CN112381742A - Single image motion blur removing method and system - Google Patents
Single image motion blur removing method and system Download PDFInfo
- Publication number
- CN112381742A CN112381742A CN202011346852.8A CN202011346852A CN112381742A CN 112381742 A CN112381742 A CN 112381742A CN 202011346852 A CN202011346852 A CN 202011346852A CN 112381742 A CN112381742 A CN 112381742A
- Authority
- CN
- China
- Prior art keywords
- image
- kernel
- fuzzy
- calculating
- length
- 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.)
- Withdrawn
Links
- 238000000034 method Methods 0.000 title claims abstract description 47
- 230000009466 transformation Effects 0.000 claims abstract description 12
- 238000004364 calculation method Methods 0.000 claims description 19
- 238000001914 filtration Methods 0.000 claims description 14
- 239000013598 vector Substances 0.000 claims description 6
- 230000006835 compression Effects 0.000 claims description 3
- 238000007906 compression Methods 0.000 claims description 3
- 238000006243 chemical reaction Methods 0.000 claims description 2
- 229910052704 radon Inorganic materials 0.000 abstract description 4
- SYUHGPGVQRZVTB-UHFFFAOYSA-N radon atom Chemical compound [Rn] SYUHGPGVQRZVTB-UHFFFAOYSA-N 0.000 abstract description 4
- 238000003384 imaging method Methods 0.000 description 6
- 238000012545 processing Methods 0.000 description 6
- 239000011159 matrix material Substances 0.000 description 3
- 230000000295 complement effect Effects 0.000 description 2
- 230000000694 effects Effects 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 2
- 238000012544 monitoring process Methods 0.000 description 2
- 238000001228 spectrum Methods 0.000 description 2
- 238000012550 audit Methods 0.000 description 1
- 230000015572 biosynthetic process Effects 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000018109 developmental process Effects 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 238000004134 energy conservation Methods 0.000 description 1
- 238000000605 extraction Methods 0.000 description 1
- 238000011835 investigation Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 239000012925 reference material Substances 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 230000011218 segmentation Effects 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
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
- 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
- G06T2207/20048—Transform domain processing
Landscapes
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Image Processing (AREA)
Abstract
The invention provides a method for removing motion blur of a single image, which comprises the following steps: and S01, calculating the direction of a fuzzy kernel, and calculating the direction of the fuzzy kernel by using a threshold method on the basis of Radon transformation according to the characteristics of the image on a frequency domain. And S02, calculating the length of a fuzzy kernel, calculating the length of the fuzzy kernel according to the direction of the fuzzy kernel and the information of the input image, calculating to obtain the fuzzy kernel according to the direction of the fuzzy kernel and the length of the fuzzy kernel, and restoring the image according to the fuzzy kernel and the information of the original image. According to the method, the direction with interference is eliminated through a threshold value method, so that an error direction is avoided being found, and the accuracy of the obtained direction is ensured.
Description
Technical Field
The invention relates to the technical field of image restoration, in particular to a method and a system for removing motion blur of a single image.
Background
With the rapid development of security monitoring technology and the wide deployment of related facilities, videos and images recorded by security monitoring camera equipment become important reference materials for responsibility tracing, case investigation and judicial appraisal. However, when the imaging device is used for shooting and imaging, the imaging device is susceptible to factors such as a shooting object, a shooting device, a shooting object or a shooting environment, and the like, so that an image with motion blur is shot, important information such as a license plate, a human face, characters and the like cannot be recognized at all, and in most cases, the shooting device cannot shoot again or cannot shoot a clear image, so that the shot image loses significance. And the image restoration is carried out on the motion blurred image to obtain a clear original image, so that the method has important practical value.
The formation of motion blurred images can be modeled as follows:
Ib＝Io*q+ε (2-1)
wherein, IoRepresenting a sharp image; q is a motion blur kernel with two parameters: angle and length; ε represents random noise; i isbRepresenting a motion-blurred image acquired by an imaging device.
The purpose of motion-blurred image restoration is to restore a blurred image from an acquired blurred image IbTo recover I as realistically as possibleoThis requires the ambiguity kernel q to be known in advance. And (5) reversely solving a clear image on the premise that q is known in advance, and is called non-blind deblurring. However, in most cases, the blur kernel q is unknown, which is called blind deblurring. In this case, it is necessary to first find the parameters of the blur kernel q and then find the sharpened image using non-blind deblurring.
The patent (201610219056.5) captures and shoots a moving object in the same field of view by using a plurality of cameras of the same type with continuous shooting function to obtain an observation image; respectively carrying out motion blur function estimation and segmentation extraction on the obtained observation images to obtain respective motion blur target images; and obtaining a clear target image by using the obtained motion blurred target image and the motion blur function and adopting a spatial domain joint restoration algorithm. On one hand, the method needs to shoot a plurality of images under the condition of accurately matching the target position, needs to know the time difference between the two images, has higher requirements on equipment and imaging conditions, and cannot carry out restoration processing on a single motion blurred image.
The patent (201910980412.9, in the audit) proposes a motion blur restoration method based on Golay sequence complementary code word set, which adopts a flash shutter imaging technology to obtain continuous inter-frame complementary images, and comprises the following steps: constructing a basic Golay pairing matrix; performing dimension expansion on the Golay pairing matrix; selecting a code word according to the Golay pairing matrix after dimension expansion; controlling a camera to acquire a fuzzy image of a corresponding frame number according to the code word; PSF estimation is carried out according to the code words and the fuzzy image; performing total variation regularization joint restoration according to the blurred image and the PSF estimation value; and outputting the final image. The method requires taking multiple consecutive frame images and requires a code word to control the camera to correspond to the blurred image frame. The method cannot effectively and quickly restore the single motion blurred image.
A method in literature (blind restoration of motion blurred images based on dark channel constraint) is based on a dark channel theory, adopts a multi-scale idea, performs dark channel constraint processing on blurred images on each layer of scale, performs nonnegative constraint and energy conservation constraint processing on a point spread function, obtains an estimated blur kernel when the scale is maximum, and finally obtains a final estimated target image when a convergence condition is met through deconvolution. The method needs to perform multi-scale and continuous cycle calculation on the image, so that the processing time is increased; in addition, the method estimates the fuzzy core under a plurality of constraint conditions, and a plurality of limits are added to influence the final processing effect. Therefore, the method cannot restore the motion-blurred image quickly and accurately.
In the literature (motion blurred image PSF parameter estimation and image restoration research), the motion blurred direction in the point spread function PSF is solved by using the Radon transformation principle, the motion blurred scale is calculated by using the central dark fringe distance on the frequency spectrum of the motion blurred image, and finally the motion blurred image is restored by adopting a wiener filtering algorithm based on the estimated PSF parameters. Although the method uses Radon transformation to solve the direction of the blur kernel, the method only uses the distance between the central dark stripes on the frequency spectrum to calculate the length of the blur kernel, so that the length of the predicted blur kernel is inaccurate, the method has no robustness, and the method cannot carry out effective sharpening processing on motion blur images of various scenes and types.
Disclosure of Invention
The technical problem to be solved by the present invention is how to provide a simple and effective restoration method for a single motion-blurred image.
The invention solves the technical problems through the following technical means:
a single image motion blur removing method comprises the following steps:
s01, calculating the direction of a fuzzy kernel to obtain a frequency domain transformation result F of the input imageg(x, y) and for Fg(x, y) compression yields:
Fp(x，y)＝ln(1+|Fg(x，y)|)
at Fp(x, y) acquiring a pixel intensity accumulated value R projected by 1-180 degrees in each angle direction of the image centeri：
Ri＝[R1，R2，R3，…，R180]，i＝1，2，3，…，180；
Obtaining an accumulated value RiDirection corresponding to maximum: dm，dmThe judgment is carried out in two cases:
(1)
take d directlyr＝dmAs a result of the judgment;
(2)dme (0,5) U (175,180) U (85,95) at [5,85]∪[95，175]D in the direction corresponding to the maximum R valuekIf, if
Then get dr＝dkAs a result, otherwise still take dr＝dm；
Wherein d isrFor fuzzy kernel direction, m is 1,2,3, …, 180;
s02, calculating the length of a fuzzy core according to the direction d of the fuzzy corerInputting the information of the image, and calculating to obtain the length of the fuzzy core;
s03, calculating to obtain a fuzzy core according to the direction and the length of the fuzzy core;
and S04, restoring the picture according to the fuzzy kernel.
Further, the specific calculation method of the frequency domain conversion result in the step S01 is as follows: firstly, carrying out two-dimensional fast Fourier transform on a gray level image to obtain a frequency domain transform result Fg(x，y)：
Wherein f isgrayRepresenting a grayed image; f. ofgray(a, b) representation of a grayed-out image fgrayPixel values at coordinates (a, b); m represents a column of the gray image and N represents a row of the gray image; j represents an imaginary unit satisfying j2＝-1。
Further, the length calculating method for calculating the blur kernel in step 02 includes: along drMaking a straight line passing through the image in the direction, calculating a pixel intensity accumulated value on a vertical line of each pixel position on the straight line, obtaining a pixel intensity curve by the pixel position of the image on the straight line and the pixel intensity accumulated value on the vertical line of the pixel position, carrying out mean value filtering on the pixel intensity curve, and merging adjacent minimum value points; the place with the minimum value is indicated to have dark stripes, the average distance delta between the dark stripes is taken, the image size is L multiplied by L, and the length L of the fuzzy kernel isr＝L/Δ。
Further, the fuzzy core calculating method in step S03 specifically includes:
wherein q represents a linear blur kernel PSF; l and theta respectively represent length kernel angles of the motion blur kernel; x is the number of1And x2Abscissa and ordinate values representing a blur kernel, x being represented by the coordinates (x) of the blur kernel1，x2) The vectors of the components.
Further, the method for restoring the picture in step S04 includes: and acquiring a sharpening result graph by using the PSF and the original picture through wiener filtering.
The invention also provides a single image motion blur removing system, which comprises
A fuzzy kernel direction calculation module for obtaining the frequency domain transformation result F of the input imageg(x, y) compression yields:
Fp(x，y)＝ln(1+|Fg(x，y)|)
at Fp(x, y) acquiring a pixel intensity accumulated value R projected by 1-180 degrees in each angle direction of the image centeri：
Ri＝[R1，R2，R3，…，R180]，i＝1，2，3，…，180；
Obtaining an accumulated value RiDirection corresponding to maximum: dm，dmThe judgment is carried out in two cases:
(1)
take d directlyr＝dmAs a result of the judgment;
(2)dme (0,5) U (175,180) U (85,95) at [5,85]∪[95，175]D in the direction corresponding to the maximum R valuekIf, if
Then get dr＝dkAs a result, otherwise still take dr＝dm；
Wherein d isrFor fuzzy kernel direction, m is 1,2,3, …, 180;
a fuzzy kernel length calculating module for calculating the length of the fuzzy kernel according to the fuzzy kernel direction drInputting the information of the image, and calculating to obtain the length of the fuzzy core;
the fuzzy kernel calculation module is used for calculating to obtain a fuzzy kernel according to the fuzzy kernel direction and the fuzzy kernel length;
and the picture restoration module restores the picture according to the fuzzy core.
Further, a specific calculation method of the frequency domain transformation result in the fuzzy kernel direction calculation module is as follows: firstly, to the gray scale imageCarrying out two-dimensional fast Fourier transform on the image to obtain a frequency domain transform result Fg(x，y)：
Wherein f isgrayRepresenting a grayed image; f. ofgray(a, b) representation of a grayed-out image fgrayPixel values at coordinates (a, b); m represents a column of the gray image and N represents a row of the gray image; j represents an imaginary unit satisfying j2＝-1。
Further, the fuzzy kernel length calculating method in the fuzzy kernel length calculating module is as follows: along drMaking a straight line passing through the image in the direction, calculating a pixel intensity accumulated value on a vertical line of each pixel position on the straight line, obtaining a pixel intensity curve by the pixel position of the image on the straight line and the pixel intensity accumulated value on the vertical line of the pixel position, carrying out mean value filtering on the pixel intensity curve, and merging adjacent minimum value points; the place with the minimum value is indicated to have dark stripes, the average distance delta between the dark stripes is taken, the image size is L multiplied by L, and the length L of the fuzzy kernel isr＝L/Δ。
Further, the fuzzy core calculation module specifically executes the following processes:
wherein q represents a linear blur kernel PSF; l and theta respectively represent length kernel angles of the motion blur kernel; x is the number of1And x2Abscissa and ordinate values representing a blur kernel, x being represented by the coordinates (x) of the blur kernel1，x2) The vectors of the components.
Further, the image restoration module acquires a sharpening result image by using a PSF and an original image and using wiener filtering.
The invention has the advantages that:
because the frequency domain graph is provided with a white line near 0 degree (or 180 degrees) and 90 degrees, the white line is easy to cause misjudgment of results, and in order to eliminate the misjudgment and enable the results to be more accurate, the method selects 0-degree interval (0,5), 90-degree interval (85,95) and 180-degree interval (175,180) as threshold value judgment, so as to avoid finding wrong directions and ensure the accuracy of the obtained directions;
after the pixel intensity accumulated value on the vertical line of the corresponding position of each pixel in the direction is obtained, noise is reduced through mean value filtering, dark fringes are obtained by using a combined minimum value method, the length of a fuzzy kernel is calculated by using the average distance and the image size between the dark fringes, the distance between the predicted length and the real length is reduced, and the predicted length is more accurate.
The invention utilizes the characteristics of the motion blurred image to perform Fourier transform under the gray level image to obtain the frequency domain amplitude image. And analyzing the characteristics and the relation of the direction and the length of the motion blur kernel according to the characteristics of the frequency domain amplitude image, predicting the angle of the blur kernel, calculating the length of the model kernel according to the angle, further obtaining a point spread function of the motion blur kernel, and finally restoring a clear image by a non-blind deblurring method. The method can directly process a single image, can adapt to images of different scenes and types compared with the existing method, is simple and feasible, and has better image restoration effect.
Drawings
FIG. 1 is a block flow diagram of a motion blur removal method in accordance with an embodiment of the present invention;
FIG. 2 is a flow chart of calculating a direction of a blur kernel according to an embodiment of the present invention;
FIG. 3 is a flowchart illustrating a method for calculating a blur kernel length according to an embodiment of the present invention.
Detailed Description
In order to make the objects, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the embodiments of the present invention, and it is obvious that the described embodiments are some embodiments of the present invention, but not all embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
The embodiment provides a method for removing motion blur of a single image, as shown in fig. 1, comprising the following steps:
the method comprises the following steps: inputting any one motion blurred image: f (x, y);
step two: if the input image is a gray scale image, fgray(x, y) ═ f (x, y); if the input image is a color image (BGR three-channel image, B is a blue channel, G is a green channel, and R is a red channel), graying the input image to obtain a grayscale image fgray(x，y)：
fgray(x，y)＝0.1140×fB(x，y)+0.5870×fG(x，y)+0.2989×fR(x，y)
Wherein f isB(x, y) is the blue channel of the color image f (x, y), fG(x, y) is the green channel of the color image f (x, y), fR(x, y) is the red channel of the color image f (x, y).
Step three: performing two-dimensional fast Fourier transform on the gray level image to obtain a frequency domain transform result Fg(x，y)：
Wherein f isgrayRepresenting a grayed image; f. ofgray(a, b) representation of a grayed-out image fgrayPixel values at coordinates (a, b); m represents a column of the gray image and N represents a row of the gray image; j represents an imaginary unit satisfying j2＝-1。
Step four: and (3) compressing the frequency domain transformation result: take FgThe modulus of (x, y), plus 1, is taken from the natural logarithm to give:
Fp(x，y)＝ln(1+|Fg(x，y)|)
step five: at Fp(x, y) acquiring a pixel intensity accumulated value R projected by 1-180 degrees in each angle direction of the image centeri：
Ri＝[R1，R2，R3，…，R180]，i＝1，2，3，…，180；
Step six: as shown in FIG. 2, the above-mentioned integrated value R is obtainediDirection corresponding to maximum: dmTo prevent the wrong direction, the judgment is carried out in two cases:
(1)
directly taking dr as dm as a judgment result;
(2)dme (0,5) U (175,180) U (85,95) at [5,85]∪[95，175]D in the direction corresponding to the maximum R valuekIf, if
Then get dr＝dkAs a result, otherwise still take dr＝dm；
Wherein d isrFor fuzzy kernel direction, m is 1,2,3, …, 180;
step seven: as shown in fig. 3, the angle d is obtained according to the aboverAnd inputting the information of the picture to obtain the length l of the fuzzy kernelr。lrThe calculation method comprises the following steps: along drThe direction is taken as a straight line through the image, and the pixel intensity integrated value on the perpendicular line of each pixel position on the straight line is calculated. The intensity curves are mean filtered and the neighboring minima points are merged. The appearance of the minimum value indicates that dark streaks appear at the position. Taking the average distance delta between dark stripes, and setting the image size as L multiplied by L, the length L of the fuzzy kernelr＝L/Δ.
Step eight: according to the predicted angle drAnd fuzzy kernel length lrObtaining the PSF of the fuzzy core according to the formula (2-2) as follows:
wherein q represents a linear blur kernel PSF; l and theta are respectively representedLength kernel angle of motion blur kernel; x is the number of1And x2Abscissa and ordinate values representing a blur kernel, x being represented by the coordinates (x) of the blur kernel1，x2) The vectors of the components.
Step nine: and acquiring a sharpening result graph by using wiener filtering according to the PSF and the original picture.
In the method provided by this embodiment, for any motion-blurred image, under the condition of unknown blur kernel: (1) the direction of the fuzzy kernel can be calculated by using a threshold method on the basis of Radon transformation according to the characteristics of the image on the frequency domain, and the method can accurately acquire the direction of the fuzzy kernel. (2) And (3) calculating to obtain an intensity curve by using the direction and the information of the fuzzy graph obtained in the step (1), obtaining minimum value points through mean value filtering, and calculating the length of the fuzzy core according to the distance between the minimum values.
The embodiment also provides a single image motion blur removing system, which comprises
A fuzzy kernel direction calculation module: inputting any one motion blurred image: f (x, y); if the input image is a gray scale image, fgray(x, y) ═ f (x, y); if the input image is a color image (BGR three-channel image, B is a blue channel, G is a green channel, and R is a red channel), graying the input image to obtain a grayscale image fgray(x，y)：
fgray(x，y)＝0.1140×fB(x，y)+0.5870×fG(x，y)+0.2989×fR(x，y)
Wherein f isB(x, y) is the blue channel of the color image f (x, y), fG(x, y) is the green channel of the color image f (x, y), fR(x, y) is the red channel of the color image f (x, y).
Performing two-dimensional fast Fourier transform on the gray level image to obtain a frequency domain transform result Fg(x，y)：
Wherein f isgrayRepresenting a grayed image; f. ofgray(a, b) representation of a grayed-out image fgrayPixel values at coordinates (a, b); m represents a column of the gray image and N represents a row of the gray image; j represents an imaginary unit satisfying j2＝-1。
And (3) compressing the frequency domain transformation result: take FgThe modulus of (x, y), plus 1, is taken from the natural logarithm to give:
Fp(x，y)＝ln(1+|Fg(x，y)|)
a fuzzy kernel length calculation module: at Fp(x, y) acquiring a pixel intensity accumulated value R projected by 1-180 degrees in each angle direction of the image centeri：
Ri＝[R1，R2，R3，…，R180]，i＝1，2，3，…，180；
Obtaining the accumulated value RiDirection corresponding to maximum: drTo prevent the wrong direction, the judgment is carried out in two cases:
(1)
take d directlyr＝dmAs a result of the judgment;
(2)dme (0,5) U (175,180) U (85,95) at [5,85]∪[95，175]D in the direction corresponding to the maximum R valuekIf, if
Then get dr＝dkAs a result, otherwise still take dr＝dm；
Wherein d isrFor fuzzy kernel direction, m is 1,2,3, …, 180;
according to the angle d obtained aboverAnd inputting the information of the picture to obtain the length l of the fuzzy kernelr。lrThe calculation method comprises the following steps: along drThe direction is taken as a straight line through the image, and the pixel intensity integrated value on the perpendicular line of each pixel position on the straight line is calculated. Mean filtering the intensity curve and merging neighboring minimaA point of value. The appearance of the minimum value indicates that dark streaks appear at the position. Taking the average distance delta between dark stripes, and setting the image size as L multiplied by L, the length L of the fuzzy kernelr＝L/Δ.
A fuzzy kernel calculation module for calculating a fuzzy kernel according to the predicted angle drAnd fuzzy kernel length lrThe PSF of the fuzzy core is obtained according to the following formula:
wherein q represents a linear blur kernel PSF; l and theta respectively represent length kernel angles of the motion blur kernel; x is the number of1And x2Abscissa and ordinate values representing a blur kernel, x being represented by the coordinates (x) of the blur kernel1，x2) The vectors of the components.
And the picture restoration module acquires a sharpening result picture by using wiener filtering according to the PSF and the original picture.
The above examples are only intended to illustrate the technical solution of the present invention, but not to limit it; although the present invention has been described in detail with reference to the foregoing embodiments, it will be understood by those of ordinary skill in the art that: the technical solutions described in the foregoing embodiments may still be modified, or some technical features may be equivalently replaced; and such modifications or substitutions do not depart from the spirit and scope of the corresponding technical solutions of the embodiments of the present invention.
Claims (10)
1. A single image motion blur removing method is characterized in that: the method comprises the following steps:
s01, calculating the direction of a fuzzy kernel to obtain a frequency domain transformation result F of the input imageg(x, y) and for Fg(x, y) compression yields:
Fp(x,y)＝ln(1+|Fg(x,y)|)
at Fp(x, y) acquiring a pixel intensity accumulated value R projected by 1-180 degrees in each angle direction of the image centeri：
Ri＝[R1,R2,R3,…,R180],i＝1,2,3,…,180；
Obtaining an accumulated value RiDirection corresponding to maximum: dm，dmTwo cases are judged:
(1)
take d directlyr＝dmAs a result of the judgment;
(2)dme (0,5) U (175,180) U (85,95) in [5, 85%]∪[95,175]D in the direction corresponding to the maximum R valuekIf, if
Then get dr＝dkAs a result, otherwise still take dr＝dm；
Wherein d isrFor fuzzy kernel direction, m is 1,2,3, …, 180;
s02, calculating the length of a fuzzy core according to the direction d of the fuzzy corerInputting the information of the image, and calculating to obtain the length of the fuzzy core;
s03, calculating to obtain a fuzzy core according to the direction and the length of the fuzzy core;
and S04, restoring the picture according to the fuzzy kernel.
2. A method for motion blur removal of a single image according to claim 1, characterized by: the specific calculation method of the frequency domain conversion result in the step S01 is as follows: firstly, carrying out two-dimensional fast Fourier transform on a gray level image to obtain a frequency domain transform result Fg(x,y)：
Wherein f isgrayRepresenting a grayed image; f. ofgray(a, b) representation of a grayed-out image fgrayPixel values at coordinates (a, b); m represents a column of the gray scale image and N represents a row of the gray scale image(ii) a j represents an imaginary unit satisfying j2＝-1。
3. A method for motion blur removal of a single image according to claim 1 or 2, characterized by: the method for calculating the length of the fuzzy core in the step 02 comprises the following steps: along drMaking a straight line passing through the image in the direction, calculating a pixel intensity accumulated value on a vertical line of each pixel position on the straight line, obtaining a pixel intensity curve by the pixel position of the image on the straight line and the pixel intensity accumulated value on the vertical line of the pixel position, carrying out mean value filtering on the pixel intensity curve, and merging adjacent minimum value points; the place with the minimum value is indicated to have dark stripes, the average distance delta between the dark stripes is taken, the image size is L multiplied by L, and the length L of the fuzzy kernel isr＝L/Δ。
4. A single image motion blur removal algorithm as claimed in claim 3, characterized by: the fuzzy core calculation method in step S03 specifically includes:
wherein q represents a linear blur kernel PSF; l and theta respectively represent length kernel angles of the motion blur kernel; x is the number of1And x2Abscissa and ordinate values representing a blur kernel, x being represented by the coordinates (x) of the blur kernel1,x2) The vectors of the components.
5. The single image motion blur removal algorithm of claim 4, wherein: the method for restoring the picture in step S04 includes: and acquiring a sharpening result graph by using the PSF and the original picture through wiener filtering.
6. A single image motion blur removal system, comprising: comprises that
A fuzzy kernel direction calculation module for obtaining the frequency domain transformation result F of the input imageg(x, y) is compressed to obtain：
Fp(x,y)＝ln(1+|Fg(x,y)|)
At Fp(x, y) acquiring a pixel intensity accumulated value R projected by 1-180 degrees in each angle direction of the image centeri：
Ri＝[R1,R2,R3,…,R180],i＝1,2,3,…,180；
Obtaining an accumulated value RiDirection corresponding to maximum: dm，dmThe judgment is carried out in two cases:
(1)
take d directlyr＝dmAs a result of the judgment;
(2)dme (0,5) U (175,180) U (85,95) in [5, 85%]∪[95,175]D in the direction corresponding to the maximum R valuekIf, if
Then get dr＝dkAs a result, otherwise still take dr＝dm；
Wherein d isrFor fuzzy kernel direction, m is 1,2,3, …, 180;
the fuzzy kernel length calculating module is used for calculating the length of the fuzzy kernel according to the direction d _ r of the fuzzy kernel and the information of the input image;
the fuzzy kernel calculation module is used for calculating to obtain a fuzzy kernel according to the fuzzy kernel direction and the fuzzy kernel length;
and the picture restoration module restores the picture according to the fuzzy core.
7. The single image motion blur removal system of claim 6, wherein: the specific calculation method of the frequency domain transformation result in the fuzzy kernel direction calculation module comprises the following steps: firstly, carrying out two-dimensional fast Fourier transform on a gray level image to obtain a frequency domain transform result Fg(x,y)：
Wherein f isgrayRepresenting a grayed image; f. ofgray(a, b) representation of a grayed-out image fgrayPixel values at coordinates (a, b); m represents a column of the gray image and N represents a row of the gray image; j represents an imaginary unit satisfying j2＝-1。
8. A single image motion blur removal system as claimed in claim 6 or 7, characterized by: the fuzzy kernel length calculating method in the fuzzy kernel length calculating module comprises the following steps: along drMaking a straight line passing through the image in the direction, calculating a pixel intensity accumulated value on a vertical line of each pixel position on the straight line, obtaining a pixel intensity curve by the pixel position of the image on the straight line and the pixel intensity accumulated value on the vertical line of the pixel position, carrying out mean value filtering on the pixel intensity curve, and merging adjacent minimum value points; the place with the minimum value is indicated to have dark stripes, the average distance delta between the dark stripes is taken, the image size is L multiplied by L, and the length L of the fuzzy kernel isr＝L/Δ。
9. A single image motion blur removal system as defined in claim 8, wherein: the fuzzy core calculation module specifically executes the following steps:
wherein q represents a linear blur kernel PSF; l and theta respectively represent length kernel angles of the motion blur kernel; x is the number of1And x2Abscissa and ordinate values representing a blur kernel, x being represented by the coordinates (x) of the blur kernel1,x2) The vectors of the components.
10. A single image motion blur removal system as defined in claim 9, wherein: the image restoration module adopts PSF and an original image to obtain a sharpening result image by using wiener filtering.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011346852.8A CN112381742A (en) | 2020-11-25 | 2020-11-25 | Single image motion blur removing method and system |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011346852.8A CN112381742A (en) | 2020-11-25 | 2020-11-25 | Single image motion blur removing method and system |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN112381742A true CN112381742A (en) | 2021-02-19 |
Family
ID=74588538
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202011346852.8A Withdrawn CN112381742A (en) | 2020-11-25 | 2020-11-25 | Single image motion blur removing method and system |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN112381742A (en) |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113807246A (en) * | 2021-09-16 | 2021-12-17 | 平安普惠企业管理有限公司 | Face recognition method, device, equipment and storage medium |
| CN115035000A (en) * | 2022-08-10 | 2022-09-09 | 山东国晟环境科技有限公司 | Road raise dust image identification method and system |
| CN115147415A (en) * | 2022-09-02 | 2022-10-04 | 山东微山湖酒业有限公司 | Wine box defect detection method based on image processing |
| CN116703785A (en) * | 2023-08-04 | 2023-09-05 | 普密特(成都)医疗科技有限公司 | Method for processing blurred image under minimally invasive surgery mirror |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102903078A (en) * | 2012-07-13 | 2013-01-30 | 南京大学 | motion-blurred image parameter estimation method based on multi-resolution Fourier analysis theory |
| US20180068430A1 (en) * | 2016-09-07 | 2018-03-08 | Huazhong University Of Science And Technology | Method and system for estimating blur kernel size |
| CN108376393A (en) * | 2018-03-16 | 2018-08-07 | 华南理工大学 | A kind of blurred picture blind restoration method towards high-speed straight-line Moving Objects |
-
2020
- 2020-11-25 CN CN202011346852.8A patent/CN112381742A/en not_active Withdrawn
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102903078A (en) * | 2012-07-13 | 2013-01-30 | 南京大学 | motion-blurred image parameter estimation method based on multi-resolution Fourier analysis theory |
| US20180068430A1 (en) * | 2016-09-07 | 2018-03-08 | Huazhong University Of Science And Technology | Method and system for estimating blur kernel size |
| CN108376393A (en) * | 2018-03-16 | 2018-08-07 | 华南理工大学 | A kind of blurred picture blind restoration method towards high-speed straight-line Moving Objects |
Non-Patent Citations (2)
| Title |
|---|
| 廖秋香等: "运动模糊图像 PSF 参数估计与图像复原研究", 《高技术通讯》 * |
| 范海菊等: "短模糊尺度下运动模糊参数的频域识别方法", 《计算机工程与设计》 * |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113807246A (en) * | 2021-09-16 | 2021-12-17 | 平安普惠企业管理有限公司 | Face recognition method, device, equipment and storage medium |
| CN115035000A (en) * | 2022-08-10 | 2022-09-09 | 山东国晟环境科技有限公司 | Road raise dust image identification method and system |
| CN115147415A (en) * | 2022-09-02 | 2022-10-04 | 山东微山湖酒业有限公司 | Wine box defect detection method based on image processing |
| CN116703785A (en) * | 2023-08-04 | 2023-09-05 | 普密特(成都)医疗科技有限公司 | Method for processing blurred image under minimally invasive surgery mirror |
| CN116703785B (en) * | 2023-08-04 | 2023-10-27 | 普密特(成都)医疗科技有限公司 | Method for processing blurred image under minimally invasive surgery mirror |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Lee et al. | Iterative filter adaptive network for single image defocus deblurring | |
| CN112381742A (en) | Single image motion blur removing method and system | |
| Liu et al. | Fast burst images denoising | |
| Shi et al. | Just noticeable defocus blur detection and estimation | |
| CN109685045B (en) | Moving target video tracking method and system | |
| CN104103050B (en) | A kind of real video restored method based on local policy | |
| CN103426182A (en) | Electronic image stabilization method based on visual attention mechanism | |
| US9143687B2 (en) | Method of analyzing motion blur using double discrete wavelet transform | |
| CN110097509B (en) | Restoration method of local motion blurred image | |
| Shen et al. | A fast algorithm for rain detection and removal from videos | |
| CN103841298B (en) | Video image stabilization method based on color constant and geometry invariant features | |
| WO2014070273A1 (en) | Recursive conditional means image denoising | |
| Zachevsky et al. | Statistics of natural stochastic textures and their application in image denoising | |
| Fan et al. | Multiscale cross-connected dehazing network with scene depth fusion | |
| Unger et al. | A convex approach for variational super-resolution | |
| US9008453B2 (en) | Blur-kernel estimation from spectral irregularities | |
| Yu et al. | Split-attention multiframe alignment network for image restoration | |
| Cristóbal et al. | Superresolution imaging: a survey of current techniques | |
| Dutta et al. | Weighted low rank approximation for background estimation problems | |
| Kamenetsky et al. | Interactive atmospheric turbulence mitigation | |
| CN113592801A (en) | Method and device for detecting stripe interference of video image | |
| Nasonov et al. | Image sharpening with blur map estimation using convolutional neural network | |
| Yue et al. | Deblur a blurred RGB image with a sharp NIR image through local linear mapping | |
| Nakamura et al. | Noise-level estimation from single color image using correlations between textures in RGB channels | |
| Nasonov et al. | Non-linear multi-frame image denoising using weighted nuclear norm minimization |
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 | ||
| WW01 | Invention patent application withdrawn after publication |
Application publication date: 20210219 |
|
| WW01 | Invention patent application withdrawn after publication |