CN110111282B - Video deblurring method based on motion vector and CNN - Google Patents
Video deblurring method based on motion vector and CNN Download PDFInfo
- Publication number
- CN110111282B CN110111282B CN201910383006.4A CN201910383006A CN110111282B CN 110111282 B CN110111282 B CN 110111282B CN 201910383006 A CN201910383006 A CN 201910383006A CN 110111282 B CN110111282 B CN 110111282B
- Authority
- CN
- China
- Prior art keywords
- video
- image block
- deblurring
- repaired
- cnn
- 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 36
- 239000013598 vector Substances 0.000 title claims abstract description 33
- 230000008439 repair process Effects 0.000 claims abstract description 20
- 238000001514 detection method Methods 0.000 claims abstract description 13
- 230000000903 blocking effect Effects 0.000 claims abstract description 11
- 238000000638 solvent extraction Methods 0.000 claims description 2
- 230000000694 effects Effects 0.000 abstract description 8
- 238000012545 processing Methods 0.000 abstract description 4
- 230000006870 function Effects 0.000 description 13
- 230000000295 complement effect Effects 0.000 description 4
- 238000010586 diagram Methods 0.000 description 4
- 230000005540 biological transmission Effects 0.000 description 2
- 238000004891 communication Methods 0.000 description 2
- 238000000354 decomposition reaction Methods 0.000 description 2
- 238000013135 deep learning Methods 0.000 description 2
- 238000010801 machine learning Methods 0.000 description 2
- 238000012544 monitoring process Methods 0.000 description 2
- 238000012216 screening Methods 0.000 description 2
- 230000000007 visual effect Effects 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000015556 catabolic process Effects 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 238000006731 degradation reaction Methods 0.000 description 1
- 238000001914 filtration Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000008707 rearrangement Effects 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
Images
Classifications
-
- G06T5/73—
-
- G06T5/77—
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/20—Analysis of motion
- G06T7/223—Analysis of motion using block-matching
- G06T7/238—Analysis of motion using block-matching using non-full search, e.g. three-step search
-
- 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
- G06T2207/10016—Video; Image sequence
-
- 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]
Abstract
The invention discloses a video deblurring method based on motion vectors and CNN, which comprises the following steps: s11, all video frames of a video to be repaired are processed in a blocking mode to generate a plurality of image blocks; s12, carrying out fuzzy detection on a plurality of image blocks of the video frame to obtain a fuzzy image block to be repaired; s13, searching for a clear image block corresponding to an adjacent frame of the to-be-repaired fuzzy image block; s14, inputting the to-be-repaired fuzzy image block and the clear image block into a CNN deblurring network; s15, the CNN deblurring network generates a deblurring result; and S16, completing the deblurring of the video to be repaired. The invention realizes the effect of efficient video fuzzy restoration, and is different from the method that the whole video frame is deblurred and only the fuzzy part in the video frame is restored. In addition, the invention also utilizes the advantages of CNN in the aspect of computer vision processing to obtain a clear video repair result.
Description
Technical Field
The invention relates to the technical field of video blurring, in particular to a video deblurring method based on motion vectors and CNN.
Background
Most information acquired by people is observed through a human visual system, and images and videos are taken as main visual information and gradually become one of main modes of information transmission in life, so that information communication modes among people are enriched and are not limited to language and character communication. Moreover, there is a need in many fields for images and video, such as medical images, to aid in understanding the condition of a patient; the remote sensing image is beneficial to monitoring the change of the environment; the video monitoring is widely applied to traffic intersections, railway stations, campuses and the like, and the safety of cities is guaranteed. However, different types of distortion may be introduced into the video during shooting and transmission, which may cause quality degradation, which may cause loss, and thus the quality of the video becomes a non-negligible problem.
The blurring distortion is a main factor causing the video quality to be reduced, so that the relevant research for developing the video deblurring algorithm has very important theoretical and practical significance. At present, existing video deblurring algorithms can be divided into two main categories: geometry-based methods and deep learning-based methods. The geometric method can be understood as a conventional block search method. Since the blurred region in the current video frame may be sharp in the video frames before and after the blurred region, the sharp region can be used to replace the blurred region, so that the main task of the algorithm is to search for the target sharp block in the video. Because the similarity between adjacent video frames is high, a common method is to search from the previous and subsequent frames of the blurred video frame. The CNN has a good effect in the field of video deblurring as a deep learning algorithm widely used in recent years.
For example, patent publication No. CN106447621A discloses a video image denoising method and apparatus based on fuzzy connection principle, the method includes: acquiring a video image signal, and extracting a first frame image of the video image signal; calculating connection parameters of the fuzzy connection complementary graph according to a preset fuzzy connection complementary graph and the first frame image; the fuzzy connection complementary graph consists of a first subsystem, a second subsystem, a plurality of amplifiers and fuzzy connection relations among the first subsystem, the second subsystem, the amplifiers and the third subsystem; the fuzzy connection relation comprises: series, parallel and feedback; and denoising the video image signals according to the connection parameters and the fuzzy connection complementary map to obtain a denoised video image. By adopting the embodiment of the invention, the filtering of the high-frequency signal noise can be realized, and the adaptability and the robustness are strong. Although it can denoise video, the repair for video screen is the whole video file and can not repair the fuzzy part in the video frame.
Disclosure of Invention
The invention aims to provide a video deblurring method based on motion vectors and CNN (compressed natural number). Firstly, positioning a fuzzy image block in a video; secondly, finding out a corresponding clear image block of the fuzzy image block in an adjacent reference frame by using the motion vector; then, screening through a designed objective function to find an optimal clear image block; and finally, inputting the fuzzy image block and the optimal clear image block into the CNN to obtain a repairing result of the fuzzy image block in the video, and replacing the fuzzy image block with the repairing result to realize the deblurring effect of the whole video.
In order to achieve the purpose, the invention adopts the following technical scheme:
a video deblurring method based on motion vectors and CNN comprises the following steps:
s1, partitioning all video frames of a video to be repaired to generate a plurality of image blocks;
s2, carrying out fuzzy detection on a plurality of image blocks of the video frame to obtain a fuzzy image block to be repaired;
s3, searching for a clear image block corresponding to an adjacent frame of the to-be-repaired fuzzy image block;
s4, inputting the to-be-repaired fuzzy image block and the clear image block into a CNN deblurring network;
s5, the CNN deblurring network generates a deblurring result;
and S6, completing the deblurring of the video to be repaired.
Further, the step S4 is preceded by the steps of:
and searching the optimal clear image block of the clear image blocks corresponding to the adjacent frames.
Further, the step S5 specifically includes:
the CNN deblurring network generates a deblurring result, and the deblurring result is a repairing result of the to-be-repaired blurred image block.
Further, the step S6 specifically includes:
and replacing the generated repairing result with the to-be-repaired fuzzy image block to generate a clear video.
Further, the step S1 specifically includes:
each video frame of a video to be repaired is processed in a blocking mode;
using F to repair videoiWherein i represents a video frame index number;
blocking each video frame to generate several image blocks, wherein the image blocks are BkAnd (4) showing.
Further, the step S2 specifically includes:
blurring detecting a number of image blocks of the video frame,
Sk＝Blur(Bk)，
wherein S iskRepresenting image blocks BkThe fuzzy score of (1).
Further, in step S3, finding a clear image block corresponding to a frame adjacent to the to-be-repaired blurred image block is performed by using a motion vector.
Further, the optimal image block of the clear image blocks corresponding to the adjacent frames is found through an objective function, where the objective function is the following formula:
F＝W1·SSIM(P,Q)+W2·PSNR(P,Q)；
wherein, P is a fuzzy image block, Q is a clear image block, SSIM () function is to calculate the similarity between the fuzzy image block P and the clear image block Q, and PSNR () function is to calculate the definition between the fuzzy image block P and the clear image block Q. W1Weight, W, representing similarity2A weight representing the sharpness.
Further, the step S4 specifically includes inputting the to-be-repaired blurred image block and the found optimal clear image block into a CNN deblurring network.
Compared with the prior art, the method has the advantages that the effect of efficient video fuzzy restoration is achieved, the motion vector is used for searching for clear image blocks in the video, the motion vector has smaller time complexity compared with the traditional characteristic point matching and geometric constraint, the method is different from the method that the whole video frame is deblurred, and only the fuzzy part in the video frame is restored; in addition, the invention also utilizes the advantages of CNN in the aspect of computer vision processing to obtain a clear video repair result.
Drawings
Fig. 1 is a flowchart of a video deblurring method based on motion vectors and CNN according to an embodiment one;
fig. 2 is a flow chart of a video deblurring method based on motion vectors and CNN according to the second embodiment;
FIG. 3 is a block candidate map found by a motion vector provided in the first and second embodiments;
FIG. 4 is a diagram of an optimal candidate block found by an objective function according to the second embodiment;
fig. 5 is a structure diagram of the CNN deblurring network provided in the first and second embodiments;
FIG. 6 is a diagram of a repair result of a blurred image block provided in the first and second embodiments;
FIG. 7 is a diagram of the repair result of the video frame provided in the first and second embodiments;
fig. 8 is a flowchart of a video deblurring method based on motion vectors and CNN according to the second embodiment.
Detailed Description
The embodiments of the present invention are described below with reference to specific embodiments, and other advantages and effects of the present invention will be easily understood by those skilled in the art from the disclosure of the present specification. The invention is capable of other and different embodiments and of being practiced or of being carried out in various ways, and its several details are capable of modification in various respects, all without departing from the spirit and scope of the present invention. It is to be noted that the features in the following embodiments and examples may be combined with each other without conflict.
The invention aims to provide a video deblurring method based on motion vectors and CNN (compressed natural number). Firstly, positioning a fuzzy image block in a video; secondly, finding out a corresponding clear image block of the fuzzy image block in an adjacent reference frame by using the motion vector; then, screening through a designed objective function to find an optimal clear image block; and finally, inputting the fuzzy image block and the optimal clear image block into the CNN to obtain a repairing result of the fuzzy image block in the video, and replacing the fuzzy image block with the repairing result to realize the deblurring effect of the whole video.
Example one
The present embodiment provides a video deblurring method based on motion vectors and CNN, as shown in fig. 1, including the steps of:
s11, all video frames of a video to be repaired are processed in a blocking mode to generate a plurality of image blocks;
s12, carrying out fuzzy detection on a plurality of image blocks of the video frame to obtain a fuzzy image block to be repaired;
s13, searching for a clear image block corresponding to an adjacent frame of the to-be-repaired fuzzy image block;
s14, inputting the to-be-repaired fuzzy image block and the clear image block into a CNN deblurring network;
s15, the CNN deblurring network generates a deblurring result;
and S16, completing the deblurring of the video to be repaired.
In step S11, all video frames of the video to be repaired are processed in blocks to generate a plurality of image blocks.
Specifically, each video frame of a video to be repaired is processed in a blocking mode, wherein the size of a block is represented by p × p;
using F to repair videoiWhere i represents the video frame index number, the ith frame is denoted as Fi；
Blocking each video frame to generate several image blocks, wherein the image blocks are BkDenotes, k ═ 1,2,3 … … N, where N denotes the video F to be repairediMiddle image block BkThe total number of image blocks, the size of the image block, is represented by p × p. In the present embodiment, p is 128。
It should be noted that, in practical application, the value of p is not limited to the value set in the embodiment, but is set according to practical situations.
In step S12, blur detection is performed on the plurality of image blocks of the video frame to obtain a blurred image block to be repaired.
Specifically, a plurality of image blocks of the video frame are subjected to fuzzy detection,
Sk＝Blur(Bk)，
wherein S iskRepresenting image blocks BkThe fuzzy score of (1).
In this embodiment, a threshold is preset for determining the image block BkIf clear, then SkWhen the value is larger than a threshold value, the image block BkThe image is a clear image block; when S iskWhen the image block is smaller than a threshold value, the image block BkAnd if the image block is a blurred image block, the obtained blurred image block is the image block to be repaired.
In this embodiment, the set threshold is 15, then when Sk>At time 15, the image block BkIf the image is a clear image block, the image does not need to be repaired; when S isk<At time 15, the image block BkIn order to blur the image blocks, repair is required.
In the present embodiment, the Blur () function is a method of Blur detection, specifically referring to the Blur detection method based on SVD decomposition of image DCT domain provided in the patent with publication number CN108510496A, which is a patent that we applied before.
In step S13, a clear image block corresponding to a frame adjacent to the to-be-repaired blurred image block is searched for.
And searching for a clear image block corresponding to an adjacent frame of the to-be-repaired blurred image block through a motion vector.
Specifically, a motion vector is used to search for a clear image block corresponding to a to-be-repaired blurred image block in an adjacent frame (i.e., a reference frame) of the blurred image block, and the found clear image block is called a candidate block.
In this embodiment, the selected phaseThe adjacent frames (i.e., reference frames) are the first two frames and the last two frames of the frame (i.e., current frame) corresponding to the current blurred image block. E.g. the current frame FiThe first two frames of (2) are Fi-2、Fi-1(ii) a The last two frames are Fi+1、Fi+2. As shown in fig. 3.
In step S14, the blurred image block to be repaired and the clear image block are input into a CNN deblurring network.
In this embodiment, a blurred image block to be repaired and the clear image block are input into a CNN deblurring network through an encoder.
After the blurred and sharp images are input into the spatio-recursive CNN deblurring network, the CNN deblurring network performs a deblurring operation.
The spatio-temporal recursive CNN deblurring network is a network in machine learning, which adds an image model to deep network learning to form the spatio-temporal recursive CNN deblurring network. After the blurred image and the sharp image are input into the spatio-recursive CNN deblurring network, the CNN deblurring network automatically executes the deblurring operation.
The CNN deblurring network structure is shown in fig. 5.
In step S15, the CNN deblurring network generates a deblurred result.
In this embodiment, the result of the CNN deblurring network is output by the decoder, and the output deblurring result is used as the repair result of the blurred image block to be repaired.
The repair results are shown in fig. 6.
In step S16, the deblurring of the video to be repaired is completed, and the deblurring of the entire video is realized.
In this embodiment, the generated restoration result is replaced with the to-be-restored blurred image block to generate a clear video, so as to achieve deblurring of the whole video.
The repair result of the video frame is shown in fig. 7.
The method realizes the effect of efficient video fuzzy restoration, uses the motion vector to search for clear image blocks in the video, has smaller time complexity compared with the traditional characteristic point matching and geometric constraint, is different from the method that the whole video frame is deblurred, and only restores the fuzzy part in the video frame; in addition, the invention also utilizes the advantages of CNN in the aspect of computer vision processing to obtain a clear video repair result.
Example two
The present embodiment provides a video deblurring method based on motion vectors and CNN, as shown in fig. 2, including the steps of:
s11, all video frames of a video to be repaired are processed in a blocking mode to generate a plurality of image blocks;
s12, carrying out fuzzy detection on a plurality of image blocks of the video frame to obtain a fuzzy image block to be repaired;
s13, searching for a clear image block corresponding to an adjacent frame of the to-be-repaired fuzzy image block;
s131, searching the optimal clear image block of the clear image blocks corresponding to the adjacent frames.
S14, inputting the to-be-repaired fuzzy image block and the optimal clear image block into a CNN deblurring network;
s15, the CNN deblurring network generates a deblurring result;
and S16, completing the deblurring of the video to be repaired.
The specific flow is shown in fig. 8.
In step S11, all video frames of the video to be repaired are processed in blocks to generate a plurality of image blocks.
Specifically, each video frame of a video to be repaired is processed in a blocking mode, wherein the size of a block is represented by p × p;
using F to repair videoiWhere i represents the video frame index number, the ith frame is denoted as Fi；
Blocking each video frame to generate several image blocks, wherein the image blocks are BkDenotes, k ═ 1,2,3 … … N, where N denotes the video F to be repairediMiddle image block BkThe total number of image blocks, the size of the image block, is represented by p × p. In this embodiment, p is 128.
It should be noted that, in practical application, the value of p is not limited to the value set in the embodiment, but is set according to practical situations.
In step S12, blur detection is performed on the plurality of image blocks of the video frame to obtain a blurred image block to be repaired.
Specifically, a plurality of image blocks of the video frame are subjected to fuzzy detection,
Sk＝Blur(Bk)，
wherein S iskRepresenting image blocks BkThe fuzzy score of (1).
In this embodiment, a threshold is preset for determining the image block BkIf clear, then SkWhen the value is larger than a threshold value, the image block BkThe image is a clear image block; when S iskWhen the image block is smaller than a threshold value, the image block BkAnd if the image block is a blurred image block, the obtained blurred image block is the image block to be repaired.
In this embodiment, the set threshold is 15, then when Sk>At time 15, the image block BkIf the image is a clear image block, the image does not need to be repaired; when S isk<At time 15, the image block BkIn order to blur the image blocks, repair is required.
In the present embodiment, the Blur () function is a method of Blur detection, specifically referring to the Blur detection method based on SVD decomposition of image DCT domain provided in the patent with publication number CN108510496A, which is a patent that we applied before.
In step S13, a clear image block corresponding to a frame adjacent to the to-be-repaired blurred image block is searched for.
And searching for a clear image block corresponding to an adjacent frame of the to-be-repaired blurred image block through a motion vector.
Specifically, a motion vector is used to search for a clear image block corresponding to a to-be-repaired blurred image block in an adjacent frame (i.e., a reference frame) of the blurred image block, and the found clear image block is called a candidate block.
In this embodiment, adjacent frames are selected (i.e., adjacent frames are selected)Reference frame) are the first two frames and the last two frames of the frame corresponding to the current blurred image block (i.e., the current frame). E.g. the current frame FiThe first two frames of (2) are Fi-2、Fi-1(ii) a The last two frames are Fi+1、Fi+2. As shown in fig. 3.
In step S131, an optimal clear image block of the clear image blocks corresponding to the adjacent frames is searched.
Since the motion vector has a certain error, the candidate block found by the motion vector is shown in fig. 3. In order to find an optimal candidate block that is similar to and clear from a blurred image block, the search range in the reference frame is expanded. The candidate block found by the motion vector is used as the center, 5 pixels are expanded in the up, down, left and right directions, and the size of the search box is 138 × 138. Finding an optimal candidate block in each reference frame by using a designed objective function, wherein the formula of the objective function is as follows:
F＝W1·SSIM(P,Q)+W2·PSNR(P,Q)；
wherein, P is a fuzzy image block, Q is a clear image block, SSIM () function is to calculate the similarity between the fuzzy image block P and the clear image block Q, and PSNR () function is to calculate the definition between the fuzzy image block P and the clear image block Q. W1Weight, W, representing similarity2A weight representing the sharpness.
In the present embodiment, W is set1＝0.6，W2The best candidate block found is shown in fig. 4, which is 0.4.
In practical applications, W is1、W2The value of (b) is not limited to the value set in the embodiment, but is set according to the actual situation.
In step S14, the blurred image block to be repaired and the optimal sharp image block are input into a CNN deblurring network.
In this embodiment, the blurred image block to be repaired and the optimal sharp image block (optimal candidate block) are input into the CNN deblurring network through an encoder.
After the blurred image and the optimal candidate block are input into the null recursive CNN deblurring network, the CNN deblurring network performs a deblurring operation.
The spatio-temporal recursive CNN deblurring network is a network in machine learning, which adds an image model to deep network learning to form the spatio-temporal recursive CNN deblurring network. After the blurred image and the optimal candidate block are input into the null recursion CNN deblurring network, the CNN deblurring network automatically executes the deblurring operation.
The CNN deblurring network structure is shown in fig. 5.
In step S15, the CNN deblurring network generates a deblurred result.
In this embodiment, the result of the CNN deblurring network is output by the decoder, and the output deblurring result is used as the repair result of the blurred image block to be repaired.
The repair results are shown in fig. 6.
In step S16, the deblurring of the video to be repaired is completed, and the deblurring of the entire video is realized.
In this embodiment, the generated restoration result is replaced with the to-be-restored blurred image block to generate a clear video, so as to achieve deblurring of the whole video.
The repair result of the video frame is shown in fig. 7.
The method realizes the effect of efficient video fuzzy restoration, uses the motion vector to search for clear image blocks in the video, has smaller time complexity compared with the traditional characteristic point matching and geometric constraint, is different from the method that the whole video frame is deblurred, and only restores the fuzzy part in the video frame; in addition, the invention also utilizes the advantages of CNN in the aspect of computer vision processing to obtain a clear video repair result.
It is to be noted that the foregoing is only illustrative of the preferred embodiments of the present invention and the technical principles employed. It will be understood by those skilled in the art that the present invention is not limited to the particular embodiments described herein, but is capable of various obvious changes, rearrangements and substitutions as will now become apparent to those skilled in the art without departing from the scope of the invention. Therefore, although the present invention has been described in greater detail by the above embodiments, the present invention is not limited to the above embodiments, and may include other equivalent embodiments without departing from the spirit of the present invention, and the scope of the present invention is determined by the scope of the appended claims.
Claims (7)
1. A video deblurring method based on motion vectors and CNN is characterized by comprising the following steps:
s1, partitioning all video frames of a video to be repaired to generate a plurality of image blocks;
s2, carrying out fuzzy detection on a plurality of image blocks of the video frame to obtain a fuzzy image block to be repaired;
s3, searching for a clear image block corresponding to an adjacent frame of the to-be-repaired fuzzy image block;
s4, inputting the to-be-repaired fuzzy image block and the clear image block into a CNN deblurring network;
s5, the CNN deblurring network generates a deblurring result;
s6, completing deblurring of the video to be repaired;
the step S4 is preceded by the steps of:
searching the optimal clear image block of the clear image blocks corresponding to the adjacent frames;
finding the optimal image block of the clear image blocks corresponding to the adjacent frames is found through an objective function, wherein the objective function is the following formula:
F＝W1·SSIM(P,Q)+W2·PSNR(P,Q)；
wherein P is a fuzzy image block, Q is a clear image block, SSIM () function is used for calculating the similarity between the fuzzy image block P and the clear image block Q, PSNR () function is used for calculating the definition between the fuzzy image block P and the clear image block Q, W1Weight, W, representing similarity2A weight representing the sharpness.
2. The video deblurring method based on motion vectors and CNN according to claim 1, wherein said step S5 specifically includes:
the CNN deblurring network generates a deblurring result, and the deblurring result is a repairing result of the to-be-repaired blurred image block.
3. The video deblurring method based on motion vectors and CNN according to claim 2, wherein said step S6 specifically includes:
and replacing the generated repairing result with the to-be-repaired fuzzy image block to generate a clear video.
4. The video deblurring method based on motion vectors and CNN according to claim 3, wherein said step S1 specifically includes:
each video frame of a video to be repaired is processed in a blocking mode;
using F to repair videoiWherein i represents a video frame index number;
blocking each video frame to generate several image blocks, wherein the image blocks are BkAnd (4) showing.
5. The video deblurring method based on motion vectors and CNN according to claim 4, wherein said step S2 specifically includes:
blurring detecting a number of image blocks of the video frame,
Sk＝Blur(Bk)，
wherein S iskRepresenting image blocks BkThe fuzzy score of (1).
6. The video deblurring method according to claim 5, wherein the step S3 of finding the clear image block corresponding to the frame adjacent to the blurred image block to be repaired is performed by using a motion vector.
7. The video deblurring method according to claim 6, wherein the step S4 specifically includes inputting the blurred image block to be repaired and the found optimal sharp image block into a CNN deblurring network.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910383006.4A CN110111282B (en) | 2019-05-09 | 2019-05-09 | Video deblurring method based on motion vector and CNN |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910383006.4A CN110111282B (en) | 2019-05-09 | 2019-05-09 | Video deblurring method based on motion vector and CNN |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN110111282A CN110111282A (en) | 2019-08-09 |
| CN110111282B true CN110111282B (en) | 2021-05-11 |
Family
ID=67489005
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910383006.4A Active CN110111282B (en) | 2019-05-09 | 2019-05-09 | Video deblurring method based on motion vector and CNN |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110111282B (en) |
Families Citing this family (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110555414B (en) * | 2019-09-05 | 2022-09-30 | 北京市商汤科技开发有限公司 | Target detection method, device, equipment and storage medium |
| CN110706176A (en) * | 2019-09-29 | 2020-01-17 | Tcl移动通信科技（宁波）有限公司 | Static processing method and device for video and computer readable storage medium |
| CN113452870B (en) * | 2020-03-26 | 2024-04-09 | 腾讯美国有限责任公司 | Video processing method and device |
| US11140416B1 (en) | 2020-03-26 | 2021-10-05 | Tencent America LLC | Method and apparatus for temporal smoothing for video |
| CN111614996B (en) * | 2020-04-07 | 2022-03-18 | 上海推乐信息技术服务有限公司 | Video repair method and system |
| CN112102212B (en) * | 2020-11-19 | 2021-02-09 | 腾讯科技（深圳）有限公司 | Video restoration method, device, equipment and storage medium |
| CN116993620B (en) * | 2023-08-31 | 2023-12-15 | 荣耀终端有限公司 | Deblurring method and electronic equipment |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103049891A (en) * | 2013-01-25 | 2013-04-17 | 西安电子科技大学 | Video image deblurring method based on self-adaption window selection |
| CN104103050A (en) * | 2014-08-07 | 2014-10-15 | 重庆大学 | Real video recovery method based on local strategies |
| CN104867111A (en) * | 2015-03-27 | 2015-08-26 | 北京理工大学 | Block-blur-kernel-set-based heterogeneous video blind deblurring method |
| CN105608446A (en) * | 2016-02-02 | 2016-05-25 | 北京大学深圳研究生院 | Video stream abnormal event detection method and apparatus |
| CN106530227A (en) * | 2016-10-27 | 2017-03-22 | 北京小米移动软件有限公司 | Image restoration method and device |
| CN107292838A (en) * | 2017-06-07 | 2017-10-24 | 汕头大学 | The image deblurring method split based on fuzzy region |
Family Cites Families (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101917550B (en) * | 2010-07-01 | 2012-11-14 | 清华大学 | High-spatial and temporal resolution video deblurring method and system |
| CN104301722B (en) * | 2014-11-03 | 2016-08-24 | 电子科技大学 | A kind of video flowing based on frequency domain obscures detection method |
| CN104469086B (en) * | 2014-12-19 | 2017-06-20 | 北京奇艺世纪科技有限公司 | A kind of video stabilization method and device |
| US10289951B2 (en) * | 2016-11-02 | 2019-05-14 | Adobe Inc. | Video deblurring using neural networks |
-
2019
- 2019-05-09 CN CN201910383006.4A patent/CN110111282B/en active Active
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103049891A (en) * | 2013-01-25 | 2013-04-17 | 西安电子科技大学 | Video image deblurring method based on self-adaption window selection |
| CN104103050A (en) * | 2014-08-07 | 2014-10-15 | 重庆大学 | Real video recovery method based on local strategies |
| CN104867111A (en) * | 2015-03-27 | 2015-08-26 | 北京理工大学 | Block-blur-kernel-set-based heterogeneous video blind deblurring method |
| CN105608446A (en) * | 2016-02-02 | 2016-05-25 | 北京大学深圳研究生院 | Video stream abnormal event detection method and apparatus |
| CN106530227A (en) * | 2016-10-27 | 2017-03-22 | 北京小米移动软件有限公司 | Image restoration method and device |
| CN107292838A (en) * | 2017-06-07 | 2017-10-24 | 汕头大学 | The image deblurring method split based on fuzzy region |
Non-Patent Citations (2)
| Title |
|---|
| Deep Video deblurring for hand-held cameras;Shuochen Su et al.;《2017 IEEE Conference on Computer Vision and Pattern Recognition (CVPR)》;IEEE;20171109;第237-246页 * |
| 基于运动分割的视频去模糊;谯从彬 等;《计算机辅助设计与图形学学报》;20151115;第27卷(第11期);第2108-2115页 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN110111282A (en) | 2019-08-09 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110111282B (en) | Video deblurring method based on motion vector and CNN | |
| CN111539879B (en) | Video blind denoising method and device based on deep learning | |
| US20200364515A1 (en) | System and method for joint image refinement and perception | |
| US10817984B2 (en) | Image preprocessing method and device for JPEG compressed file | |
| JP2006507775A (en) | Method and apparatus for measuring the quality of a compressed video sequence without criteria | |
| CN108900848B (en) | Video quality enhancement method based on self-adaptive separable convolution | |
| US8582915B2 (en) | Image enhancement for challenging lighting conditions | |
| Moorthy et al. | Efficient motion weighted spatio-temporal video SSIM index | |
| CN106210448B (en) | Video image jitter elimination processing method | |
| CN1471306A (en) | Scene change detectorand method thereof | |
| JP2011515981A (en) | Method and apparatus for encoding or decoding video signal | |
| CN108364282B (en) | Image mosaic detection method and image mosaic detection system | |
| US20200143517A1 (en) | Non-uniformity correction method for infrared image based on guided filtering and high-pass filtering | |
| CN107292830A (en) | Low-light (level) image enhaucament and evaluation method | |
| JP5364264B2 (en) | Location detection of block defect using neural network | |
| US8121199B2 (en) | Reducing the block effect in video file compression | |
| CN115035172A (en) | Depth estimation method and system based on confidence degree grading and inter-stage fusion enhancement | |
| JP4443767B2 (en) | Motion estimation method for reducing motion vector transmission cost | |
| CN107248143B (en) | Depth image restoration method based on image segmentation | |
| US7274366B2 (en) | Method and apparatus for detecting edge | |
| KR101707337B1 (en) | Multiresolution non-local means filtering method for image denoising | |
| CN104182940A (en) | Blurred image restoration method and system | |
| CN113256541B (en) | Method for removing water mist from drilling platform monitoring picture by machine learning | |
| Kaur et al. | Blind image deconvolution technique for image restoration using ant colony optimization | |
| CN112132757B (en) | General image restoration method based on neural network |
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 |