CN111709887A - Image rain removing method based on sparse blind detection and image multiple feature restoration - Google Patents

Image rain removing method based on sparse blind detection and image multiple feature restoration Download PDF

Info

Publication number
CN111709887A
CN111709887A CN202010464988.2A CN202010464988A CN111709887A CN 111709887 A CN111709887 A CN 111709887A CN 202010464988 A CN202010464988 A CN 202010464988A CN 111709887 A CN111709887 A CN 111709887A
Authority
CN
China
Prior art keywords
image
rain
method based
multiple feature
blind detection
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.)
Granted
Application number
CN202010464988.2A
Other languages
Chinese (zh)
Other versions
CN111709887B (en
Inventor
张亚松
张琦
戚云西
华妮娜
相林
范媛媛
陈华松
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Huaiyin Institute of Technology
Original Assignee
Huaiyin Institute of Technology
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Huaiyin Institute of Technology filed Critical Huaiyin Institute of Technology
Priority to CN202010464988.2A priority Critical patent/CN111709887B/en
Publication of CN111709887A publication Critical patent/CN111709887A/en
Application granted granted Critical
Publication of CN111709887B publication Critical patent/CN111709887B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T5/00—Image enhancement or restoration
    • G06T5/73—Deblurring; Sharpening
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T5/00—Image enhancement or restoration
    • G06T5/77—Retouching; Inpainting; Scratch removal
    • 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
    • G06T2207/20048—Transform domain processing
    • G06T2207/20056—Discrete and fast Fourier transform, [DFT, FFT]
    • 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
    • Y02A90/10—Information and communication technologies [ICT] supporting adaptation to climate change, e.g. for weather forecasting or climate simulation

Landscapes

  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Image Processing (AREA)

Abstract

The invention relates to the technical field of image processing, and discloses an image rain removing method based on sparse blind detection and image multiple feature restoration, which comprises the following steps of: (1) detecting the distribution position of rain in the image through a rain mark detection model, and generating a rain mark distribution template; (2) and carrying out rain removing treatment and image information repairing treatment on the image area with rain by using the rain drop distribution template through the image rain removing and repairing model, and reserving original information on the area without rain pollution to realize rain removing and repairing of the image. Compared with the prior art, the invention can realize higher-performance image rain removal and image information protection.

Description

Image rain removing method based on sparse blind detection and image multiple feature restoration
Technical Field
The invention relates to the technical field of image processing, in particular to an image rain removing method based on sparse blind detection and image multiple feature restoration.
Background
At present, photoelectric imaging systems are widely used for outdoor information acquisition perception, for example: traffic monitoring, public safety, satellite remote sensing, and the like. When the imaging system works outdoors, imaging is inevitably needed to be performed in a rainy environment, however, the imaging quality of the imaging system is reduced in the rainy environment due to the influence of rain marks formed in the process of falling under the rain, and partial information of the image is shielded by the rain marks and even lost. This would present significant difficulties in the post-analysis and use of the images, which would require de-raining of the images obtained in a rainy environment. The existing image rain removing method mainly comprises a rain removing method based on image priori knowledge and an image rain removing method based on deep learning; the method based on the image priori knowledge mainly utilizes an image decomposition technology to erase rain layer information from a degraded image to remove rain, but the method has the problems of incomplete rain removal or over-smooth image in the rain removing process, because the intensity distribution of rain in the image is random; the deep learning method needs a sufficient number of clear/rain degradation image training sample sets, and the rain removal of the method can only be effective in removing rain to a certain degree and type; and the methods do not consider the distribution characteristics of rain marks in the image.
Aiming at the problem of rain marks of images, the invention provides an image rain removing method based on sparse blind detection and image information restoration, which finds that the rain marks are sparse in distribution in the images (as shown in figure 1), firstly detects the distribution position of rain in the images by using a sparse statistical model by using the distribution characteristic of rain space, and then carries out rain removing processing and image information restoration processing on image areas with rain, thereby realizing the rain removing of the images and the image information protection with higher performance.
Disclosure of Invention
The purpose of the invention is as follows: aiming at the problem of rain mark removal of an image, the invention provides an image rain removing method based on sparse blind detection and image multiple characteristic restoration, which comprises the steps of firstly detecting the distribution position of rain in the image by using a rain mark vertical gradient based on a sparse statistical model, and generating a function distribution template; then, the rain drop distribution template is utilized to carry out rain removal processing and image information restoration processing on the image area with rain, original information is reserved on the area without rain pollution, and high-performance image rain removal and image information protection are achieved.
The technical scheme is as follows: the invention provides an image rain removing method based on sparse blind detection and image multiple feature restoration, which comprises the following steps of:
(1) detecting the distribution position of rain in the image through a rain mark detection model, and generating a rain mark distribution template;
(2) and carrying out rain removing treatment and image information repairing treatment on the image area with rain by using the rain drop distribution template through the image rain removing and repairing model, and reserving original information on the area without rain pollution to realize rain removing and repairing of the image.
Further, in the step (1), the raindrop detection model is:
Figure BDA0002512300400000021
in the formula, s is a rain mark image to be detected, r is a rain pollution image, | | · luminance |L0Is a norm of L0 and,
Figure BDA0002512300400000022
and
Figure BDA0002512300400000023
are horizontal and vertical gradient operators; lambda [ alpha ]1，λ2，λ3，λ4And λ5Non-negative regularization coefficients.
Further, the solving process of the raindrop detection model is as follows:
introducing an auxiliary variable d1,d2,d3,d4,d5And make it possible to
Figure BDA0002512300400000024
According to the semiquadratic splitting principle, the rain drop detection model is transformed into the following unconstrained condition problem:
Figure BDA0002512300400000025
in the formula, α1,α2,α3,α4,α5Is a non-negative iteration coefficient introduced during the half-quadratic splitting.
Further, with respect to sk+1Solving the subproblems:
Figure BDA0002512300400000026
solving the formula (3) to obtain:
Figure BDA0002512300400000027
and (3) sorting and utilizing Fourier transform and inverse Fourier transform, wherein the sorting comprises the following steps:
Figure BDA0002512300400000028
in the formula, fft2 (-) and ifft2 (-) represent fourier transform and inverse fourier transform, respectively.
Further, as to
Figure BDA0002512300400000029
And
Figure BDA00025123004000000210
solving the subproblems:
Figure BDA00025123004000000211
Figure BDA0002512300400000031
Figure BDA0002512300400000032
Figure BDA0002512300400000033
Figure BDA0002512300400000034
using the hard threshold iteration principle, the following iteration formula can be obtained:
Figure BDA0002512300400000035
Figure BDA0002512300400000036
Figure BDA0002512300400000037
Figure BDA0002512300400000038
Figure BDA0002512300400000039
further, according to the iterative formula, the image rain detection process can be summarized as follows:
step 1-the initial setup is performed,
Figure BDA00025123004000000310
iteration stop condition error limit tol is 10-5The initial error is 0; the maximum iteration number itermax is 100, and the initial value iter of the iteration number is 1;
step2 while iter < itermax or error > tol
And (3) calculating:
Figure BDA00025123004000000311
Figure BDA00025123004000000312
Figure BDA00025123004000000313
Figure BDA00025123004000000314
Figure BDA0002512300400000041
Figure BDA0002512300400000042
Figure BDA0002512300400000043
step3: outputting a rain detection image s, and generating a rain distribution position template P by binarization^。
Further, in the step (2), the image rain removing and repairing model is:
Figure BDA0002512300400000044
in the above formula, D is a tight wavelet frame sparse transform operator, u is a clear image to be restored, and β is a non-negative real number coefficient.
Further, the solving process of the model equation (16) is as follows:
let p be Du, with a half-quadratic split, equation (16) can be transformed into:
Figure BDA0002512300400000045
then there is uk+1The optimization problem of (2):
Figure BDA0002512300400000046
obtaining by solution:
uk+1＝(P^+γI)-1(P^f+γDTpk)
pk+1the optimization problems are as follows:
Figure BDA0002512300400000047
with hard threshold filtering there are:
Figure BDA0002512300400000048
further, the process of performing rain removing processing and image information restoration processing on the image area with rain by the image rain removing and restoration model is as follows:
step1: initial setting, p0＝u0Iteration stop condition error limit tol is 0 and 10-4The initial error is 0; the maximum iteration number itermax is 100, and the initial value iter of the iteration number is 1;
step2: while iter < itermax or error > tol
And (3) calculating:
uk+1＝(P^+γI)-1(P^f+γDTpk)
Figure BDA0002512300400000051
Figure BDA0002512300400000052
step3: and outputting a repaired image result u.
Has the advantages that: aiming at the problem of rain mark removal of an image, the invention provides an image rain removing method based on sparse blind detection and image multiple characteristic restoration, which comprises the steps of firstly detecting the distribution position of rain in the image by using a rain mark vertical gradient based on a sparse statistical model, and generating a function distribution template; then, the rain drop distribution template is utilized to carry out rain removal processing and image information restoration processing on the image area with rain, original information is reserved on the area without rain pollution, and high-performance image rain removal and image information protection are achieved.
Drawings
Fig. 1 shows the spatial distribution characteristics of rain: (a) a map of rain intensity information, (b) a spatial distribution map of rain;
FIG. 2 is a process summarizing an image rain removal method based on sparse blind detection and image multi-feature restoration according to the present invention: (a) an image needing rain removal, (b) a rain distribution template detected in the first step, (c) an image area to be repaired after being mapped by the rain distribution template, and (d) an image after rain removal and repair;
fig. 3 shows a comparison of rain removal and image restoration: first column: rain drop degradation diagram, second column: neural network rain removal results, third column: rain removal and restoration results based on the total variation of directivity, fourth column: the method removes rain and recovers the result.
Detailed Description
The present invention will be described in detail with reference to the accompanying drawings.
Embodiment 1:
the embodiment provides an image rain removing method based on sparse blind detection and image multiple feature restoration, which comprises the following steps of:
(1) and detecting the distribution position of rain in the image through the rain mark detection model, and generating a rain mark distribution template.
The rain mark detection model comprises the following steps:
Figure BDA0002512300400000053
where s is a rain mark image to be detected (see fig. 1(a)), r is a rain pollution image (see fig. 2(a)), | | · | | luminous |, andL0is a norm of L0 and,
Figure BDA0002512300400000054
and
Figure BDA0002512300400000055
are horizontal and vertical gradient operators; lambda [ alpha ]1，λ2，λ3，λ4And λ5Non-negative regularization coefficients.
The solving process of the rain mark detection model is as follows:
introducing an auxiliary variable d1,d2,d3,d4,d5And make it possible to
Figure BDA0002512300400000061
Then the half-quadratic splitting principle, the above model can be transformed into the following unconstrained problem:
Figure BDA0002512300400000062
in the formula, α1,α2,α3,α4,α5Is a non-negative iteration coefficient introduced during the half-quadratic splitting.
Equation (1) is a multivariable optimization problem that can be transformed into the following univariate optimization subproblems:
① pertaining to sk+1Solving the subproblems:
Figure BDA0002512300400000063
solving the formula (3) to obtain:
Figure BDA0002512300400000064
and (3) sorting and utilizing Fourier transform and inverse Fourier transform, wherein the sorting comprises the following steps:
Figure BDA0002512300400000065
in the formula, fft2 (-) and ifft2 (-) represent fourier transform and inverse fourier transform, respectively.
② about
Figure BDA0002512300400000066
And
Figure BDA0002512300400000067
solving the subproblems:
Figure BDA0002512300400000068
Figure BDA0002512300400000069
Figure BDA00025123004000000610
Figure BDA00025123004000000611
Figure BDA0002512300400000071
using the hard threshold iteration principle, the following iteration formula can be obtained:
Figure BDA0002512300400000072
Figure BDA0002512300400000073
Figure BDA0002512300400000074
Figure BDA0002512300400000075
Figure BDA0002512300400000076
according to the above iterative formula of variables, the image rain detection process can be summarized as follows:
step 1-the initial setup is performed,
Figure BDA0002512300400000077
iteration stop condition error limit tol is 10-5The initial error is 0; the maximum iteration number itermax is 100, and the initial iteration number iter is 1.
Step2 while iter < itermax or error > tol
And (3) calculating:
Figure BDA0002512300400000078
Figure BDA0002512300400000079
Figure BDA00025123004000000710
Figure BDA00025123004000000711
Figure BDA00025123004000000712
Figure BDA00025123004000000713
Figure BDA00025123004000000714
step3, outputting a rain detection image s, and generating a rain distribution position template by binarizationP^(see FIG. 2 (b)).
(2) By using the rain drop distribution template, the rain removing processing and the image information repairing processing are carried out on the image area with rain through the image rain removing and repairing model, original information is reserved for the area without rain pollution, and the rain removing and repairing of the image are realized (as shown in fig. 2 (d)).
The image rain removal and restoration model comprises the following steps:
Figure BDA0002512300400000081
in the above formula, D is a tight wavelet frame sparse transform operator, u is a clear image to be restored, and β is a non-negative real number coefficient.
The solving process of the model formula (16) is as follows:
let p be Du, with a half-quadratic split, equation (16) can be transformed into:
Figure BDA0002512300400000082
then there is uk+1The optimization problem of (2):
Figure BDA0002512300400000083
obtaining by solution:
uk+1＝(P^+γI)-1(P^f+γDTpk)
pk+1the optimization problems are as follows:
Figure BDA0002512300400000084
with hard threshold filtering there are:
Figure BDA0002512300400000085
the image rain removal and restoration process in the second step can be summarized as follows:
step1 initialization setting, p0＝u0Iteration stop condition error limit tol is 0 and 10-4The initial error is 0; the maximum iteration number itermax is 100, and the initial iteration number iter is 1.
Step2 while iter < itermax or error > tol
And (3) calculating:
uk+1＝(P^+γI)-1(P^f+γDTpk)
Figure BDA0002512300400000091
Figure BDA0002512300400000092
step3, outputting a repaired image result u.
The rain detection parameters of the above image take values in the experiment as follows: lambda [ alpha ]1＝0.01,λ2＝0.001,λ3＝0.004,λ4＝3×10-4,λ50.19, β -0.3, and 3-3 × 10-4. The above parameters may be preferred to be ideal values according to the actual application.
The results of the rain removing process performed on the pictures shown in the first column of fig. 3 according to the above method are compared as shown in fig. 3 (wherein, the second column of the image is the rain removing result of the deep neural network, the third column is the rain removing result based on the directional total variation method, and the last column is the rain removing result of the present invention). The image after rain removal still has a certain amount of rain marks and is not completely removed based on the deep neural network and the method based on the directional total variation, and some images are blurred after rain removal, so that the target characteristic protection is insufficient; the image after rain is removed by the method of the invention has the advantages that rain marks are sufficiently removed, and the target information of the processed image is well protected. Therefore, the method of the invention only carries out the rain removing treatment and the image information repairing treatment on the image area with rain, and reserves the original information for the area without rain pollution, thereby realizing the higher-performance rain removing of the image and the image information protection.
The above embodiments are merely illustrative of the technical concepts and features of the present invention, and the purpose of the embodiments is to enable those skilled in the art to understand the contents of the present invention and implement the present invention, and not to limit the protection scope of the present invention. All equivalent changes and modifications made according to the spirit of the present invention should be covered within the protection scope of the present invention.

Claims (9)

1. An image rain removing method based on sparse blind detection and image multiple feature restoration is characterized by comprising the following steps:
(1) detecting the distribution position of rain in the image through a rain mark detection model, and generating a rain mark distribution template;
(2) and carrying out rain removing treatment and image information repairing treatment on the image area with rain by using the rain drop distribution template through the image rain removing and repairing model, and reserving original information on the area without rain pollution to realize rain removing and repairing of the image.
2. The image rain removing method based on sparse blind detection and image multiple feature restoration according to claim 1, wherein in the step (1), the rain drop detection model is as follows:
Figure FDA0002512300390000011
in the above formula, s is the rain mark image to be detected, r is the rain pollution image, | · luminance |L0Is a norm of L0 and,
Figure FDA0002512300390000014
and
Figure FDA0002512300390000015
are horizontal and vertical gradient operators; lambda [ alpha ]1，λ2，λ3，λ4And λ5Non-negative regularization coefficients.
3. The image rain removing method based on sparse blind detection and image multiple feature restoration according to claim 2, wherein the solution process of the rain drop detection model is as follows:
introducing an auxiliary variable d1,d2,d3,d4,d5And make it possible to
Figure FDA0002512300390000016
d5＝s。
According to the semiquadratic splitting principle, the rain drop detection model is transformed into the following unconstrained condition problem:
Figure FDA0002512300390000012
in the formula, α1,α2,α3,α4,α5Is a non-negative iteration coefficient introduced during the half-quadratic splitting.
4. The image rain removal method based on sparse blind detection and image multiple feature restoration as claimed in claim 3, wherein regarding sk+1Solving the subproblems:
Figure FDA0002512300390000013
solving the formula (3) to obtain:
Figure FDA0002512300390000021
and (3) sorting and utilizing Fourier transform and inverse Fourier transform, wherein the sorting comprises the following steps:
Figure FDA0002512300390000022
in the formula, fft2 (-) and ifft2 (-) represent fourier transform and inverse fourier transform, respectively.
5. The image rain removal method based on sparse blind detection and image multiple feature restoration as claimed in claim 3, wherein the method relates to
Figure FDA0002512300390000023
And
Figure FDA0002512300390000024
solving the subproblems:
Figure FDA0002512300390000025
Figure FDA0002512300390000026
Figure FDA0002512300390000027
Figure FDA0002512300390000028
Figure FDA0002512300390000029
using the hard threshold iteration principle, the following iteration formula can be obtained:
Figure FDA00025123003900000210
Figure FDA00025123003900000211
Figure FDA00025123003900000212
Figure FDA00025123003900000213
Figure FDA0002512300390000031
6. the image rain removing method based on sparse blind detection and image multiple feature restoration according to claim 4, wherein according to the iterative formula, the image rain detection process can be summarized as follows:
at Step1, the setup is initialized,
Figure FDA0002512300390000032
iteration stop condition error limit tol is 10-5The initial error is 0; the maximum iteration number itermax is 100, and the initial value iter of the iteration number is 1;
step2 while iter < itermax or error > tol
And (3) calculating:
Figure FDA0002512300390000033
Figure FDA0002512300390000034
Figure FDA0002512300390000035
Figure FDA0002512300390000036
Figure FDA0002512300390000037
Figure FDA0002512300390000038
Figure FDA0002512300390000039
iter＝iter+1；
step3: outputting a rain detection image s, and generating a rain distribution position template P by binarization^。
7. The image rain removing method based on sparse blind detection and image multiple feature restoration according to any one of claims 1 to 6, wherein in the step (2), the image rain removing and restoration model is as follows:
Figure FDA00025123003900000310
in the above formula, D is a tight wavelet frame sparse transform operator, u is a clear image to be restored, and β is a non-negative real number coefficient.
8. The image rain removing method based on sparse blind detection and image multiple feature restoration as claimed in claim 7, wherein the solving process of the model formula (16) is as follows:
let p be Du, with a half-quadratic split, equation (16) can be transformed into:
Figure FDA0002512300390000041
then there is uk+1The optimization problem of (2):
Figure FDA0002512300390000042
obtaining by solution:
uk+1＝(P^+γI)-1(P^f+γDTpk)
pk+1the optimization problems are as follows:
Figure FDA0002512300390000043
with hard threshold filtering there are:
Figure FDA0002512300390000044
9. the image rain removing method based on sparse blind detection and image multiple feature restoration according to claim 7, wherein the rain removing processing and the image information restoration processing are performed on the image area with rain through an image rain removing and restoration model as follows:
step1: initial setting, p0＝u0Iteration stop condition error limit tol is 0 and 10-4The initial error is 0; the maximum iteration number itermax is 100, and the initial value iter of the iteration number is 1;
step2: while iter < itermax or error > tol
And (3) calculating:
uk+1＝(P^+γI)-1(P^f+γDTpk)
Figure FDA0002512300390000045
Figure FDA0002512300390000046
iter＝iter+1；
step3: and outputting a repaired image result u.
CN202010464988.2A 2020-05-28 2020-05-28 Image rain removing method based on sparse blind detection and image multiple feature restoration Active CN111709887B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202010464988.2A CN111709887B (en) 2020-05-28 2020-05-28 Image rain removing method based on sparse blind detection and image multiple feature restoration

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202010464988.2A CN111709887B (en) 2020-05-28 2020-05-28 Image rain removing method based on sparse blind detection and image multiple feature restoration

Publications (2)

Publication Number Publication Date
CN111709887A true CN111709887A (en) 2020-09-25
CN111709887B CN111709887B (en) 2023-09-22

Family

ID=72538044

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202010464988.2A Active CN111709887B (en) 2020-05-28 2020-05-28 Image rain removing method based on sparse blind detection and image multiple feature restoration

Country Status (1)

Country Link
CN (1) CN111709887B (en)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113902931A (en) * 2021-09-17 2022-01-07 淮阴工学院 Image rain removing method based on learning type convolution sparse coding

Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20150287223A1 (en) * 2014-04-04 2015-10-08 The Board Of Trustees Of The University Of Illinois Highly accelerated imaging and image reconstruction using adaptive sparsifying transforms
CN108520501A (en) * 2018-03-30 2018-09-11 西安交通大学 A kind of video and removes rain snow method based on multiple dimensioned convolution sparse coding
CN110544215A (en) * 2019-08-23 2019-12-06 淮阴工学院 traffic monitoring image rain removing method based on anisotropic sparse gradient

Patent Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20150287223A1 (en) * 2014-04-04 2015-10-08 The Board Of Trustees Of The University Of Illinois Highly accelerated imaging and image reconstruction using adaptive sparsifying transforms
CN108520501A (en) * 2018-03-30 2018-09-11 西安交通大学 A kind of video and removes rain snow method based on multiple dimensioned convolution sparse coding
CN110544215A (en) * 2019-08-23 2019-12-06 淮阴工学院 traffic monitoring image rain removing method based on anisotropic sparse gradient

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113902931A (en) * 2021-09-17 2022-01-07 淮阴工学院 Image rain removing method based on learning type convolution sparse coding

Also Published As

Publication number Publication date
CN111709887B (en) 2023-09-22

Similar Documents

Publication Publication Date Title
CN109035149B (en) License plate image motion blur removing method based on deep learning
CN109784333B (en) Three-dimensional target detection method and system based on point cloud weighted channel characteristics
CN108230264B (en) Single image defogging method based on ResNet neural network
CN108520501B (en) Video rain and snow removing method based on multi-scale convolution sparse coding
CN110503610B (en) GAN network-based image rain and snow trace removing method
CN110517203B (en) Defogging method based on reference image reconstruction
CN110544215B (en) Traffic monitoring image rain removing method based on anisotropic sparse gradient
CN105701785A (en) Image smog removing method based on sky region division and transmissivity optimization of weighting TV
CN111652812A (en) Image defogging and rain removing algorithm based on selective attention mechanism
CN113673590A (en) Rain removing method, system and medium based on multi-scale hourglass dense connection network
CN111833328A (en) Aircraft engine blade surface defect detection method based on deep learning
CN104050637A (en) Quick image defogging method based on two times of guide filtration
CN102129670B (en) Method for detecting and repairing movie scratch damage
CN111079822A (en) Method for identifying dislocation fault image of middle rubber and upper and lower plates of axle box rubber pad
CN107886065A (en) A kind of Serial No. recognition methods of mixing script
CN108665468B (en) Device and method for extracting tangent tower insulator string
CN115908354A (en) Photovoltaic panel defect detection method based on double-scale strategy and improved YOLOV5 network
CN112419163B (en) Single image weak supervision defogging method based on priori knowledge and deep learning
CN116309483A (en) DDPM-based semi-supervised power transformation equipment characterization defect detection method and system
CN111598814A (en) Single image defogging method based on extreme scattering channel
CN111709887A (en) Image rain removing method based on sparse blind detection and image multiple feature restoration
CN114663352A (en) High-precision detection method and system for defects of power transmission line and storage medium
CN106934344B (en) quick pedestrian detection method based on neural network
CN112862729B (en) Remote sensing image denoising method based on characteristic curve guidance
CN107464220A (en) A kind of highway surface layer disease geo-radar image Enhancement Method based on gravity Additive Model

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