table look-up is carried out according to the numerical value of each pixel of the brightness image to obtain a final processing result; for the cases of fig. 3(a) and fig. 4(a), the processing result graph should be restored to a three-channel color image according to the rules of HSV space. The processing results show that the method of the invention is suitable for three different classesThe processing results of the type pictures are natural, rich in details and strong in self-adaptive capacity.
The invention provides a double-histogram equalization method based on an optimization model, which adaptively controls the dynamic range of double-histogram equalization by calculating weighted average probability density and improves the adaptive surface of the algorithm; the algorithm provided by the invention is simple and rapid, is suitable for video real-time processing and hardware integration, and has great market value.
The above description is only an embodiment of the present invention, but the design concept of the present invention is not limited thereto, and any insubstantial modifications made by using the design concept should fall within the scope of infringing the present invention.
Claims (7)
1. A dual histogram equalization method based on optimized model is characterized by inputting digital image with bit B and level L2 ^ B, obtaining brightness image, counting initial histogram of the brightness image, dividing point S according to preset datadsDividing the initial histogram into a left histogram and a right histogram, and respectively carrying out normalization and accumulation and operation to obtain a left accumulation array CL(j) And right accumulation array CR(k) J is equal to or less than data dividing point SdsK is greater than the data dividing point SdsThe gray level of (a); segmenting points S from a predetermined dynamic rangersInitial value L ofmin+1 Start, LminFor minimum gray level, S is performedrsTraversing operation:
1) according to the current SrsValue, combined with left accumulation array CL(j) And right accumulation array CR(k) Calculating to obtain an integer lookup table, looking up the initial histogram to obtain a temporary histogram, and dividing the point S according to the dynamic rangersDividing the histogram into a temporary left histogram and a temporary right histogram;
2) respectively calculating corresponding weighted average probability densities of the temporary left histogram and the temporary right histogram, and performing weighted summation on the two obtained weighted average probability densities according to the set weight to obtain a total weighted average probability density sum; the average probability densityThe calculation method is as follows: defining a pre-emphasis index r, wherein r is more than or equal to 0 and less than or equal to 1, calculating the r power of each numerical value in the temporary left histogram, summing, and dividing by SrsObtaining the weighted average probability density of the temporary left histogram; calculating the r power of each value in the temporary right histogram, then summing and dividing by L-SrsObtaining the weighted average probability density of the temporary right histogram;
3) storing the calculated sum of the weighted average probability densities into the S-th arrayrsIn each element, the length of the array is L, let Srs＝Srs+1, if Srs＜Lmax，LmaxReturning to step 1) for maximum gray level, if S isrs≥LmaxStep 4) is entered;
4) s corresponding to the minimum value of the array in the step 3) by the total weighted average probability densityrsAnd dividing the initial histogram as an optimal division point, calculating a lookup table to obtain an output lookup table, and performing lookup operation on each numerical value of the initial brightness image to obtain a final result image.
2. The optimized model-based dual histogram equalization method of claim 1, wherein said digital image is a multi-channel image, and its luminance channel is used as said luminance image, and if it is a single-channel image, it is directly used as said luminance image.
3. An optimized model based dual histogram equalization method as claimed in claim 1 characterized in that said data partitioning points SdsBy a maximum between-class variance algorithm or a one-dimensional maximum entropy algorithm or using the mean, median or inflection point of the initial histogram of the luminance image.
4. An optimized model based dual histogram equalization method as claimed in claim 1, wherein in step 1), said left accumulation combined array CL(k) And right accumulation array CR(k) Calculating to obtain an integer lookup table, specifically: firstly, according to left accumulation array CL(j) And right accumulation array CR(k) Respectively calculating the left lookup table LutL(j) And right lookup table LutR(k) Is of the formula
LutL(j)＝(Srs-Lmin)*CL(j)+Lmin
LutR(k)＝(Lmax-Srs)*CR(k)+Srs
Wherein L ismin≤j≤Sds，Sds<k≤Lmax(ii) a Then merging the left lookup table LutL(j) And right lookup table LutR(k) And obtaining the integer lookup table.
5. An optimized model based dual histogram equalization method as claimed in claim 1, characterized in that in step 2), the weights are taken as the ratio of the dynamic ranges of the left histogram and the right histogram in step 1).
6. The dual histogram equalization method based on optimized model as claimed in claim 1, wherein in the step 4), the calculation lookup table is specifically: according to left accumulation array CL(j) And right accumulation array CR(k) Respectively calculating the left lookup table LutL(j) And right lookup table LutR(k) Is of the formula
LutL(j)＝(Srs-Lmin)*CL(j)+Lmin
LutR(k)＝(Lmax-Srs)*CR(k)+Srs
Wherein L ismin≤j≤Sds，Sds<k≤Lmax(ii) a Then merge the left lookup table LutL(j) And right lookup table LutR(k) And obtaining the output lookup table.
7. The optimized model-based dual histogram equalization method as claimed in claim 1, wherein in said step 4), if said digital image of initial input is a color image, the final result is restored to the color image.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810044231.0A CN108305234B (en) | 2018-01-17 | 2018-01-17 | Double-histogram equalization method based on optimization model |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810044231.0A CN108305234B (en) | 2018-01-17 | 2018-01-17 | Double-histogram equalization method based on optimization model |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN108305234A CN108305234A (en) | 2018-07-20 |
| CN108305234B true CN108305234B (en) | 2021-08-03 |
Family
ID=62865658
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810044231.0A Active CN108305234B (en) | 2018-01-17 | 2018-01-17 | Double-histogram equalization method based on optimization model |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108305234B (en) |
Families Citing this family (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110956933B (en) * | 2018-09-27 | 2021-02-12 | 中科创达软件股份有限公司 | Power consumption adjusting method and device for display equipment |
| CN114264661A (en) * | 2021-12-06 | 2022-04-01 | 浙江大学台州研究院 | Definition self-adaptive coiled material detection method, device and system |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103188433A (en) * | 2011-12-30 | 2013-07-03 | 株式会社日立制作所 | Image demisting device and image demisting method |
| CN103530847A (en) * | 2013-09-24 | 2014-01-22 | 电子科技大学 | Infrared image enhancing method |
| CN103606137A (en) * | 2013-11-13 | 2014-02-26 | 天津大学 | Histogram equalization method for maintaining background and detail information |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US9208556B2 (en) * | 2010-11-26 | 2015-12-08 | Quantitative Insights, Inc. | Method, system, software and medium for advanced intelligent image analysis and display of medical images and information |
-
2018
- 2018-01-17 CN CN201810044231.0A patent/CN108305234B/en active Active
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103188433A (en) * | 2011-12-30 | 2013-07-03 | 株式会社日立制作所 | Image demisting device and image demisting method |
| CN103530847A (en) * | 2013-09-24 | 2014-01-22 | 电子科技大学 | Infrared image enhancing method |
| CN103606137A (en) * | 2013-11-13 | 2014-02-26 | 天津大学 | Histogram equalization method for maintaining background and detail information |
Non-Patent Citations (2)
| Title |
|---|
| Intensity exposure-based bi-histogram equalization for image enhancement;Jing Rui TANG 等;《Turkish Journal of Electrical Engineering &uter Sciences》;20160131;第26卷(第5期);第3564-3585页 * |
| 基于双直方图均衡的自适应图像增强算法;武英;《计算机工程》;20110228;第37卷(第4期);第250-251页 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN108305234A (en) | 2018-07-20 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110232661B (en) | Low-illumination color image enhancement method based on Retinex and convolutional neural network | |
| CN109410126B (en) | Tone mapping method of high dynamic range image with detail enhancement and brightness self-adaption | |
| Celik | Spatial mutual information and PageRank-based contrast enhancement and quality-aware relative contrast measure | |
| KR100694153B1 (en) | Method and Apparatus for Image Processing based on a Mapping Function | |
| US7889280B2 (en) | Image processing apparatus and method thereof | |
| CN105335933A (en) | Image contrast enhancement method and apparatus | |
| CN108305234B (en) | Double-histogram equalization method based on optimization model | |
| TW202001700A (en) | Method for quantizing an image, a method for training a neural network and a neural network training system | |
| Huang et al. | An advanced gradient histogram and its application for contrast and gradient enhancement | |
| KR101329133B1 (en) | Texture reproduction system and method and chroma correction system for method for immersion enhancement | |
| KR101516632B1 (en) | Bipartite histogram equalization apparatus maintain the rate of mean brightness of video using visual threshold | |
| US7978927B2 (en) | Image processing apparatus | |
| Alex et al. | Underwater image enhancement using CLAHE in a reconfigurable platform | |
| CN109003241B (en) | Linear image enhancement method with double control function | |
| CN109801246B (en) | Global histogram equalization method for adaptive threshold | |
| WO2023284528A1 (en) | Image enhancement method and apparatus, computer device, and storage medium | |
| CN115511737A (en) | Infrared image enhancement method based on semantic segmentation | |
| Mosny et al. | Cubical gamut mapping colour constancy | |
| JP2017530578A (en) | How to choose a compression algorithm depending on the type of image | |
| Chang et al. | Perceptual contrast enhancement of dark images based on textural coefficients | |
| KR20090024375A (en) | System and method for enhancing visual feel of texture according to texture property | |
| Tiwari et al. | Analysis of color contrast enhancement techniques | |
| Singh et al. | Performance Evaluation of Various Histogram Equalization Techniques on Thermal Images | |
| Li et al. | Enhancement of the quality of images based on multiple threshold segmentation and adaptive Gamma correction | |
| Amoah et al. | A bi-histogram shifting contrast enhancement for color images |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |