CN112950589A - Dark channel prior defogging algorithm of multi-scale convolution neural network - Google Patents
Dark channel prior defogging algorithm of multi-scale convolution neural network Download PDFInfo
- Publication number
- CN112950589A CN112950589A CN202110234825.XA CN202110234825A CN112950589A CN 112950589 A CN112950589 A CN 112950589A CN 202110234825 A CN202110234825 A CN 202110234825A CN 112950589 A CN112950589 A CN 112950589A
- Authority
- CN
- China
- Prior art keywords
- image
- dark channel
- neural network
- fog
- convolutional neural
- 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
- 238000013528 artificial neural network Methods 0.000 title claims abstract description 9
- 238000013527 convolutional neural network Methods 0.000 claims abstract description 25
- 238000002834 transmittance Methods 0.000 claims abstract description 25
- 238000001914 filtration Methods 0.000 claims abstract description 10
- 238000005286 illumination Methods 0.000 claims abstract description 9
- 238000012545 processing Methods 0.000 claims abstract description 7
- 238000000034 method Methods 0.000 claims description 14
- 238000012549 training Methods 0.000 claims description 9
- 238000010586 diagram Methods 0.000 claims description 5
- 238000012163 sequencing technique Methods 0.000 claims description 4
- 239000011159 matrix material Substances 0.000 claims description 3
- 238000007781 pre-processing Methods 0.000 claims description 3
- 238000004364 calculation method Methods 0.000 claims description 2
- 230000005540 biological transmission Effects 0.000 abstract description 2
- 238000007670 refining Methods 0.000 abstract description 2
- 238000011176 pooling Methods 0.000 description 5
- 238000005516 engineering process Methods 0.000 description 3
- 230000004927 fusion Effects 0.000 description 3
- 238000010521 absorption reaction Methods 0.000 description 2
- 238000000605 extraction Methods 0.000 description 2
- 239000002245 particle Substances 0.000 description 2
- 230000010287 polarization Effects 0.000 description 2
- 238000005070 sampling Methods 0.000 description 2
- 230000000007 visual effect Effects 0.000 description 2
- 238000013459 approach Methods 0.000 description 1
- 230000002238 attenuated effect Effects 0.000 description 1
- 230000015556 catabolic process Effects 0.000 description 1
- 238000006731 degradation reaction Methods 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 239000000284 extract Substances 0.000 description 1
- 238000003384 imaging method Methods 0.000 description 1
- 238000013507 mapping Methods 0.000 description 1
- 239000003595 mist Substances 0.000 description 1
- 238000012634 optical imaging Methods 0.000 description 1
- 238000011084 recovery Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/0002—Inspection of images, e.g. flaw detection
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/213—Feature extraction, e.g. by transforming the feature space; Summarisation; Mappings, e.g. subspace methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- 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
- G06T2207/10004—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
- G06T2207/20—Special algorithmic details
- G06T2207/20081—Training; Learning
-
- 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
- G06T2207/20084—Artificial neural networks [ANN]
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- General Engineering & Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- Evolutionary Computation (AREA)
- Biophysics (AREA)
- Biomedical Technology (AREA)
- Computing Systems (AREA)
- General Health & Medical Sciences (AREA)
- Computational Linguistics (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Molecular Biology (AREA)
- Health & Medical Sciences (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Evolutionary Biology (AREA)
- Quality & Reliability (AREA)
- Image Processing (AREA)
Abstract
The invention discloses a dark channel prior defogging algorithm of a multi-scale convolutional neural network, which is characterized in that the original fog image is subjected to minimum filtering processing by using a dark channel prior theory, and the illumination intensity mean value of the pixel point position with the brightness of 0.1 percent in the dark color image is taken as a global atmospheric light value; and then estimating the fine transmittance by using a multi-scale convolution neural network, wherein the proposed algorithm consists of a coarse-scale network for performing global transmission map estimation on the whole image and a fine-scale network for locally refining the transmittance, and finally, the global atmospheric light value and the fine transmittance are substituted into an atmospheric scattering model to recover a fog-free image.
Description
Technical Field
The invention relates to the technical field of image processing, in particular to a dark channel prior defogging algorithm of a multi-scale convolution neural network.
Background
Under the severe weather conditions of fog, thunder and the like, the suspended particles in the air can influence the outdoor shooting condition of the optical imaging equipment, and the absorption, scattering and refraction of the turbid medium on the reflected light can cause the attenuation of incident light of a camera for acquiring an image of an object, so that the imaging quality is turbid, and the extraction of valuable information in the surrounding environment by the camera is seriously influenced.
Based on the defogging method of the traditional image enhancement technology, the problem of image degradation caused by attenuation of incident light rays under the foggy weather condition is not generally considered, but the image enhancement technology is directly adopted to appropriately enhance the contrast, brightness or saturation of the foggy weather image so as to highlight valuable information in the degraded image; the defogging method based on the polarization characteristic generally needs more than two specific cameras to realize, and the core idea is to utilize a plurality of images shot from different polarization angles to perform defogging. The defogging method based on the fusion strategy characteristics extracts the characteristics of the input foggy images, then fuses the foggy images by utilizing the pyramid multi-scale technology, thereby realizing the improvement of the visual effect of the foggy images, and classifies the extracted characteristics into a single-scale fusion strategy (SSR) and a multi-scale fusion strategy (MSR) according to the difference that the extracted characteristics come from a single image or a plurality of images. The methods mentioned above all simply improve the visual subjectivity of the foggy image by some method, and do not really realize physical defogging.
Disclosure of Invention
The invention aims to provide a dark channel prior defogging algorithm of a multi-scale convolutional neural network, and aims to solve the technical problems of poor defogging effect and long defogging time of a defogging method in the prior art.
In order to achieve the above object, the dark channel prior defogging algorithm of the multi-scale convolutional neural network adopted by the invention comprises the following steps:
acquiring an original fog image, preprocessing the original fog image, and calculating a dark channel value of the original fog image;
calculating a dark channel image and recording the parameter of each pixel point of the dark channel image;
comparing the brightness parameters of the pixel points and sequencing to obtain a global atmospheric light value;
selecting a multi-scale convolution neural network, inputting the original fog diagram for training, and obtaining refined transmittance;
and recovering the fog-free image.
And in the process of calculating the dark channel value of the original fog image, performing minimum filtering operation on the original fog image, enlarging the radius of an original matrix, filling the excessive pixel points with 255 values, and then obtaining the minimum value by using double-cycle traversal.
In the process of calculating a dark channel image and recording parameters of each pixel point of the dark channel image, the original fog image is subjected to minimum filtering processing to obtain the dark channel image, and then the brightness and the coordinates of each pixel point of the dark channel image are recorded.
The specific steps of comparing the brightness parameters of the pixel points and sequencing are that the pixel points of the dark channel image are sequenced in a descending order according to the brightness, the illumination value intensity of the pixel point position with the brightness of 0.1% in the front is found, and the coordinates of the pixel values are recorded.
And the global atmospheric light value is the average value of the illumination intensity of the pixel points with the brightness of 0.1 percent in the front.
The multi-scale convolutional neural network consists of a coarse-scale convolutional neural network and a fine-scale convolutional neural network, and the original fog diagram is input into the coarse-scale convolutional neural network for training at first and then fed into the fine-scale convolutional neural network for training together with a training result.
And in the process of recovering the fog-free image, substituting the global atmospheric light value and the refined transmittance into an atmospheric scattering model to recover the fog-free image.
According to the dark channel prior defogging algorithm of the multi-scale convolution neural network, the original fog image is subjected to minimum filtering processing by using a dark channel prior theory, and the illumination intensity mean value of the pixel point position with the brightness of 0.1% in the dark color image is taken as a global atmospheric light value; and then estimating the fine transmittance by using a multi-scale convolution neural network, wherein the proposed algorithm consists of a coarse-scale network for performing global transmission map estimation on the whole image and a fine-scale network for locally refining the transmittance, and finally, the global atmospheric light value and the fine transmittance are substituted into an atmospheric scattering model to recover a fog-free image.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the embodiments or the prior art will be briefly described below, it is obvious that the drawings in the following description are only some embodiments of the present invention, and for those skilled in the art, other drawings can be obtained according to the drawings without creative efforts.
FIG. 1 is a flow chart diagram of the dark channel prior defogging algorithm of the multi-scale convolutional neural network of the present invention.
FIG. 2 is a flow chart of the multi-scale convolutional neural network structure of the present invention.
Detailed Description
Reference will now be made in detail to embodiments of the present invention, examples of which are illustrated in the accompanying drawings, wherein like or similar reference numerals refer to the same or similar elements or elements having the same or similar function throughout. The embodiments described below with reference to the drawings are illustrative and intended to be illustrative of the invention and are not to be construed as limiting the invention.
In the description of the present invention, it is to be understood that the terms "length", "width", "upper", "lower", "front", "rear", "left", "right", "vertical", "horizontal", "top", "bottom", "inner", "outer", and the like, indicate orientations or positional relationships based on the orientations or positional relationships illustrated in the drawings, and are used merely for convenience in describing the present invention and for simplicity in description, and do not indicate or imply that the devices or elements referred to must have a particular orientation, be constructed in a particular orientation, and be operated, and thus, are not to be construed as limiting the present invention. Further, in the description of the present invention, "a plurality" means two or more unless specifically defined otherwise.
Referring to fig. 1, the present invention provides a dark channel prior defogging algorithm for a multi-scale convolutional neural network, which comprises the following steps:
s1: acquiring an original fog image, preprocessing the original fog image, and calculating a dark channel value of the original fog image;
s2: calculating a dark channel image and recording the parameter of each pixel point of the dark channel image;
s3: comparing the brightness parameters of the pixel points and sequencing to obtain a global atmospheric light value;
s4: selecting a multi-scale convolution neural network, inputting the original fog diagram for training, and obtaining refined transmittance;
s5: and recovering the fog-free image.
Optionally, in the process of calculating the dark channel value of the original fog map, performing minimum filtering operation on the original fog map, enlarging the radius of the original matrix, filling the excess pixel points with 255 values, and then obtaining the minimum value by using double-loop traversal.
Optionally, in the process of calculating a dark channel image and recording the parameter of each pixel point of the dark channel image, the original fog image is subjected to minimum filtering to obtain the dark channel image, and then the brightness and the coordinate of each pixel point of the dark channel image are recorded.
Further optionally, the specific step of comparing the brightness parameters of the pixel points and sorting includes sorting the pixel points of the dark channel image in a descending order according to the brightness, finding the illumination value intensity of the pixel point position with the brightness being 0.1% of the first brightness, and recording the coordinates of the pixel values.
Optionally, the global atmospheric light value is an illumination intensity average value of a pixel point with the brightness of 0.1% in the front.
According to the atmospheric scattering model, the image captured by the high-definition camera in the foggy environment is shielded by haze, and incident light for acquiring the scene image is attenuated due to absorption, scattering and refraction of reflected light by suspended particles in the air. Therefore, the atmospheric scattering model of the degraded image in the foggy weather condition is represented by the following formula (1):
I(x)＝J(x)t(x)+A(1-t(x)) (1)
wherein, I (x) is a photographed foggy image; j (x) is a clear image after defogging; a is the global atmospheric light value; t (x) is a medium transmittance, and when the mist concentration is uniform, it can be further represented as t (x) e-βd(x)(β is the atmospheric scattering coefficient, d (x) is the scene depth); j (x) t (x) is a direct attenuation term reflecting the degree of attenuation of the object in the medium; a (1-t (x)) is atmospheric scattered light with a color shift in the scene.
The hong Kong Chinese university Hokeming et al makes statistics of a large number of outdoor images to summarize a basic rule: in most clear outdoor images, most local areas contain object shadows with pixel intensity values approaching 0, and further assuming that there are one or more points in all local areas where the pixel intensity values approach 0 on a certain color channel, the process of obtaining a dark primary image for one image J through a minimum filtering operation is described by the following formula:
wherein JcEach channel representing a color image, c ∈ { R, G, B } representingΩ (x, y) represents a region centered on the pixel (x, y) as a color label of the image.
And for the global atmospheric light value A, performing minimum filtering operation on the input fog image to obtain a dark primary color image, and calculating the illumination intensity average value of the pixel point position corresponding to the front 0.1% of the brightness value in the dark primary color image to be used as the global atmospheric light value A required for recovering the fog-free image.
For the transmittance t (x), we use a multi-scale convolutional neural network to estimate the transmittance t (x), and the overall working flow is shown in the figure I. The coarse-scale convolutional neural network is used for obtaining the coarse transmittance of the atomized image, and the network mainly comprises four parts, namely a convolutional layer, a maximum pooling layer, an upsampling layer and a linear combination layer.
Wherein the convolutional layers are composed of filters convolved with the input feature map, and the response of each convolutional layer is composed of
Is given in
And
is the feature map of the l-th layer and the l + 1-th layer;
the maximum pooling layer is used for performing down-sampling by using a filter with the size of 2 x 2 after each convolution layer because the number of characteristic images after convolution operation is large and the processing time is long if the calculation is directly performed, so that the image dimensionality reduction complexity is reduced;
the upsampling layer, because the original outdoor transmittance dimension should be the same as the input fog map transmittance dimension, and the input image is transmitted to the max-pooling layer by way of convolution kernel feature mapping, the transmittance dimension will be reduced by half, so an upsampling layer needs to be added to ensure that the final output transmittance is the same as the original input transmittance dimension. The upsampling layer, following the pooling layer, has the function of sub-sampling feature size recovery while preserving the non-linearity of the network. Each corresponding upsampling layer is defined as:
the above equation copies the feature pixel value after the maximum pooling at the (x, y) position into a 2 × 2 block of the subsequent upsampling layer;
the linear combination layer, which needs to output the multi-feature channels in the upsampling layer, performs linear combination to obtain the final output result (transmittance obtained by the coarse-scale network), and the linear combination expression is as follows:
referring to fig. 2, since the coarse-scale convolutional neural network adopts a network model with a larger size, shallow feature information of an image may be unconsciously ignored due to excessive attention on feature extraction of scene information, and thus, the transmittance t (x) obtained by the coarse-scale convolutional neural network cannot be directly used for image defogging processing. A fine-scale convolutional neural network is required to be closely connected behind the coarse-scale network model, and the obtained coarse transmittance of the coarse-scale network model is refined to obtain transmittance information closer to a real scene.
And deforming the atmospheric scattering model to obtain the following formula:
and (3) substituting the obtained atmospheric light value and refined transmittance into a formula (5) to remove haze in the image and recover a clear image.
While the invention has been described with reference to a preferred embodiment, it will be understood by those skilled in the art that various changes in form and detail may be made therein without departing from the spirit and scope of the invention as defined by the appended claims.
Claims (7)
1. A dark channel prior defogging algorithm for a multi-scale convolutional neural network is characterized by comprising the following steps:
acquiring an original fog image, preprocessing the original fog image, and calculating a dark channel value of the original fog image;
calculating a dark channel image and recording the parameter of each pixel point of the dark channel image;
comparing the brightness parameters of the pixel points and sequencing to obtain a global atmospheric light value;
selecting a multi-scale convolution neural network, inputting the original fog diagram for training, and obtaining refined transmittance;
and recovering the fog-free image.
2. The dark channel a priori defogging algorithm for a multiscale convolutional neural network as recited in claim 1, wherein in the calculation of the dark channel values of said original fog map,
and performing minimum filtering operation on the original fog image, expanding the radius of the original matrix, filling the excessive pixel points with 255 values, and then obtaining the minimum value by using double-cycle traversal.
3. The dark channel prior defogging algorithm according to claim 2 wherein, in the process of computing a dark channel image and recording the parameters of each pixel point of said dark channel image,
and carrying out minimum filtering processing on the original fog image to obtain the dark channel image, and then recording the brightness and the coordinates of each pixel point of the dark channel image.
4. The dark channel prior defogging algorithm according to claim 3, wherein the step of comparing the brightness parameters of the pixels and sorting the pixels comprises sorting the pixels of said dark channel image in descending order according to the brightness, finding the illumination intensity at the pixel position with the brightness of the first 0.1%, and recording the coordinates of the pixel values.
5. The dark channel a priori defogging algorithm of the multi-scale convolutional neural network according to claim 4, wherein said global atmospheric light value is an average of illumination intensities of pixels with brightness of the first 0.1% in magnitude.
6. The dark channel prior defogging algorithm according to claim 5, wherein said multi-scale convolutional neural network is composed of a coarse-scale convolutional neural network and a fine-scale convolutional neural network, and said raw fog pattern is input into said coarse-scale convolutional neural network for training and then fed into said fine-scale convolutional neural network for training together with the training result.
7. The dark channel a priori defogging algorithm for a multi-scale convolutional neural network according to claim 6, wherein in the process of recovering a fog-free image, the global atmospheric light value and the refined transmittance are substituted into an atmospheric scattering model to recover a fog-free image.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110234825.XA CN112950589A (en) | 2021-03-03 | 2021-03-03 | Dark channel prior defogging algorithm of multi-scale convolution neural network |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110234825.XA CN112950589A (en) | 2021-03-03 | 2021-03-03 | Dark channel prior defogging algorithm of multi-scale convolution neural network |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN112950589A true CN112950589A (en) | 2021-06-11 |
Family
ID=76247259
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110234825.XA Pending CN112950589A (en) | 2021-03-03 | 2021-03-03 | Dark channel prior defogging algorithm of multi-scale convolution neural network |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN112950589A (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113920018A (en) * | 2021-09-13 | 2022-01-11 | 中国人民解放军战略支援部队航天工程大学 | Improved MSCNN underwater image defogging method |
| CN114648467A (en) * | 2022-05-18 | 2022-06-21 | 中山大学深圳研究院 | Image defogging method and device, terminal equipment and computer readable storage medium |
| CN115331264A (en) * | 2022-10-13 | 2022-11-11 | 山东济矿鲁能煤电股份有限公司阳城煤矿 | Mine worker safety identification management method based on neural network |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109712083A (en) * | 2018-12-06 | 2019-05-03 | 南京邮电大学 | A kind of single image to the fog method based on convolutional neural networks |
| CN109978799A (en) * | 2019-04-15 | 2019-07-05 | 武汉理工大学 | A kind of maritime affairs UAV Video image defogging method based on deep learning |
| US20200245969A1 (en) * | 2019-01-31 | 2020-08-06 | Qisda Corporation | Ultrasound Needle Positioning System and Ultrasound Needle Positioning Method Utilizing Convolutional Neural Networks |
| CN111563852A (en) * | 2020-04-24 | 2020-08-21 | 桂林电子科技大学 | Dark channel prior defogging method based on low-complexity MF |
-
2021
- 2021-03-03 CN CN202110234825.XA patent/CN112950589A/en active Pending
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109712083A (en) * | 2018-12-06 | 2019-05-03 | 南京邮电大学 | A kind of single image to the fog method based on convolutional neural networks |
| US20200245969A1 (en) * | 2019-01-31 | 2020-08-06 | Qisda Corporation | Ultrasound Needle Positioning System and Ultrasound Needle Positioning Method Utilizing Convolutional Neural Networks |
| CN109978799A (en) * | 2019-04-15 | 2019-07-05 | 武汉理工大学 | A kind of maritime affairs UAV Video image defogging method based on deep learning |
| CN111563852A (en) * | 2020-04-24 | 2020-08-21 | 桂林电子科技大学 | Dark channel prior defogging method based on low-complexity MF |
Non-Patent Citations (2)
| Title |
|---|
| 寇大磊等: "基于多尺度卷积网络的快速图像去雾算法", 《计算机工程与应用》 * |
| 金仙力等: "基于引导滤波和自适应容差的图像去雾算法", 《通信学报》 * |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113920018A (en) * | 2021-09-13 | 2022-01-11 | 中国人民解放军战略支援部队航天工程大学 | Improved MSCNN underwater image defogging method |
| CN114648467A (en) * | 2022-05-18 | 2022-06-21 | 中山大学深圳研究院 | Image defogging method and device, terminal equipment and computer readable storage medium |
| CN115331264A (en) * | 2022-10-13 | 2022-11-11 | 山东济矿鲁能煤电股份有限公司阳城煤矿 | Mine worker safety identification management method based on neural network |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111915530B (en) | End-to-end-based haze concentration self-adaptive neural network image defogging method | |
| CN112950589A (en) | Dark channel prior defogging algorithm of multi-scale convolution neural network | |
| CN103020920B (en) | Method for enhancing low-illumination images | |
| CN108510451B (en) | Method for reconstructing license plate based on double-layer convolutional neural network | |
| Kuanar et al. | Night time haze and glow removal using deep dilated convolutional network | |
| CN112184604B (en) | Color image enhancement method based on image fusion | |
| CN110097522B (en) | Single outdoor image defogging method based on multi-scale convolution neural network | |
| CN109389569B (en) | Monitoring video real-time defogging method based on improved DehazeNet | |
| CN112614136A (en) | Infrared small target real-time instance segmentation method and device | |
| CN114627269A (en) | Virtual reality security protection monitoring platform based on degree of depth learning target detection | |
| CN110223240A (en) | Image defogging method, system and storage medium based on color decaying priori | |
| CN115775376A (en) | Crowd counting method based on low-light image enhancement | |
| CN115019340A (en) | Night pedestrian detection algorithm based on deep learning | |
| CN112907461B (en) | Defogging enhancement method for infrared foggy-day degraded image | |
| CN115063434A (en) | Low-low-light image instance segmentation method and system based on feature denoising | |
| CN107301625B (en) | Image defogging method based on brightness fusion network | |
| CN110738624A (en) | area self-adaptive image defogging system and method | |
| Guo et al. | Marg-unet: a single image dehazing network based on multimodal attention residual group | |
| CN110070007A (en) | Video smoke recognition methods, device, computer equipment and storage medium | |
| Chung et al. | A Study of Single Image Haze Removal Using a Novel White-Patch RetinexBased Improved Dark Channel Prior Algorithm. | |
| CN111932470A (en) | Image restoration method, device, equipment and medium based on visual selection fusion | |
| CN111932469A (en) | Significance weight quick exposure image fusion method, device, equipment and medium | |
| CN112598777A (en) | Haze fusion method based on dark channel prior | |
| Unnikrishnan et al. | Non-Local Retinex Based Dehazing and Low Light Enhancement of Images. | |
| Guo et al. | Improved dark channel prior single image defogging |
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 | ||
| RJ01 | Rejection of invention patent application after publication | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20210611 |