CN106991650A - A kind of method and apparatus of image deblurring - Google Patents

Abstract

The present application discloses an image deblurring method and device, comprising: acquiring a blurred picture and a DVS event set within an exposure time; using the blurred picture as a current blurred picture; and performing deblurring processing on the blurred picture according to the DVS event set. By applying the present application, the effect of image deblurring can be improved.

Description

Method and device for image deblurring

technical field

The present application relates to image processing technology, and in particular to an image deblurring method and device.

Background technique

When the camera is taking pictures, sometimes due to various reasons such as hand shaking, the picture may be blurred. In order to remove or reduce image blur caused by camera motion, deblurring processing is usually performed on the imaged blurred image to obtain a clearer image.

At present, the existing deblurring methods can achieve deblurring effects only for a small number of images. The reason for this is that the existing technology uses deconvolution to achieve deblurring, and the process of using deconvolution to achieve deblurring itself is a pathological process, because the clear image and the cause of the blurring of the clear image are unknown in advance. of. It is precisely because the reason for the blurring of the image is unknown, so many methods will make some assumptions about the clear image to be solved. However, when the actual image does not match the assumption, the deblurring method will fail.

Contents of the invention

The present application provides a DVS-based image deblurring method and device, which can improve the effect of image deblurring.

In order to achieve the above object, the application adopts the following technical solutions:

An image deblurring method, comprising:

Obtain the blurred picture to be processed and the DVS event set recorded by the DVS sensor within the exposure time of the blurred picture;

Perform deblurring processing on the blurred picture according to the DVS event set.

Preferably, according to the DVS event set, performing deblurring processing on the blurred picture includes:

Estimate a DVS edge estimation image according to the DVS event set, and perform the deblurring process according to the DVS edge estimation image.

Preferably, the manner of performing deblurring processing on the DVS edge estimation image includes one of the following:

Align the blurred picture with the DVS edge estimation image, and perform deblurring processing according to the alignment result;

Align the blurred picture with the DVS edge estimation image, and perform deblurring processing according to the alignment result. When the deblurring processing result does not reach the preset deblurring processing effect and does not reach the set maximum number of iterations, according to The deblurring processing result determines an average edge map, and uses the average edge map to re-align the blurred picture and the DVS edge estimation image until the deblurring processing result reaches a preset deblurring processing effect or reaches a preset The specified maximum number of iterations, the fuzzy processing result is output as the defuzzification result;

Aligning the blurred picture with the DVS edge estimation image, performing deblurring processing according to the alignment result, and determining an average edge map according to the deblurring processing result when the current maximum number of iterations has not been reached, and Re-aligning the blurred picture and the DVS edge estimation image according to the average edge map until a set maximum number of iterations is reached, and outputting the blurred processing result as a deblurring result;

Align the blurred picture with the DVS edge estimation image, and perform deblurring processing according to the alignment result. When the deblurring processing result does not reach the preset deblurring processing effect and currently does not reach the set maximum number of iterations, An average edge map is determined according to the deblurring processing result, and the blurred picture and the DVS edge estimation image are realigned according to the average edge map until the deblurring processing result reaches a preset deblurring processing effect or reaches The maximum number of iterations is set, and the blurred processing result is output as the deblurring result.

Preferably, according to the DVS event set, performing deblurring processing on the blurred picture includes:

Estimating the camera motion trajectory and the DVS edge estimation image within the exposure time according to the DVS event set; and performing the deblurring process according to the camera motion trajectory and the DVS edge estimation image.

Preferably, the manner of performing deblurring processing according to the camera motion trajectory and the DVS edge estimation image includes one of the following:

Align the blurred picture with the DVS edge estimation image, and perform deblurring processing according to the alignment result, and when the deblurring processing result does not reach the preset deblurring processing effect, determine according to the deblurring processing result an average edge map in the direction of the camera motion track, and re-align the blurred picture and the DVS edge estimation image according to the average edge map until the deblurring processing result reaches the preset deblurring processing effect, and the The blurred processing result is output as the deblurred result;

Align the blurred picture with the DVS edge estimation image, and perform deblurring processing according to the alignment result. an average edge map in the track direction, and re-align the blurred picture and the DVS edge estimation image according to the average edge map until a set maximum number of iterations is reached, and output the blurred processing result as a deblurring result;

Align the blurred picture with the DVS edge estimation image, and perform deblurring processing according to the alignment result. When the deblurring processing result does not reach the preset deblurring processing effect and currently does not reach the set maximum number of iterations, Determine the average edge map in the direction of the camera motion trajectory according to the deblurring processing result, and realign the blurred picture and the DVS edge estimation image according to the average edge map until the deblurring processing result reaches The preset deblurring effect or the set maximum number of iterations are reached, and the blurring result is output as the deblurring result.

Preferably, the performing deblurring processing according to the alignment result includes: performing deblurring processing on the currently aligned blurred picture according to the currently aligned DVS edge estimation image using deconvolution transformation to obtain this iteration The clear picture of is used as the result of the deblurring process;

If the error between the clear picture of this iteration and the previous iteration is smaller than a predetermined threshold, it is determined that the preset deblurring effect is achieved; otherwise, it is determined that the preset deblurring effect is not achieved.

Preferably, the estimating the camera motion trajectory according to the DVS event set includes: dividing the exposure time into N time segments in chronological order, and dividing the DVS events in the same time segment in the DVS event set independently The imaging is taken as a picture, and the camera movement trajectory within the exposure time is determined according to the pictures imaged in N time segments.

Preferably, the method of determining the camera movement trajectory includes: in the pictures imaged in every two consecutive time segments, according to the positional relationship of the DVS event, determining the camera displacement in the two consecutive time segments, and calculating the All camera displacements in every two consecutive time segments within the exposure time are connected in chronological order to form a camera motion trajectory within the exposure time.

Preferably, the estimating the DVS edge estimation image according to the DVS event set includes: dividing the exposure time into N time segments in chronological order, and dividing the DVS events in the same time segment in the DVS event set Unify imaging into pictures, spatially align and superimpose the pictures imaged in all time segments, and calculate the skeleton diagram of the superimposed pictures as the DVS edge estimation image.

Preferably, the method of spatially aligning pictures of any two time segments includes: for pictures A and B of any two time segments, calculating Among them, (x, y) is the two-dimensional coordinates of the pixels in the picture, A(x, y) and B(x, y) represent the values of the pixels in pictures A and B respectively, and (Δx, Δy) is The two-dimensional displacement value required by picture B when picture B is aligned to picture A, argmin( ) indicates the value of the independent variable when ( ) takes the minimum value; according to the calculated (Δx, Δy), picture B and / Or picture A is displaced to achieve the alignment of picture A and picture B.

Preferably, said spatially aligning and superimposing the pictures imaged in all time slices includes one of the following:

Calculate (Δx, Δy) for every two consecutive time segments of pictures; in chronological order, sequentially align and superimpose the pictures at the back to the pictures at the front;

Calculate (Δx, Δy) for every two consecutive time segments of pictures; in chronological order, sequentially align and superimpose the picture in front of the time to the picture in time behind.

Preferably, the way to calculate the blur kernel when performing the deblurring process is: blur kernel in, Indicates that the matrix x is represented as a vector form, Indicates the calculation of the gradient, I is the clear image after deblurring processing, C is the blurred image after the current alignment, E is the DVS edge estimation image after the current alignment, ||vec|| ₂ indicates the 2 norm of the calculation vector, ||vec|| ₁ means to calculate the 1-norm of the vector, λ ₁ and λ ₂ are two preset weight values, and argmin( ) means the value of the independent variable when ( ) takes the minimum value.

Preferably, the method of determining the average edge map includes:

For the clear picture of this iteration, calculate the edge map in each segment direction in the camera motion trajectory, and calculate a skeleton map after superimposing all the edge maps, which is used as the average edge map.

An image deblurring device, comprising: an acquisition module and an image deblurring module;

The acquisition module is used to acquire the blurred picture to be processed and the DVS event set recorded by the DVS sensor within the exposure time of the blurred picture;

The image deblurring module is configured to perform deblurring processing on the blurred picture according to the DVS event set.

It can be seen from the above technical solution that in this application, the DVS event recorded by the DVS sensor is introduced to perform deblurring processing on the blurred picture. Specifically, the camera motion trajectory can be estimated, and the motion trajectory can be used to estimate the edge of the blurred image; at the same time, the DVS event can be used to perform DVS edge estimation, and the blurred image or the estimated edge estimation image of the blurred image can be compared with the DVS edge estimation result After alignment, participate in the deblurring process. Through the above processing, the effect of deblurring can be improved.

Description of drawings

FIG. 1 is a schematic diagram of the basic flow of the image deblurring method in the present application;

Fig. 2 is a schematic diagram of the relationship between the imaging of the DVS event set and the images imaged in each time segment;

Fig. 3 is a schematic diagram of camera displacement in two consecutive time segments;

Fig. 4 is a schematic diagram of performing DVS edge estimation;

Fig. 5 is the processing schematic diagram of RGB-DVS image alignment;

Fig. 6 is a schematic diagram of calculating a clear picture;

Fig. 7 is a schematic diagram of generating an average edge map;

FIG. 8 is a schematic diagram of the basic structure of an image deblurring device in the present application;

9 is a schematic structural diagram of the image deblurring module in the image deblurring device of the present application;

FIG. 10 is a schematic diagram of the comparison between the image deblurring method in this application and the prior art.

detailed description

In order to make the purpose, technical means and advantages of the present application clearer, the present application will be further described in detail below in conjunction with the accompanying drawings.

The DVS sensor is an ultra-high-speed camera with microsecond-level time resolution, which can record and output DVS events. The DVS time is used to represent the changes that occur between microsecond-level time units in the camera lens. Since the part that changes in the lens is usually the edge of the image, when the DVS sensor is fixed with the ordinary camera, a relatively clear edge image under the fast motion of the ordinary camera can be obtained. In this application, the relatively clear edge image obtained by the DVS sensor is used to participate in the deblurring process of the blurred image captured by the ordinary camera, thereby greatly improving the deblurring effect.

The most basic image deblurring methods in this application include:

Step 1. Obtain the blurred picture to be processed and the set of DVS events recorded by the DVS sensor within the exposure time of the blurred picture.

Step 2. Perform deblurring processing on the blurred picture according to the DVS event set.

In step 2, preferably, the DVS edge estimation image can be estimated according to the DVS event set, and then the deblurring process is performed according to the DVS edge estimation image. Alternatively, the camera motion trajectory and the DVS edge estimation image may also be estimated according to the DVS event set, and then the deblurring process is performed according to the camera motion trajectory and the DVS edge estimation image.

FIG. 1 is a schematic flow chart of an image deblurring method in this application. Wherein, the processing of an RGB image is taken as an example for description. As shown in Figure 1, the method includes:

Step 101, acquire the blurred picture to be processed and the set of DVS events recorded by the DVS sensor within the exposure time of the blurred picture.

Obtain the RGB fuzzy picture taken by the ordinary camera, the DVS sensor is fixed with the ordinary camera that takes the picture, so the record of the same viewing angle as the ordinary camera can be obtained, and the recorded DVS events can reflect the movement of the ordinary camera that takes the picture. Obtain the DVS event set within the exposure time of the blurred picture, which can reflect the camera motion condition within the corresponding exposure time.

Step 102, according to the DVS event set, estimate the camera motion track and the DVS edge estimation image within the exposure time.

The exposure time is divided into N time segments in chronological order, and N is a preset positive integer, which can be set according to needs, and the application does not limit the specific setting basis. The DVS event set within the exposure time includes many DVS events, and the corresponding DVS events have time marks. According to the divided time segments, according to the time marks of DVS events, the DVS events in the same time segment are independently imaged. The specific independent imaging method Any implementable manner may be adopted, which is not limited in the present application. As shown in FIG. 2 , it is the relationship between the imaging of the DVS event set and the images imaged in each time segment.

Firstly, the generation of blurred pictures and the effect of pictures of DVS event imaging are analyzed. If the subject within the exposure time of the camera is regarded as stationary, then the blurring of the image is caused by the movement of the camera itself relative to the stationary subject. Based on this, the images imaged in each time segment are relatively clear edge images under the rapid relative motion of the camera in the corresponding time range; for the pictures in different time segments, the objects to be photographed are the same, that is to say, the edge images The contours of should also be the same, except that the edge image has a relative displacement in the background, and this displacement just represents the relative motion trajectory of the camera between different time segments. At the same time, the images imaged in different time segments represent the same edge images at different time points, but due to the influence of noise, the edges of each picture may not be the most accurate and complete; if all N The pictures in the time segments are restored to the same time point, and then N pictures of the same time point are superimposed together to obtain a relatively clearer edge map.

Simply put, for pictures in two different time segments, there should be the same shape on the picture, and the change of the positional relationship of the shape in the whole image represents the linear motion relationship of the camera between the two time segments. One of the pictures is reversely moved according to the motion relationship, and the picture in the time period of the other picture can be obtained, and the edge map with enhanced effect can be obtained by superimposing these two pictures.

Based on the content of the above analysis, in this application, according to the images imaged in N time segments, the camera motion track and DVS edge estimation image within the exposure time are determined.

Specifically, the method of determining the camera motion trajectory within the exposure time can be as follows: in the pictures imaged in every two consecutive time segments, according to the positional relationship of the DVS event, determine the camera displacement in the two consecutive time segments, and All the camera displacements in every two consecutive time segments within the exposure time are connected in chronological order to form a camera motion trajectory within the exposure time. Among them, the camera displacement in two consecutive time segments is shown in Figure 3.

Here, an exemplary method for determining the camera displacement in two consecutive time segments is given: For pictures A and B of any two consecutive time segments, calculate

Among them, (x, y) is the two-dimensional coordinates of the pixels in the picture, A(x, y) and B(x, y) represent the values of the pixels in pictures A and B respectively, and (Δx, Δy) is The two-dimensional displacement of the camera from the time point represented by picture B to the time point represented by picture A, according to the two-dimensional displacement, the motion vector shown in Figure 3 can be obtained; here, it is assumed that the camera makes a straight line in two consecutive time segments sports. By connecting all motion vectors in two consecutive time periods in chronological order, the camera trajectory within the exposure time can be obtained.

The method of determining the DVS edge estimation image may be as follows: spatially align and superimpose images imaged in all time segments, and calculate a skeleton image of the superimposed images as the DVS edge estimation image. As shown in Figure 4.

Among them, when performing spatial alignment, the pictures of all time segments can be aligned to the same time segment, and then superimposed uniformly; or, the pictures of a certain time segment can be aligned to another time segment, and the two are superimposed , and then align the superimposed results to other time segments, and so on. Here, the above formula (1) can be used to calculate the two-dimensional displacement value (Δx, Δy) between the pictures of any two time segments, and the two-dimensional displacement value represents the two-dimensional displacement value required for aligning picture B to picture A. You can align the displacement (Δx, Δy) of picture B to picture A, or you can align the displacement (-Δx, -Δy) of picture A to picture B, or you can also shift and align picture A and picture B together to other time slices between picture A and picture B.

In addition, since the movement between two time segments is assumed to be a linear motion when formula (1) is used to calculate the displacement, in fact, as the time interval increases, the deviation between the linear motion and the actual motion may be greater, and the estimation is less accurate. Accurate, therefore, preferably, (Δx, Δy) can be calculated for the pictures of every two consecutive time segments when performing picture alignment and superimposition, and in chronological order, sequentially move the time-behind/front-time pictures to the time-before/ The final images are aligned and superimposed. Such processing can increase the accuracy of the estimation. Where "/" means or.

After this step, iterations begin to achieve image deblurring.

Step 103: Align the obtained blurred picture with the DVS edge estimation image to obtain the current aligned blurred picture and the current aligned DVS edge estimation image.

The processing of RGB-DVS image alignment is shown in Figure 5. Firstly, the RGB image and the DVS edge estimation image are dewarped separately. Then find the corresponding key points in the two images, and finally align the two images by calculating the affine transformation model. Specifically, the processes of finding out the distortion, determining the corresponding key points, and aligning the images through the model are all existing technologies, and will not be repeated here. Wherein, it is preferable to include de-distortion processing in this step, and if due to implementation complexity or other considerations, de-distortion may not be performed, and key points may be directly determined for image alignment. When performing alignment, the RGB image can be aligned to the DVS edge estimation image, and the DVS edge estimation image can also be aligned to the RGB image. Preferably, the RGB image can be aligned to the DVS edge estimation image, and the alignment effect is better.

In addition, in the first iteration, the RGB blurred image and the DVS edge estimation image are directly used for alignment processing, and in the subsequent iterations, the RGB blurred image is aligned with the DVS edge estimation image according to the RGB edge estimation image determined in the previous iteration. processing to improve the registration accuracy of RGB and DVS.

Step 104, according to the currently aligned DVS edge estimation image, use deconvolution transformation to perform deblurring processing on the currently aligned blurred picture to obtain a clear picture of this iteration.

The processing of the image deblurring module is to estimate the blur kernel of the image, and use the deconvolution transformation to achieve deblurring. As shown in Figure 6. Wherein, the currently aligned DVS edge is used to estimate the image E when calculating the blur kernel. Specifically, the deblurring process includes:

Calculate the blur kernel k such that

in Indicates that the matrix x is represented as a vector form, Indicates the calculation of the gradient, I is the clear RGB image after processing, C is the blurred image after the current alignment, E is the DVS edge estimation image after the current alignment, ||vec|| ₂ indicates the 2 norm of the calculation vector, | |vec|| ₁ means to calculate the 1 norm of the vector, λ ₁ and λ ₂ are the two preset weight values, argmin( ) means the value of the independent variable when ( ) takes the minimum value;

After obtaining the blur kernel k, obtain a clear RGB image I,

Indicates that the matrix x is expressed in vector form, ||vec|| ₂ indicates the calculation of the 2-norm of the vector, ||vec|| ₁ indicates the calculation of the 1-norm, and B' is the blurred image after the current alignment.

The above process of obtaining the blur kernel and obtaining the clear image is carried out alternately until the set number of alternations is reached, or the clear image does not change, and the calculated I is used as the clear image obtained by this iteration.

In the processing of this step, the method of calculating I is the same as that of the prior art, and the blur kernel k is calculated according to the currently aligned DVS edge estimation image E. In this way, by introducing the image E into the calculation of the blur kernel k, the blurred state of the image can be reflected more completely, and the calculated clear picture I is closer to the original picture.

So far, this process can be ended. The image was deblurred based on DVS edge estimation. If the processing capability allows, preferably, the following steps may be continued to optimize the deblurred processing result through an iterative processing process.

Step 105, judge whether the clear picture of this iteration is the same as that of the previous iteration, if yes, use the clear picture as the deblurring result, and end the process; otherwise, execute step 106.

When the clear image of the previous two iterations remains unchanged, it is considered that the best deblurring effect has been achieved, and the clear image is output, and the deblurring process ends; otherwise, continue to calculate the RGB edge image for the next iteration. Among them, the judgment method of whether the clear pictures of the two iterations are the same can be as follows: when the error between the clear pictures of this iteration and the previous iteration is less than a predetermined threshold, the clear pictures of the two iterations can be considered to be the same; otherwise, the clear pictures of the two iterations can be considered as Clear pictures are not the same. Of course, other determination methods may also be used, which is not limited in the present application.

Step 106, determine the average edge map according to the clear picture in this iteration, return to step 103, and use the average edge map as the basis for aligning the RGB fuzzy picture and the DVS edge estimation image in the next iteration.

The processing in this step is used to generate a directed edge map of the RGB image, which is called an average edge map in this application. Specifically, based on the clear picture estimated in this iteration, the edge map of the clear picture in the set direction is obtained, and then the average edge map is obtained by averaging the multiple edge maps. Wherein, the set direction may be several randomly designated directions, or, preferably, may be the direction of the camera motion track.

Then, this average edge map is used as the basis for aligning the RGB image with the DVS edge estimation image in the next iteration. Specifically, the average edge map can be compared with the DVS edge estimation image in the next iteration to determine the alignment displacement, and then the RGB blurred image and/or the DVS edge estimation image can be moved to align, thereby improving the matching of RGB and DVS images. quasi-precision. Wherein, the process of aligning the RGB image with the DVS edge estimation image according to the average edge map can be implemented in an existing manner, which is not limited in this application.

When the direction is set as the direction of the camera motion trajectory, the average edge map can be generated specifically as follows: segment the camera motion trajectory within the exposure time, calculate the edge map of the clear picture of this iteration in each segment direction, and divide After all the obtained edge maps are superimposed, the skeleton map is calculated as the average edge map.

Among them, in the simplest way, when segmenting the camera motion trajectory, it can be directly performed according to each segment when constructing the motion trajectory in step 102 . The method for generating the average edge map can be shown in FIG. 7 .

The above is a specific implementation of the image deblurring method in this application. In the above iterative processing flow, when the clear images of the two iterations are the same, the iterative process ends and the final deblurred processing result is obtained. Of course, it is also possible to set the maximum number of iterations more simply, iterate repeatedly before reaching the maximum number of iterations, and end the iterative process after reaching the maximum number of iterations to obtain the final processing result of deblurring. Alternatively, it is also possible to combine the clear pictures of the two iterations before and after the above comparison with the setting of the maximum number of iterations, and when any of the conditions is met, the iterative process ends, otherwise, repeated iterations are performed. Wherein, the maximum number of iterations may be set according to actual needs and device processing capabilities, which is not limited in this application.

In the above-mentioned deblurring process, the DVS event recorded by the DVS sensor can be introduced to estimate the camera motion trajectory, and use the motion trajectory to estimate the edge of the blurred image; at the same time, use the DVS event to perform DVS edge estimation, and the blurred image or estimate to get After the edge estimation image of the blurred image is aligned with the DVS edge estimation result, it participates in the deblurring process. Through the above processing, it is possible to use the camera motion trajectory and the DVS edge estimation image to provide a motion hypothesis closer to reality, thereby improving the effect of deblurring.

The present application also provides an image deblurring device, which can be used to implement the above-mentioned deblurring method shown in FIG. 1 . FIG. 8 is a schematic diagram of the basic structure of an image deblurring device in the present application. As shown in FIG. 8 , the device includes: an acquisition module and an image deblurring module.

Wherein, the obtaining module is used to obtain the blurred picture to be processed and the DVS event set recorded by the DVS sensor within the exposure time of the blurred picture. The image deblurring module is configured to perform deblurring processing on the blurred picture according to the DVS event set.

More preferably, the image deblurring module may further include a trajectory estimation submodule, an image registration submodule, an image deblurring submodule and an average edge map generation submodule. As shown in Figure 9.

Wherein, the track estimation sub-module is used for estimating the camera motion track and the DVS edge estimation image within the exposure time according to the DVS event set. It is specifically used to implement the processing of steps 101-102.

The image registration sub-module is used for aligning the fuzzy picture and the DVS edge estimation image to obtain the currently aligned fuzzy picture and the current aligned DVS edge estimation image. It is specifically used to implement the processing of step 103.

The image deblurring sub-module is used to estimate the image according to the DVS edge after the current alignment, and use the deconvolution transformation to perform deblurring processing on the blurred picture after the current alignment to obtain the clear picture of the current time. If this and the previous iteration If the error of the clear picture is smaller than the predetermined threshold, then output the clear picture as the deblurring result; otherwise, output the current clear picture to the average edge map output module. It is specifically used to implement the processing of steps 104-105.

The average edge map output module is used to determine the average edge map according to the clear picture input by the image deblurring module. And the average edge map is input to the image registration module, which is used as the basis for aligning the blurred picture with the DVS edge estimation image next time. Preferably, the average edge map in the direction of the camera motion track is determined, specifically for implementing the processing of step 106 .

Deblurring experiments were carried out using the deblurring method of the present application and the existing deblurring method, and the results were compared. FIG. 10 is a schematic diagram of the comparison between the present application and the existing deblurring method. It can be seen from Fig. 10 that the deblurring effect obtained by the present application is better.

The above descriptions are only preferred embodiments of the present invention, and are not intended to limit the present invention. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principles of the present invention shall be included in the present invention. within the scope of protection.

Claims (14)

1. An image deblurring method, characterized in that, comprising: Obtain the blurred picture to be processed and the DVS event set recorded by the DVS sensor within the exposure time of the blurred picture; Perform deblurring processing on the blurred picture according to the DVS event set. 2. The method according to claim 1, wherein, according to the DVS event set, performing deblurring processing on the blurred picture includes: Estimate a DVS edge estimation image according to the DVS event set, and perform the deblurring process according to the DVS edge estimation image. 3. The method according to claim 2, wherein the manner of performing deblurring processing according to the DVS edge estimation image comprises one of the following: Align the blurred picture with the DVS edge estimation image, and perform deblurring processing according to the alignment result; Align the blurred picture with the DVS edge estimation image, and perform deblurring processing according to the alignment result. When the deblurring processing result does not reach the preset deblurring processing effect and does not reach the set maximum number of iterations, according to The deblurring processing result determines an average edge map, and uses the average edge map to re-align the blurred picture and the DVS edge estimation image until the deblurring processing result reaches a preset deblurring processing effect or reaches a preset The specified maximum number of iterations, the fuzzy processing result is output as the defuzzification result; Aligning the blurred picture with the DVS edge estimation image, performing deblurring processing according to the alignment result, and determining an average edge map according to the deblurring processing result when the current maximum number of iterations has not been reached, and Re-aligning the blurred picture and the DVS edge estimation image according to the average edge map until a set maximum number of iterations is reached, and outputting the blurred processing result as a deblurring result; Align the blurred picture with the DVS edge estimation image, and perform deblurring processing according to the alignment result. When the deblurring processing result does not reach the preset deblurring processing effect and currently does not reach the set maximum number of iterations, An average edge map is determined according to the deblurring processing result, and the blurred picture and the DVS edge estimation image are realigned according to the average edge map until the deblurring processing result reaches a preset deblurring processing effect or reaches The maximum number of iterations is set, and the blurred processing result is output as the deblurring result. 4. The method according to claim 1, wherein, according to the DVS event set, performing deblurring processing on the blurred picture includes: Estimating the camera motion trajectory and the DVS edge estimation image within the exposure time according to the DVS event set; and performing the deblurring process according to the camera motion trajectory and the DVS edge estimation image. 5. method according to claim 4, is characterized in that, the mode that carries out deblurring processing according to described camera motion locus and DVS edge estimation image comprises one of following: Align the blurred picture with the DVS edge estimation image, and perform deblurring processing according to the alignment result, and when the deblurring processing result does not reach the preset deblurring processing effect, determine according to the deblurring processing result an average edge map in the direction of the camera motion track, and re-align the blurred picture and the DVS edge estimation image according to the average edge map until the deblurring processing result reaches the preset deblurring processing effect, and the The blurred processing result is output as the deblurring result; Align the blurred picture with the DVS edge estimation image, and perform deblurring processing according to the alignment result. an average edge map in the track direction, and re-align the blurred picture and the DVS edge estimation image according to the average edge map until a set maximum number of iterations is reached, and output the blurred processing result as a deblurring result; Align the blurred picture with the DVS edge estimation image, and perform deblurring processing according to the alignment result. When the deblurring processing result does not reach the preset deblurring processing effect and currently does not reach the set maximum number of iterations, Determine the average edge map in the direction of the camera motion trajectory according to the deblurring processing result, and realign the blurred picture and the DVS edge estimation image according to the average edge map until the deblurring processing result reaches The preset deblurring effect or the set maximum number of iterations are reached, and the blurring result is output as the deblurring result. 6. The method according to claim 3 or 5, wherein said performing deblurring processing according to the alignment result comprises: according to the DVS edge estimation image after the current alignment, using deconvolution transformation to perform deblurring on the currently aligned The final blurred image is deblurred, and the clear image of this iteration is obtained as the result of deblurred processing; If the error between the clear picture of this iteration and the previous iteration is smaller than a predetermined threshold, it is determined that the preset deblurring effect is achieved; otherwise, it is determined that the preset deblurring effect is not achieved. 7. The method according to claim 4, wherein the estimating the camera trajectory according to the DVS event set comprises: dividing the exposure time into N time segments according to chronological order, dividing the DVS event The DVS events in the same time segment in the set are independently imaged as pictures, and the camera movement track within the exposure time is determined according to the pictures imaged in N time segments. 8. The method according to claim 7, wherein the method of determining the camera motion track comprises: in the pictures imaged in every two consecutive time segments, according to the positional relationship of the DVS event, determine the two The camera displacement in the continuous time segments, and all the camera displacements in every two continuous time segments within the exposure time are connected in chronological order to form a camera motion trajectory within the exposure time. 9. The method according to claim 2 or 4, wherein the estimating the DVS edge estimation image according to the DVS event set comprises: dividing the exposure time into N time segments in chronological order, dividing the The DVS events in the same time segment in the DVS event set are uniformly imaged as pictures, the images imaged in all time segments are spatially aligned and superimposed, and the skeleton diagram of the superimposed pictures is calculated as the DVS edge estimation image. 10. The method according to claim 9, wherein the method of spatially aligning pictures of any two time segments comprises: for pictures A and B of any two time segments, calculating $\[\mathrm{\Δ}x,\mathrm{\Δ}\mathrm{the\; y}\]=\arg mi\mathrm{no}\underset{x,\mathrm{the\; y}}{\Σ}{\left(B\right(x+\mathrm{\Δ}x,\mathrm{the\; y}+\mathrm{\Δ}\mathrm{the\; y})-A(x,\mathrm{the\; y}\left)\right)}^2,$ Among them, (x, y) is the two-dimensional coordinates of the pixels in the picture, A(x, y) and B(x, y) represent the values of the pixels in pictures A and B respectively, and (Δx, Δy) is The two-dimensional displacement value required by picture B when picture B is aligned to picture A, argmin( ) indicates the value of the independent variable when ( ) takes the minimum value; according to the calculated (Δx, Δy), picture B and / Or picture A is displaced to achieve the alignment of picture A and picture B. 11. The method according to claim 9, wherein said spatially aligning and superimposing the pictures of all time slice imaging comprises one of the following: Calculate (Δx, Δy) for every two consecutive time segments of pictures; in chronological order, sequentially align and superimpose the pictures at the back to the pictures at the front; Calculate (Δx, Δy) for every two consecutive time segments of pictures; in chronological order, sequentially align and superimpose the picture in front of the time to the picture in time behind. 12. The method according to claim 6, wherein the method of calculating the blur kernel when performing the deblurring process is: blur kernel $\hat{k}=\arg mi\mathrm{no}\left|\right|\∂\hat{I}*\hat{k}-\∂\hat{C}|{|}_2+{\mathrm{\λ}}_1|\left|\hat{k}\right|{|}_1+{\mathrm{\λ}}_2\left|\right|\∂\hat{I}-\widehat{\mathrm{E.}}|{|}_2,$ in, Indicates that the matrix x is represented as a vector form, Indicates the calculation of the gradient, I is the clear image after deblurring processing, C is the blurred image after the current alignment, E is the DVS edge estimation image after the current alignment, ||vec|| ₂ indicates the 2 norm of the calculation vector, ||vec|| ₁ means to calculate the 1-norm of the vector, λ ₁ and λ ₂ are two preset weight values, and argmin( ) means the value of the independent variable when ( ) takes the minimum value. 13. The method according to claim 1, wherein the mode of determining the average edge map comprises: For the clear picture of this iteration, calculate the edge map in each segment direction in the camera motion trajectory, and calculate a skeleton map after superimposing all the edge maps, which is used as the average edge map. 14. An image deblurring device, comprising: an acquisition module and an image deblurring module; The acquisition module is used to acquire the blurred picture to be processed and the DVS event set recorded by the DVS sensor within the exposure time of the blurred picture; The image deblurring module is configured to perform deblurring processing on the blurred picture according to the DVS event set.