CN110766616B - Underwater image dodging algorithm based on single-scale Retinex method - Google Patents
Underwater image dodging algorithm based on single-scale Retinex method Download PDFInfo
- Publication number
- CN110766616B CN110766616B CN201910863568.9A CN201910863568A CN110766616B CN 110766616 B CN110766616 B CN 110766616B CN 201910863568 A CN201910863568 A CN 201910863568A CN 110766616 B CN110766616 B CN 110766616B
- Authority
- CN
- China
- Prior art keywords
- image
- eme
- channel
- value
- channels
- 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
- 238000000034 method Methods 0.000 title claims abstract description 80
- 238000004422 calculation algorithm Methods 0.000 title abstract description 34
- 230000003044 adaptive effect Effects 0.000 claims abstract description 3
- 238000004364 calculation method Methods 0.000 claims description 8
- 230000008569 process Effects 0.000 claims description 8
- 239000011159 matrix material Substances 0.000 claims description 6
- 238000001914 filtration Methods 0.000 claims description 5
- 230000004075 alteration Effects 0.000 claims description 3
- 230000000694 effects Effects 0.000 abstract description 10
- 238000010276 construction Methods 0.000 abstract 1
- 238000013139 quantization Methods 0.000 abstract 1
- 238000012937 correction Methods 0.000 description 17
- 238000012545 processing Methods 0.000 description 12
- 238000002474 experimental method Methods 0.000 description 6
- 230000006870 function Effects 0.000 description 5
- 238000010586 diagram Methods 0.000 description 4
- 230000007547 defect Effects 0.000 description 2
- 230000006872 improvement Effects 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 238000011084 recovery Methods 0.000 description 2
- 241001481833 Coryphaena hippurus Species 0.000 description 1
- XEEYBQQBJWHFJM-UHFFFAOYSA-N Iron Chemical group [Fe] XEEYBQQBJWHFJM-UHFFFAOYSA-N 0.000 description 1
- 238000004458 analytical method Methods 0.000 description 1
- 238000000889 atomisation Methods 0.000 description 1
- 230000004456 color vision Effects 0.000 description 1
- 239000002131 composite material Substances 0.000 description 1
- 238000007796 conventional method Methods 0.000 description 1
- 239000006185 dispersion Substances 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 230000004438 eyesight Effects 0.000 description 1
- 230000001795 light effect Effects 0.000 description 1
- 230000014759 maintenance of location Effects 0.000 description 1
- 239000000463 material Substances 0.000 description 1
- 238000010606 normalization Methods 0.000 description 1
- 230000008447 perception Effects 0.000 description 1
- 210000001525 retina Anatomy 0.000 description 1
- 230000011218 segmentation Effects 0.000 description 1
- 238000004088 simulation Methods 0.000 description 1
- 239000000126 substance Substances 0.000 description 1
- 230000000007 visual effect Effects 0.000 description 1
Images
Classifications
-
- G06T5/77—
-
- 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
- G06T2207/10004—Still image; Photographic image
-
- 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
-
- 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
- G06T2207/20—Special algorithmic details
- G06T2207/20024—Filtering details
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02A—TECHNOLOGIES FOR ADAPTATION TO CLIMATE CHANGE
- Y02A90/00—Technologies having an indirect contribution to adaptation to climate change
- Y02A90/30—Assessment of water resources
Abstract
The invention discloses an underwater image dodging algorithm based on a single-scale Retinex method, which calculates EME values of all channels of an image through different step sizes; taking the value as an adaptive cut-off frequency of a construction Gaussian filter; acquiring the ratio of the sum of pixel value intensities of all channels of the underwater image RGB to the sum of pixel intensities of three channels; the ratio is used as the weight of three channels during quantization to correct the color shift problem introduced by the native Retinex method. The invention has the advantages of better light homogenizing effect, avoiding manual parameter adjustment and correcting color cast phenomenon.
Description
Technical Field
The invention belongs to the technical field of underwater image processing, and relates to an underwater image dodging algorithm based on a single-scale Retinex method, which is improved by an EME method.
Background
Underwater image processing, which is a branch in digital image processing, has been a relatively complex one of image processing techniques. Because the underwater image has the problems of large light contrast, serious color distortion, large image noise, serious atomization effect and the like, the processing of the underwater image is obviously more complicated than that of other image enhancement technologies. The method mainly aims at solving the problem of overlarge light ratio of the underwater image, and is called a dodging algorithm. Common underwater dodging algorithms include Mask dodging algorithm, retinex dodging algorithm, wallis dodging algorithm and homomorphism filtering method. The Retinex dodging algorithm is widely applied. In 1963, e.land proposed a calculation theory of color constancy perception, i.e. Retinex theory, based on a model of luminance and color perception of human vision. Retinex is a composite word that is composed of two words, retina and cotex. Over 40 years, the Retinex algorithm was developed by J.J.McCann and D.J.Jobson, zia-Ur Rahman, G.A.Woodell et al, working in IS & T, NASA, mimicking the human visual system, and was modified from a single-scale Retinex algorithm (single scale Retinex, SSR) to a multi-scale Retinex algorithm (MSR) with multi-scale weight average, and developed again into a multi-scale Retinex algorithm (multi-scale Retinex with color restoration, MSRCR) with color recovery.
The basic method of the Retinex dodging algorithm is as follows: first, the image to be homogenized is regarded as being the product between the incident light portion and the reflected light portion, where the reflected light portion corresponds to the original appearance of the image and the incident light portion corresponds to the noise portion in the image. Then, the incident light part is calculated in a specific mode, the incident light part is removed from the image to be homogenized, the reflection component is obtained, the original appearance of the image is restored, and the image homogenizing processing is completed. Jobson, zia-Ur Rahman and g.a. woodell et al propose single-scale Retinex algorithm (SSR) and multi-scale Retinex (MSR) algorithms that use the concept of center-wrapping to estimate the incident portion of the image, which later improves the MSR (MultiScales Retinex with Color Restoration) method of color recovery. The single-scale Retinex (SSR) can only process gray level images, has limitation on the application range of uniform light, and cannot adaptively adjust parameters, so that the uniform light effect is poor. The multi-scale Retinex (MSR) can only process the gray level image, but basic self-adaption is realized, and the light homogenizing effect is stronger than that of manually setting filtering parameters. The Retinex (MSR) with color channels (RGB images) can process gray-scale and color maps, resulting in a great improvement in algorithm adaptability.
Disclosure of Invention
The invention aims to provide an underwater image dodging algorithm based on a single-scale Retinex method, which has the advantages of better dodging effect, avoiding manual parameter adjustment and correcting color cast phenomenon.
The technical scheme adopted by the invention comprises the following steps:
step one: acquiring an input image, wherein the input image I (w, h, 3) is a three-dimensional matrix with the size of w multiplied by h multiplied by 3, w represents width, h represents height, 3 represents the number of channels of the image, and is usually three channels of RGB (red, green and blue) of the image, and each channel can extract a two-dimensional matrix with the size of w multiplied by h to represent spatial information contained in each channel in a source image;
step two: taking any channel in the source image, taking R channel as an example, and recording as I R (x, y) for calculation;
step three: three step values are respectively s 1 ，s 2 ，s 3 The step value is used to calculate EME values at different scales. The EME value is collectively referred to as an image enhancement measure (Measure of Image Enhancement), describing the dynamic range of the image: the larger the EME value, the larger the average dynamic range of the image. Image I R (x, y) is as follows 1 ×s 1 Is divided into k 1 ×k 2 Block then calculate EME 1 The method of the value is as follows:
i in formula (1) max|m，n And I max|m，n Respectively refer to the maximum and minimum values of pixel intensities within the (m, n) th region;
step four: respectively divide the image I R (x, y) is as per s 2 ×s 2 Sum s 3 ×s 3 Is divided into sizes, and the corresponding EME is calculated 2 And EME 3 ；
Step five: r channel image I R Final EME value of (x, y) is
EME R ＝avg(EME 1 +EME 2 +EME 3 ) (2)
Step six: by EME R Value as cut-off frequency to construct a gaussian filter F R R channel image I R (x, y) and the Gaussian filter F R Performing convolution operation to obtain an incident light image B R (x，y)；
Step seven: obtaining R channel image R by difference in logarithmic domain I A reflected image R (x, y) of (x, y),
R(x，y)＝log(R I (x，y))-log(B R (x，y)) (3)
step eight: according to the method from the third step to the seventh step, respectively calculating the reflected images G (x, y) and B (x, y) of the G channel and the B channel;
step nine: weights for the pixel intensities of the three channels of the original image are calculated respectively:
the weight reflects the approximate proportion of three color channels in the initial image, and determines the dominant hue of the original image, so that the image chromatic aberration generated by the original Retinex method can be restrained by the weight;
step ten: will w R R(x，y)，w G G(x，y)，w B B (x, y) are respectively taken as three channels of the final image, and are splicedAnd normalize the pixel intensities to [0,255 ]]And obtaining a final image output.
Further, step three: three step values are respectively s 1 ，s 2 ，s 3 Wherein there is a constraint relationship S 3 ≥2S 2 And S is 2 ≥2S 1 。
Further, in the third to fifth steps, the first three steps s are used 1 ，s 2 ，s 3 And cutting the image, then calculating EME value, and calculating final self-adaptive parameters by adopting an arithmetic mean value method.
Further, in step six, the calculated EME value is used as a sigma value to generate a gaussian filter, and the sigma value of the gaussian filter, that is, the standard deviation, describes the degree of dispersion of the whole filter on the two-dimensional plane: the larger the sigma is, the more discrete the filter is, and the better the overall control of the image is; the smaller σ, the more concentrated the filter, the better the control over the details of the image, when the EME value is used as the adaptive parameter σ of the filter: when the EME value is larger, the average dynamic range of the image is larger, and the contrast of the whole image needs to be revised as a whole, namely a larger sigma value is needed; the smaller the EME value, the smaller the average dynamic range of the image, the more detail information of the image can be preserved by reducing the sigma value when filtering.
Further, in step nine, a channel weight w is calculated by calculating the ratio of the sum of the pixel intensities in each channel in the original image to the sum R ，w G ，w B The color shift phenomenon caused by information loss caused by logarithmic operation in the process of light homogenizing is corrected by describing the contrast of the integral intensity of each channel in an original image, in the process of the light homogenizing algorithm, the pixel values of three finally calculated channels are approximately at the same level caused by logarithmic operation, and the pixel intensity of the B channel in an underwater image is usually inconsistent, so that color difference can be caused, the introduced weight can describe the integral intensity contrast of each channel in the original image to a certain extent, and the color shift phenomenon can be restrained by introducing the introduced weight into a final spliced image.
Further, in step nine, the channel weight w is given to the pixel intensity of the output channel as a whole R ，w G ，w B Thus resulting in a lower overall brightness of the image, but here the brightness range of the image cannot be adjusted to 0,255 by simple stretching]Because the image after being homogenized does not necessarily have its pixel intensity distribution in each channel covering the whole [0,255 ]]Selecting a method of normalizing with weights to optimize the weights, i.e. such that:
w′ K ＝w K /max(w R ，w G ，w B ) (8)
in the formula (8), K is R, G and B respectively, and the problem of excessively low overall brightness of the image can be avoided through the optimized weight.
Drawings
FIG. 1 is a block diagram of an algorithm of the present invention;
FIG. 2 is a block diagram of the steps of the algorithm of the present invention;
FIG. 3 is a schematic illustration of cutting an image according to the EME method;
FIG. 4 is a diagram of the comparison of the results of the algorithm of the present invention with the results of a conventional algorithm (no color correction) shown in FIG. 1;
FIG. 5 is a diagram of the comparison of the results of the algorithm of the present invention with the conventional algorithm (with color correction) of FIG. 1;
FIG. 6 is a graph of the comparison of the results of the algorithm of the present invention with the results of a conventional algorithm (no color correction) 2;
FIG. 7 is a graph of the comparison of the results of the algorithm of the present invention with the conventional algorithm (with color correction) in FIG. 2.
Detailed Description
The present invention will be described in detail with reference to the following embodiments.
As shown in fig. 1, the algorithm of the present invention can be divided into two large modules, namely a color correction module (right) and a Retinex dodging module (left). The system comprises an EME value calculation module, a Retinex dodging module, a color correction module and a combination processing module according to specific functions.
The main function of the EME value calculation module is to give EME values of three channels of the image respectively according to different step sizes. The Retinex dodging module has the main function of constructing a filter by using the EME value calculated by the EME value calculation module, filtering a source image and obtaining a dodging image through calculation. The main function of the color correction module is to obtain the pixel intensity of each channel by calculating the pixel intensity ratio of each channel of the image, thereby correcting the Retinex dodging result. The main functions of the combining processing module are to combine the results made by the color correction module (right) and the Retinex dodging module (left), including channel weighting, channel combining and the like.
And finally, obtaining the dodging image after color correction through the processing.
In order to achieve the above results, detailed embodiments of the present invention will be further described. The specific steps of an embodiment are also shown in fig. 2.
Step one: the input image I (w, h, 3) is obtained as a three-dimensional matrix of size w×h×3, w representing the width and h representing the height, and 3 representing the number of channels of the image, typically RGB three channels of the image. Each channel can extract a two-dimensional matrix with the size of w×h, which represents the spatial information contained in each channel in the source image.
Step two: taking any channel in the source image, taking R channel as an example, and recording as I R (x, y) for calculation;
step three: three step values are respectively s 1 ，s 2 ，s 3 (set to 8, 16, 32, respectively, in this example). The step value is used for calculating EME values under different scales and can be manually adjusted according to actual effects; empirically, 8, 16, 32 are optimal values. Image I R (x, y) is as follows 1 ×s 1 Is divided into k 1 ×k 2 The block, the specific segmentation method is shown in figure 3. The method of calculating the EME value is:
i in formula (1) max|m，n And I min|m，n Respectively refer to the maximum and minimum values of pixel intensities within the (m, n) th region. The constant on the denominator is to avoid the situation where the denominator is zero. It should be noted that if the remaining part of the image is insufficient to make up s 1 ×s 1 The size of the defect s is not needed to be supplemented, and only the defect s is needed to be taken 1 ×s 1 The maximum and minimum pixel intensities within the size region are sufficient.
Step four: according to the method in the third step, respectively converting the images I R (x, y) is as per s 2 ×s 2 Sum s 3 ×s 3 Is divided into sizes, and the corresponding EME is calculated 2 And EME 3 。
Step five: r channel image I R Final EME value of (x, y) is
EME R ＝avg(EME 1 +EME 2 +EME 3 ) (2)
Step six: by EME R The value as cut-off frequency (sigma value) constructs a gaussian filter F R R channel image I R (x, y) and the Gaussian filter F R Performing convolution operation to obtain an incident light image B R (x，y)。
Step seven: obtaining R channel image R by difference in logarithmic domain I Reflected image R (x, y) of (x, y):
R(x，y)＝log(R I (x，y))-log(B R (x，y)) (3)
step eight: according to the methods of the third to seventh steps, reflected images G (x, y) and B (x, y) of the G and B channels are calculated, respectively.
Step nine: the proportion (weight) of the pixel intensities of the three channels of the original image is calculated respectively:
the weights reflect the approximate proportions of the three color channels within the original image, determining the dominant hue of the original image. Thus, the image chromatic aberration generated by the native Retinex method can be suppressed by the weight.
Step ten: will w R R(x，y)，w G G(x，y)，w B B (x, y) are taken as three channels of the final image respectively, they are spliced, and the pixel intensity is normalized to [0,255]And obtaining a final image output.
Since the channel weight w is given to the pixel intensity of the output channel as a whole R ，w G ，w B Thus resulting in a lower overall brightness of the image. But here the brightness range of the image cannot be adjusted to 0,255 by simple stretching]Because the image after being homogenized does not necessarily have its pixel intensity distribution in each channel covering the whole [0,255 ]]Is defined as the pixel intensity axis of (a). The method of weight normalization is chosen here to optimize the weights, i.e. such that:
w′ K ＝w K /max(w R ，w G ，w B ) (8)
in the formula (8), K is R, G and B respectively. The optimized weight can avoid the problem of excessively low overall brightness of the image.
Algorithm performance analysis
A small number of images will be selected to verify the algorithm and to illustrate the effectiveness of the algorithm.
1. Experimental conditions
In the experiment, an underwater image and a part of internet image acquired by a homemade underwater three-eye camera are selected as experimental data, MATLAB R2018a under Windows10 is adopted as a simulation tool, and a computer is configured as Intel Core i7-8550U@1.80GHz.
2. Experimental details
Firstly, selecting part of self-made camera to shoot images, processing the images by using a traditional Retinex method and an EME-Retinex method respectively under the condition of not using color correction, and comparing experimental results. Then, in the case of using color correction, it is processed using the conventional Retinex method and the EME-Retinex method, respectively, and experimental results are compared.
And then selecting part of the Internet image as experimental data, processing the Internet image by using a traditional Retinex method and an EME-Retinex method respectively under the condition of not using color correction, and comparing experimental results. Then, in the case of using color correction, it is processed using the conventional Retinex method and the EME-Retinex method, respectively, and experimental results are compared.
3. Experimental results
Experiment 1 was performed on an image taken from a camera, and processed by the conventional Retinex method and the EME-Retinex method, respectively, without using color correction. As shown in fig. 4, the source image, the image processed by the conventional Retinex method, and the image processed by the EME-Retinex method are sequentially shown from left to right. It is apparent that the images 4 (b) and (c) show significant color distortion. In terms of the dodging results, fig. 4 (c) is significantly better than fig. 4 (b), especially in the bright and dark parts of the iron ring, with a great improvement in detail retention results.
Then, color correction is added to the method. As shown in fig. 5, the source image, the image processed by the conventional Retinex method, and the image processed by the EME-Retinex method are sequentially shown from left to right. In the experimental result, the color distortion phenomenon in fig. 4 is obviously inhibited, and the color of the image is more similar to that of the original image. Fig. 5 (c) is still better than fig. 5 (b) in terms of the dodging effect.
Experiment 2 takes part of the internet pictures as data, and the experimental method remains the same as experiment 1. Fig. 6 and 7 are the results of processing a source image using the conventional Retinex method and the EME-Retinex method, respectively, with and without color correction applied, and are, from left to right, the source image, the image processed by the conventional Retinex method, and the image processed by the EME-Retinex method, respectively. In experiment 2, the light homogenizing effect of the conventional method and the EME method is almost the same, and when the light homogenizing effect is carefully observed, the EME method is more excellent in treatment at the halation around the dolphin. Considering that the core of the Retinex theory is to remove the effect of the incident light, it is apparent that the set of color corrected fig. 7 is more closely shielded from the effect of the incident light.
The invention has the advantages that:
1. the invention adopts EME value as the self-adaptive parameter of the Gaussian filter, avoids uncertainty introduced by artificial subjective parameter adjustment, and saves manpower and material resources.
2. In the process of calculating EME, the invention adopts the mode of dividing images with different sizes and adopts the mode of weighted average to reduce the error generated in the process of calculating EME.
3. The invention provides a color correction method, which is used for simply correcting the color shift problem caused by the Retinex method and expanding the usable range and the universality of the method.
4. The method provided by the invention has good adaptability to the underwater image, can better perform dodging, and better reserves the details of the underwater bright part and the underwater dark part.
The foregoing description is only of the preferred embodiments of the present invention, and is not intended to limit the invention in any way, and any simple modification, equivalent variation and modification made to the above embodiments according to the technical substance of the present invention falls within the scope of the technical solution of the present invention.
Claims (6)
1. An underwater image dodging method based on a single-scale Retinex method is characterized by comprising the following steps of:
step one: acquiring an input image, wherein the input image I (w, h, 3) is a three-dimensional matrix with the size of w multiplied by h multiplied by 3, w represents width, h represents height, 3 represents the number of channels of the image, and is usually three channels of RGB (red, green and blue) of the image, and each channel can extract a two-dimensional matrix with the size of w multiplied by h to represent spatial information contained in each channel in a source image;
step two: taking any channel in the source image, taking R channel as an example, and recording as I R (x, y) for calculation;
step three: three step values are respectively s 1 ，s 2 ，s 3 The step value is used for calculating EME values under different scales, and the image I is obtained R (x, y) is as follows 1 ×s 1 Is divided into k 1 ×k 2 The method for calculating EME value is as follows:
i in formula (1) max|m，n And I min|m，n Respectively refer to the maximum and minimum values of pixel intensities within the (m, n) th region;
step four: respectively divide the image I R (x, y) is as per s 2 ×s 2 Sum s 3 ×s 3 Is divided into sizes, and the corresponding EME is calculated 2 And EME 3 ；
Step five: r channel image I R Final EME value of (x, y) is
EME R ＝avg(EME 1 +EME 2 +EME 3 ) (2)
Step six: by EME R Value as cut-off frequency to construct a gaussian filter F R R channel image I R (x, y) and the Gaussian filter F R Performing convolution operation to obtain an incident light image B R (x，y)；
Step seven: obtaining R channel image R by difference in logarithmic domain I A reflected image R (x, y) of (x, y),
R(x，y)＝log(R I (x，y))-log(B R (x，y)) (3)
step eight: according to the method from the third step to the seventh step, respectively calculating the reflected images G (x, y) and B (x, y) of the G channel and the B channel;
step nine: weights for the pixel intensities of the three channels of the original image are calculated respectively:
the weight reflects the approximate proportion of three color channels in the initial image, and determines the dominant hue of the original image, so that the image chromatic aberration generated by the original Retinex method can be restrained by the weight;
step ten: will w R R(x，y)，w G G(x，y)，w B B (x, y) are taken as three channels of the final image respectively, they are spliced, and the pixel intensities are normalized to 0,255]And obtaining a final image output.
2. The underwater image dodging method based on the single-scale Retinex method according to claim 1, wherein the method comprises the following steps: and step three: three step values are respectively s 1 ，s 2 ，s 3 Wherein there is a constraint relationship S 3 ≥2S 2 And S is 2 ≥2S 1 。
3. The underwater image dodging method based on the single-scale Retinex method according to claim 1, wherein the method comprises the following steps: in the third to fifth steps, the first three steps s are used 1 ，s 2 ，s 3 And cutting the image, then calculating EME value, and calculating final self-adaptive parameters by adopting an arithmetic mean value method.
4. The underwater image dodging method based on the single-scale Retinex method according to claim 1, wherein the method comprises the following steps: in the sixth step, the calculated EME value is used as a sigma value to generate a gaussian filter, and the sigma value of the gaussian filter, namely, the standard deviation, describes the discrete degree of the whole filter on a two-dimensional plane: the larger the sigma is, the more discrete the filter is, and the better the overall control of the image is; the smaller σ, the more concentrated the filter, the better the control over the details of the image, when the EME value is used as the adaptive parameter σ of the filter: when the EME value is larger, the average dynamic range of the image is larger, and the contrast of the whole image needs to be revised as a whole, namely a larger sigma value is needed; the smaller the EME value, the smaller the average dynamic range of the image, the more detail information of the image can be preserved by reducing the sigma value when filtering.
5. The underwater image dodging method based on the single-scale Retinex method according to claim 1, wherein the method comprises the following steps: in step nine, the channel weight w is calculated by calculating the ratio of the sum of the pixel intensities in each channel in the original image to the sum R ，w G ，w B The method is used for describing the comparison of the overall intensity of each channel in the original image, so as to correct the color cast phenomenon caused by information loss due to logarithmic operation in the process of dodging.
6. The underwater image dodging method based on the single-scale Retinex method according to claim 1, wherein the method comprises the following steps: in the step nine, a method of normalizing the weights is selected to optimize the weights, namely, to enable:
w′ K ＝w K /max(w R ，w G ，w B ) (8)
in the formula (8), K is R, G, B three image channels respectively, and the problem of excessively low overall brightness of the image can be avoided through the optimized weight.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910863568.9A CN110766616B (en) | 2019-09-12 | 2019-09-12 | Underwater image dodging algorithm based on single-scale Retinex method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910863568.9A CN110766616B (en) | 2019-09-12 | 2019-09-12 | Underwater image dodging algorithm based on single-scale Retinex method |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN110766616A CN110766616A (en) | 2020-02-07 |
| CN110766616B true CN110766616B (en) | 2023-05-09 |
Family
ID=69329480
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910863568.9A Active CN110766616B (en) | 2019-09-12 | 2019-09-12 | Underwater image dodging algorithm based on single-scale Retinex method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110766616B (en) |
Families Citing this family (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115482216B (en) * | 2022-09-20 | 2023-05-16 | 河南大学第一附属医院 | Laparoscopic image enhancement method and system |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106934775A (en) * | 2017-03-08 | 2017-07-07 | 中国海洋大学 | A kind of non local image de-noising method recovered based on low-rank |
| CN108053374A (en) * | 2017-12-05 | 2018-05-18 | 天津大学 | A kind of underwater picture Enhancement Method of combination bilateral filtering and Retinex |
| CN108520539A (en) * | 2018-03-13 | 2018-09-11 | 中国海洋大学 | A kind of image object detection method based on sparse study variable model |
| CN110175964A (en) * | 2019-05-30 | 2019-08-27 | 大连海事大学 | A kind of Retinex image enchancing method based on laplacian pyramid |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US7590303B2 (en) * | 2005-09-29 | 2009-09-15 | Samsung Electronics Co., Ltd. | Image enhancement method using local illumination correction |
-
2019
- 2019-09-12 CN CN201910863568.9A patent/CN110766616B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106934775A (en) * | 2017-03-08 | 2017-07-07 | 中国海洋大学 | A kind of non local image de-noising method recovered based on low-rank |
| CN108053374A (en) * | 2017-12-05 | 2018-05-18 | 天津大学 | A kind of underwater picture Enhancement Method of combination bilateral filtering and Retinex |
| CN108520539A (en) * | 2018-03-13 | 2018-09-11 | 中国海洋大学 | A kind of image object detection method based on sparse study variable model |
| CN110175964A (en) * | 2019-05-30 | 2019-08-27 | 大连海事大学 | A kind of Retinex image enchancing method based on laplacian pyramid |
Non-Patent Citations (2)
| Title |
|---|
| shenghui rong.An improved non-uniformity correction algorithm and its GPU parallel implementation,web of science.《web of science》.2018,全文. * |
| 刘贵杰 ; 王猛 ; 何波 ; .基于Adams与Matlab/Simulink的水下自航行器协同仿真.机械工程学报.2009,(第10期),全文. * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN110766616A (en) | 2020-02-07 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110505459B (en) | Image color correction method, device and storage medium suitable for endoscope | |
| 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 | |
| Parthasarathy et al. | An automated multi scale retinex with color restoration for image enhancement | |
| Vonikakis et al. | Fast centre–surround contrast modification | |
| CN107798661B (en) | Self-adaptive image enhancement method | |
| CN111210395B (en) | Retinex underwater image enhancement method based on gray value mapping | |
| CN111415304A (en) | Underwater vision enhancement method and device based on cascade deep network | |
| Fang et al. | A method to improve the image enhancement result based on image fusion | |
| CN108648160B (en) | Underwater sea cucumber image defogging enhancement method and system | |
| CN114399442B (en) | Nonlinear image enhancement method and system based on parameter self-adaption | |
| CN113947553B (en) | Image brightness enhancement method and device | |
| CN107169942B (en) | Underwater image enhancement method based on fish retina mechanism | |
| CN109886885B (en) | Image enhancement method and system based on Lab color space and Retinex | |
| CN110766616B (en) | Underwater image dodging algorithm based on single-scale Retinex method | |
| WO2020107308A1 (en) | Low-light-level image rapid enhancement method and apparatus based on retinex | |
| CN116993616A (en) | Single low-illumination scene image enhancement method and enhancement system | |
| JP5410378B2 (en) | Video signal correction apparatus and video signal correction program | |
| Ismail et al. | Adapted single scale retinex algorithm for nighttime image enhancement | |
| Mahmood et al. | A color image enhancement technique using multiscale retinex | |
| CN109417616B (en) | Method and apparatus for image processing | |
| CN115049561A (en) | Real image reproduction method based on non-ideal illumination image | |
| CN115018717A (en) | Improved Retinex-Net low-illumination and dark vision image enhancement method | |
| Fan et al. | Underwater image enhancement algorithm combining color correction and multi-scale fusion | |
| CN110675332A (en) | Method for enhancing quality of metal corrosion image |
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 |