WO2017100971A1 - 一种失焦模糊图像的去模糊方法和装置 - Google Patents
一种失焦模糊图像的去模糊方法和装置 Download PDFInfo
- Publication number
- WO2017100971A1 WO2017100971A1 PCT/CN2015/097237 CN2015097237W WO2017100971A1 WO 2017100971 A1 WO2017100971 A1 WO 2017100971A1 CN 2015097237 W CN2015097237 W CN 2015097237W WO 2017100971 A1 WO2017100971 A1 WO 2017100971A1
- Authority
- WO
- WIPO (PCT)
- Prior art keywords
- blur
- image
- pixel
- amount
- map
- Prior art date
Links
- 238000000034 method Methods 0.000 title claims abstract description 85
- 238000012545 processing Methods 0.000 claims abstract description 38
- 230000002159 abnormal effect Effects 0.000 claims description 13
- 238000012805 post-processing Methods 0.000 claims description 11
- 238000013507 mapping Methods 0.000 claims description 7
- 238000004364 calculation method Methods 0.000 claims description 5
- 238000001914 filtration Methods 0.000 claims description 3
- 238000003709 image segmentation Methods 0.000 claims description 3
- 238000010586 diagram Methods 0.000 abstract description 8
- 230000000694 effects Effects 0.000 description 5
- 230000006870 function Effects 0.000 description 5
- 230000015556 catabolic process Effects 0.000 description 3
- 238000006731 degradation reaction Methods 0.000 description 3
- 239000011159 matrix material Substances 0.000 description 3
- 238000004458 analytical method Methods 0.000 description 1
- 230000002547 anomalous effect Effects 0.000 description 1
- 238000003708 edge detection Methods 0.000 description 1
- 238000003384 imaging method Methods 0.000 description 1
- 230000003287 optical effect Effects 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
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F17/00—Digital computing or data processing equipment or methods, specially adapted for specific functions
- G06F17/10—Complex mathematical operations
- G06F17/15—Correlation function computation including computation of convolution operations
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T5/00—Image enhancement or restoration
- G06T5/20—Image enhancement or restoration using local operators
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/13—Edge detection
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
- H04N1/00—Scanning, transmission or reproduction of documents or the like, e.g. facsimile transmission; Details thereof
- H04N1/40—Picture signal circuits
- H04N1/409—Edge or detail enhancement; Noise or error suppression
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
- H04N1/00—Scanning, transmission or reproduction of documents or the like, e.g. facsimile transmission; Details thereof
- H04N1/40—Picture signal circuits
- H04N1/409—Edge or detail enhancement; Noise or error suppression
- H04N1/4092—Edge or detail enhancement
-
- 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
- G06T2207/20004—Adaptive image processing
- G06T2207/20012—Locally adaptive
Definitions
- the present application relates to the field of image processing, and in particular, to a deblurring method and apparatus for an out-of-focus blurred image.
- I and L respectively represent an out-of-focus blur image and a clear image
- Is a convolutional symbol
- N random noise
- k a fuzzy kernel.
- the fuzzy kernel is often seen as a Gaussian model:
- (x, y) is the coordinate of the pixel in the image
- ⁇ is the standard deviation
- ⁇ can measure the degree of blurring of the image, also known as the amount of blur.
- the estimation of the fuzzy kernel is a key step in the deblurring process.
- deblurring the image in this case is relatively simple.
- the blur kernel corresponding to each pixel in the image is different. This situation is more complicated, and the difficulty of deblurring is relatively large.
- there are two main methods for fuzzy kernel estimation of variable spatial blurred images One is to divide the image into equal-sized rectangular regions according to the similarity of fuzzy kernels, and use space sparse constraint in each region.
- the fuzzy kernel is constrained to obtain the local fuzzy kernel k of the image; the other method is to treat the fuzzy kernel as a disk model or a Gaussian model, and estimate the standard deviation of the disk radius or Gaussian model corresponding to each pixel, and then obtain the blur. nuclear. Since the degree of blurring of the local area of the image is related to the depth, dividing the image into equal-sized rectangular regions is likely to cause an excessive difference in the actual blur degree between different pixels in the same rectangular region, so the first method may appear. Local fuzzy kernel estimation is inaccurate.
- the second method is for every pixel. A corresponding fuzzy kernel, but the deconvolution calculation for each pixel with different fuzzy kernels in the image restoration process is too expensive, so how to use these fuzzy kernels for image restoration is a key issue.
- the present application provides a deblurring method and apparatus for an out-of-focus image, which solves the problem of poor deblurring effect on the out-of-focus image in the prior art.
- the present application provides a deblurring method for an out-of-focus image, comprising:
- the original image is blurred by using a preset fuzzy kernel to obtain a re-blurred image
- Deblurring processing is performed according to the complete blur quantity map to obtain a deblurred image.
- the present application further provides a deblurring device for an out-of-focus image, comprising:
- the sparse fuzzy quantity estimation module is used for blurring the original image by using a preset fuzzy kernel to obtain a re-blurred image; and estimating the blur amount of the edge region pixel of the original image according to the change of the image edge information in the blur processing process. , obtaining a sparse fuzzy quantity map;
- a complete fuzzy quantity map mapping module is configured to estimate a blur quantity of a non-edge area pixel of the original image according to the sparse fuzzy quantity map to obtain a complete blur quantity map;
- the deblurring processing module is configured to perform deblurring processing according to the complete fuzzy quantity map to obtain a deblurred image.
- the deblurring method and device for the defocus blur image provided by the present application firstly adopts a preset fuzzy kernel to perform fuzzy processing on the input original image to obtain a re-blurred image; according to the change of the image edge information in the blur processing process, the edge region of the original image is obtained.
- the blur quantity of the pixel is estimated to obtain a sparse blur quantity map; the blur quantity of the non-edge area pixel of the original image is estimated according to the sparse fuzzy quantity map to obtain a complete blur quantity map; and the complete blur quantity map is used for deblurring processing, Get the deblurred image.
- the blur map since the blur map is obtained by using the change of the edge information after the image blur, the obtained blur map can be made more accurate, thereby improving the quality of the deblurred image.
- FIG. 1 is a schematic flow chart of a method for deblurring an out-of-focus blur image according to an embodiment of the present application
- FIG. 2 is a schematic flow chart of a method for deblurring an out-of-focus blur image according to another embodiment of the present application
- FIG. 3 is a schematic flow chart of a method for deblurring an out-of-focus blur image according to another embodiment of the present application.
- FIG. 4 is a schematic flowchart of post-processing in a deblurring method of an out-of-focus blur image according to an embodiment of the present application
- FIG. 5 is a schematic block diagram of a deblurring device for an out-of-focus blur image according to an embodiment of the present application
- FIG. 6 is a schematic diagram of a unit of a deblurring processing module in a deblurring device of an out-of-focus blur image according to an embodiment of the present application.
- the embodiment provides a deblurring method for an out-of-focus blur image, which includes the following steps:
- Step 1.1 Enter the original image.
- the original image is a depth-based single-space variable defocus blur image.
- Step 1.2 Blur the original image by using a preset fuzzy kernel to obtain a re-blurred image.
- the input original image is re-blurred with a Gaussian kernel (preset fuzzy kernel) having a standard deviation of ⁇ 0 (fuzzy amount).
- Gaussian kernel preset fuzzy kernel having a standard deviation of ⁇ 0 (fuzzy amount).
- the defocusing fuzzy kernel is regarded as a Gaussian model to describe the present application.
- the defocus blur fuzzy kernel is regarded as another model (for example, a circle)
- the disc model the deblurring method provided by the present application is also applicable.
- Step 1.3 According to step 1.2, the change of image edge information during the blur process is original The amount of blur of the pixels in the edge region of the image is estimated to obtain a sparse blur amount map.
- the edges of the original image can be expressed as:
- u(x, y) is a step function
- the edge sharpness is defined as:
- the gradient ⁇ I R of the blurred image can be expressed by the following convolution process:
- equation (4) can be written as:
- the fuzzy amount ⁇ can be obtained by the formula (7):
- the ⁇ corresponding to each pixel constitutes a sparse fuzzy quantity map.
- the sparse fuzzy amount map is corrected first.
- a Guided Filter method is used to correct the sparse fuzzy amount map. Using the edge image as the guide image, the corrected image r is:
- E is an edge image
- W is a kernel function
- w z is the window centered on the pixel z, ⁇ z , And
- w z window are the mean gray pixels, the variance of the total number of pixels within the window and E w z, ⁇ is a regularization parameter.
- the windows w z and ⁇ are set to 5x5 and 10 -9 , respectively .
- Step 1.4 Estimating the amount of blurring of the non-edge region pixels of the original image according to the sparse fuzzy amount map to obtain a complete blur amount map.
- the sparse fuzzy amount map only includes the amount of blurring of the pixels in the edge region, and therefore, it is necessary to obtain the amount of blurring including the pixels of the edge region and the pixels of the non-edge region.
- the K-most adjacent interpolation method (KNN matting interpolation) is used to estimate the blur amount of the non-edge region pixels of the original image according to the sparse blur amount map, and a complete blur amount map is obtained.
- the complete fuzzy quantity map can be obtained by the following equation:
- a vector representing a complete fuzzy quantity map and a sparse fuzzy quantity map respectively.
- ⁇ is the regularization parameter
- L is the Laplacian matrix of the sparse associative matrix A
- T is the transposed matrix.
- ⁇ can be set to 0.8. Be constraint. Solving equation (13) by preconditioned conjugate gradient (PCG)
- Step 1.5 Deblurring according to the obtained complete fuzzy quantity map to obtain a deblurred image.
- Step 1.5 may employ any one of the prior art deblurring methods, such as the two deblurring methods mentioned in the background of the present application.
- the defocusing reduces the sharpness and contrast of the edge
- the amount of blur at the high frequency will change relatively large. Therefore, the original image is blurred by the preset blur kernel, and the image edge information is changed. The amount of blur of the edge region pixels of the original image is estimated. In this way, a more accurate blur map can be obtained, which makes the image obtained after deblurring more clear.
- the degree of blurring of the image is related to the depth.
- the greater the depth the greater the degree of blur.
- the degree of blurring is also Has local consistency.
- the amount of blurring of each pixel is different, in order to ensure the local consistency of the fuzzy kernel and reduce the influence of the outliers, in this embodiment
- the super pixel-based method is used to divide the image into a plurality of super pixels, and the blur amount of the super pixels is obtained according to the amount of blur of all the pixels in each super pixel.
- this embodiment provides another deblurring method for the out-of-focus blur image, which is different from the first embodiment in that step 1.5 includes the following sub-steps:
- Sub-step 2.1 The super-pixel based method divides the complete blur map into multiple superpixels.
- Sub-step 2.2 Processing the amount of blur of all pixels in the current superpixel to obtain the amount of blur of the current superpixel.
- the average value of the blur amount ⁇ of all the pixels in each super pixel is used as the blur amount of the super pixel, that is:
- n is the label of the super pixel
- ⁇ n and m j are the blur amount of the nth super pixel and the blur amount of the pixel j, respectively.
- M n represents the n-th superpixel area
- t is the number of pixels in M n.
- the total number l of super pixels is a parameter set in advance.
- the amount of blur of the super pixel may also be determined in other manners according to actual conditions.
- Sub-step 2.3 Obtain the blur kernel of each super pixel according to the amount of blur of each super pixel.
- the fuzzy kernel of the nth superpixel is:
- Sub-step 2.4 Deblurring each superpixel according to the blur kernel of each superpixel.
- performing deblurring processing on each super pixel according to the fuzzy kernel of each super pixel includes: performing deconvolution calculation on each super pixel according to each fuzzy pixel of the super pixel.
- each super pixel is separately subjected to deblurring processing by using a three-dimensional block matching (BM3D)-based non-blind deconvolution method.
- BM3D three-dimensional block matching
- Sub-step 2.5 Combining each deblurred superpixel to obtain a deblurred image.
- the resulting deblurred image is:
- the super pixel segmentation method is adopted, and each super pixel is separately subjected to deblurring processing, and then all the super pixels are combined into a clear image.
- the spatially variable deblurring can be transformed into a local space invariant deblurring problem.
- the image segmentation method based on superpixel can better adapt to the depth change of the scene and the edge region of the image, which can distinguish the foreground and the background more accurately and conform to the depth consistency.
- the method provided by the embodiment can be more accurate based on the local depth consistency fuzzy image refinement. Fuzzy core.
- the deblurred image obtained by the method provided in the second embodiment described above may have ringing and noise, especially in the edge region of the image.
- the first is that the parameter l is set too small, resulting in a super-pixel area that is too large, including a region with a large depth change.
- the second is that the blur amount ⁇ at the smaller edge is smaller than the blur amount ⁇ n of the current super pixel.
- the present embodiment provides another deblurring method for the defocus blur image, which is different from Embodiments 1 and 2 in that after step 1.5, step 1.6 is also included.
- the blurred image is post-processed to obtain the final sharp image.
- the post-processing step includes the following sub-steps:
- Sub-step 3.1 Blurring the deblurred image according to the obtained fuzzy check of each super pixel to obtain a second blurred image.
- Sub-step 3.2 calculating the difference e(x, y) of the region corresponding to the original image of the second blurred image, specifically:
- Sub-step 3.3 When the obtained difference is greater than the preset threshold ⁇ , it is determined that the corresponding pixel is an abnormal pixel.
- Sub-step 3.4 using the minimum value of the blur amount of each pixel in the super pixel in which the abnormal pixel is located as the blur amount of the super pixel, and deblurring the second blurred image according to the blur amount to obtain a restored image.
- Sub-step 3.5 replacing the restored result of the abnormal pixel in the restored image with the corresponding pixel in the blurred image to obtain a final clear image.
- e(x, y) is greater than a preset threshold ⁇
- the pixel (x, y) is considered to be an abnormal pixel.
- ⁇ n is larger than the blur amount m j of the pixel j
- image restoration is performed by using the minimum value of the blur amount in each super pixel in m as the blur amount of the super pixel (that is, deblurring the second blurred image) to obtain a restored image.
- the anomalous pixels (ringing and noise pixels) in the final image are:
- various parameters used may be as shown in the following table. In other embodiments, these parameters may be selected using corresponding empirical values or based on actual needs.
- the deblurring method of the defocus blur image provided by the embodiment provides the quality of the clear image The amount is higher and can effectively remove ringing and noise.
- the embodiment provides a deblurring device for the defocus blur image, which includes an input module 101 and a sparse fuzzy amount map estimation module 102.
- the input module 101 is for inputting an original image.
- the sparse fuzzy map estimation module 102 is configured to perform blur processing on the original image by using a preset blur kernel to obtain a re-blurred image; and estimate the blur amount of the edge region pixel of the original image according to the change of the image edge information in the blur processing process, Get a sparse fuzzy map.
- the complete fuzzy quantity map mapping module 104 is configured to estimate the blur quantity of the non-edge area pixels of the original image according to the sparse fuzzy quantity map to obtain a complete blur quantity map.
- the deblurring processing module 105 is configured to perform deblurring processing according to the complete blur quantity map to obtain a deblurred image.
- the deblurring device of the defocus blur image further comprises a sparse blur map correction module 103 for correcting the sparse blur map.
- the complete fuzzy quantity map mapping module 104 is configured to estimate the blur quantity of the non-edge area pixels of the original image according to the corrected sparse blur quantity map to obtain a complete blur quantity map.
- the sparse fuzzy map correcting module 103 is configured to correct the sparse fuzzy amount map by using a guided filtering method.
- the complete fuzzy quantity map mapping module 104 is configured to estimate the blur quantity of the non-edge area pixels of the original image according to the sparse fuzzy quantity map by using an interpolation method to obtain a complete blur quantity map.
- the sparse fuzzy amount map estimation module 102 is configured to estimate the blur amount of the edge region pixel of the original image according to the change of the image edge information in the blur processing process, and obtain the blur amount of the edge region pixel when the sparse blur amount map is obtained.
- ⁇ is obtained by the following formula,
- the deblurring processing module includes an image dividing unit 201 and an image restoring unit 202.
- the image segmentation unit 201 is configured to segment the complete blur amount map into a plurality of super pixels based on a super pixel method.
- the image restoring unit 202 is configured to process the blur amount of all the pixels in the current super pixel to obtain the blur amount of the current super pixel; and obtain the blur kernel of each super pixel according to the blur amount of each super pixel;
- the super pixel fuzzy kernel performs deblurring processing on each super pixel, and synthesizes each deblurred super pixel to obtain a deblurred image.
- the image restoration unit 202 is configured to use an average value of the blur amounts of all the pixels in the current super pixel as the blur amount of the current super pixel.
- the image restoration unit 202 is configured to perform deconvolution calculation for each super pixel according to the blur kernel of each super pixel.
- the image restoration unit 202 is configured to perform deblurring processing on each super pixel by using a non-blind deconvolution method based on three-dimensional block matching according to a blur kernel of each super pixel.
- the deblurring device of the defocus blur image further comprises a post processing module 106 for post processing the deblurred image to obtain a final sharp image.
- the post-processing module 106 is configured to perform blur processing on the deblurred image according to the obtained fuzzy check of each super pixel to obtain a second blurred image; and calculate a difference between regions corresponding to the original image of the second blurred image; When the difference is greater than the preset threshold, determining that the corresponding pixel is an abnormal pixel; using a minimum value of the blur amount of each pixel in the super pixel in which the abnormal pixel is located as the blur amount of the super pixel, and using the blur amount to the second blurred image Deblurring is performed to obtain a restored image; the restored result of the abnormal pixel in the restored image is replaced with the corresponding pixel in the blurred image to obtain a final clear image.
- the deblurring device of the defocusing blur image provided in this embodiment corresponds to the deblurring method of the defocus blur image provided in the third embodiment.
- the third embodiment which is not described in detail in this embodiment.
- the deblurring device of the defocus blur image provided by the embodiment can obtain a more accurate fuzzy quantity map, and the fuzzy quantity map refinement based on the local depth consistency can obtain a more accurate fuzzy kernel, and can effectively remove the ringing. And noise to get a clearer picture.
- the program may be stored in a computer readable storage medium, and the storage medium may include: a read only memory. Random access memory, disk or optical disk, etc.
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Pure & Applied Mathematics (AREA)
- Mathematical Physics (AREA)
- Mathematical Analysis (AREA)
- Mathematical Optimization (AREA)
- Computational Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Multimedia (AREA)
- Signal Processing (AREA)
- Algebra (AREA)
- Computing Systems (AREA)
- Databases & Information Systems (AREA)
- Software Systems (AREA)
- General Engineering & Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Image Processing (AREA)
Abstract
Description
σ0 | ε τ | λ ζ |
3 | 10-6 8*10-4 | 0.8 10-9 |
Claims (22)
- 一种失焦模糊图像的去模糊方法,其特征在于,包括:输入原始图像;采用预设模糊核对原始图像进行模糊处理,得到再模糊图像;根据所述模糊处理过程中图像边缘信息的变化对原始图像的边缘区域像素的模糊量进行估计,得到稀疏模糊量图;根据稀疏模糊量图对原始图像的非边缘区域像素的模糊量进行估计,得到完整模糊量图;根据所述完整模糊量图进行去模糊处理,得到去模糊图像。
- 如权利要求1所述的方法,其特征在于,先对所述稀疏模糊量图进行校正,根据校正后的稀疏模糊量图对原始图像的非边缘区域像素的模糊量进行估计,得到完整模糊量图。
- 如权利要求2所述的方法,其特征在于,采用引导滤波方法对所述稀疏模糊量图进行校正。
- 如权利要求1所述的方法,其特征在于,采用k最邻近插值法根据稀疏模糊量图对原始图像的非边缘区域像素的模糊量进行估计,得到完整模糊量图。
- 如权利要求1-5任意一项所述的方法,其特征在于,根据所述完整模糊量图进行去模糊处理,得到去模糊图像,包括:基于超像素的方法将完整模糊量图分割成多个超像素,对当前超像素中所有像素的模糊量进行处理,得到当前超像素的模糊量;根据每个超像素的模糊量得到每个超像素的模糊核;根据每个超像素的模糊核对每个超像素进行去模糊处理,并将每个 去模糊处理后的超像素进行合成,得到去模糊图像。
- 如权利要求6所述的方法,其特征在于,对当前超像素中所有像素的模糊量进行处理,得到当前超像素的模糊量,包括:将当前超像素中所有像素的模糊量的平均值作为当前超像素的模糊量。
- 如权利要求6所述的方法,其特征在于,根据每个超像素的模糊核对每个超像素进行去模糊处理,包括:根据每个超像素的模糊核,分别对每个超像素进行反卷积计算。
- 如权利要求8所述的方法,其特征在于,根据每个超像素的模糊核,采用基于三维块匹配的非盲反卷积方法对每个超像素进行去模糊处理。
- 如权利要求6所述的方法,其特征在于,还包括对所述去模糊图像进行后处理,以得到最终的清晰图像。
- 如权利要求10所述的方法,其特征在于,对所述去模糊图像进行后处理,以得到最终的清晰图像,包括:根据得到的每个超像素的模糊核对所述去模糊图像进行模糊处理,得到第二模糊图像;计算第二模糊图像与原始图像对应的区域的差值;当所述差值大于预设阈值时,判断对应的像素为异常像素;将异常像素所在超像素中各个像素的模糊量的最小值作为该超像素的模糊量,并根据该模糊量对第二模糊图像进行去模糊处理,得到复原图像;将所述复原图像中异常像素的复原结果替换所述去模糊图像中对应的像素,以得到最终的清晰图像。
- 一种失焦模糊图像的去模糊装置,其特征在于,包括:输入模块,用于输入原始图像;稀疏模糊量图估计模块,用于采用预设模糊核对原始图像进行模糊处理,得到再模糊图像;并根据所述模糊处理过程中图像边缘信息的变化对原始图像的边缘区域像素的模糊量进行估计,得到稀疏模糊量图;完整模糊量图映射模块,用于根据稀疏模糊量图对原始图像的非边缘区域像素的模糊量进行估计,得到完整模糊量图;去模糊处理模块,用于根据所述完整模糊量图进行去模糊处理,得到去模糊图像。
- 如权利要求12所述的装置,其特征在于,还包括稀疏模糊量图校正模块,用于对所述稀疏模糊量图进行校正;完整模糊量图映射模块用于根据校正后的稀疏模糊量图对原始图像的非边缘区域像素的模糊量进行估计,得到完整模糊量图。
- 如权利要求13所述的装置,其特征在于,稀疏模糊量图校正模 块用于采用引导滤波方法对所述稀疏模糊量图进行校正。
- 如权利要求12所述的装置,其特征在于,完整模糊量图映射模块用于采用k最邻近插值法根据稀疏模糊量图对原始图像的非边缘区域像素的模糊量进行估计,得到完整模糊量图。
- 如权利要求12-16任意一项所述的装置,其特征在于,去模糊处理模块包括:图像分割单元,用于基于超像素的方法将完整模糊量图分割成多个超像素;图像复原单元,用于对当前超像素中所有像素的模糊量进行处理,得到当前超像素的模糊量;并根据每个超像素的模糊量得到每个超像素的模糊核;之后,根据每个超像素的模糊核对每个超像素进行去模糊处理,将每个去模糊处理后的超像素进行合成,得到去模糊图像。
- 如权利要求17所述的装置,其特征在于,图像复原单元用于将当前超像素中所有像素的模糊量的平均值作为当前超像素的模糊量。
- 如权利要求17所述的装置,其特征在于,图像复原单元用于根据每个超像素的模糊核,分别对每个超像素进行反卷积计算。
- 如权利要求19所述的装置,其特征在于,图像复原单元用于根据每个超像素的模糊核,采用基于三维块匹配的非盲反卷积方法对每个超像素进行去模糊处理。
- 如权利要求17所述的装置,其特征在于,还包括后处理模块,用于对所述去模糊图像进行后处理,以得到最终的清晰图像。
- 如权利要求18所述的装置,其特征在于,后处理模块用于:根据得到的每个超像素的模糊核对所述去模糊图像进行模糊处理,得到第二模糊图像;计算第二模糊图像与原始图像对应的区域的差值;当所述差值大于预设阈值时,判断对应的像素为异常像素;将异常像素所在超像素中各个像素的模糊量的最小值作为该超像素的模糊量,并根据该模糊量对第二模糊图像进行去模糊处理,得到复原图像;将所述复原图像中异常像素的复原结果替换所述去模糊图像中对应的像素,以得到最终的清晰图像。
Priority Applications (2)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
PCT/CN2015/097237 WO2017100971A1 (zh) | 2015-12-14 | 2015-12-14 | 一种失焦模糊图像的去模糊方法和装置 |
US15/753,971 US10628924B2 (en) | 2015-12-14 | 2015-12-14 | Method and device for deblurring out-of-focus blurred images |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
PCT/CN2015/097237 WO2017100971A1 (zh) | 2015-12-14 | 2015-12-14 | 一种失焦模糊图像的去模糊方法和装置 |
Publications (1)
Publication Number | Publication Date |
---|---|
WO2017100971A1 true WO2017100971A1 (zh) | 2017-06-22 |
Family
ID=59055581
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
PCT/CN2015/097237 WO2017100971A1 (zh) | 2015-12-14 | 2015-12-14 | 一种失焦模糊图像的去模糊方法和装置 |
Country Status (2)
Country | Link |
---|---|
US (1) | US10628924B2 (zh) |
WO (1) | WO2017100971A1 (zh) |
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110135576A (zh) * | 2018-02-02 | 2019-08-16 | 辉达公司 | 一种用于视频去模糊的无监督学习方法 |
CN112766190A (zh) * | 2021-01-25 | 2021-05-07 | 北京有竹居网络技术有限公司 | 对抗样本生成方法、装置、存储介质及电子设备 |
Families Citing this family (8)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JP6849430B2 (ja) * | 2016-12-27 | 2021-03-24 | キヤノン株式会社 | 画像処理装置、画像処理方法、及びプログラム |
CN110807745B (zh) * | 2019-10-25 | 2022-09-16 | 北京小米智能科技有限公司 | 图像处理方法及装置、电子设备 |
CN111311562B (zh) * | 2020-02-10 | 2023-10-10 | 浙江华创视讯科技有限公司 | 虚焦图像的模糊度检测方法及装置 |
JP7434032B2 (ja) * | 2020-03-31 | 2024-02-20 | キヤノン株式会社 | 情報処理装置、情報処理方法、およびプログラム |
JP2022142932A (ja) * | 2021-03-17 | 2022-10-03 | ラピステクノロジー株式会社 | ヘッドアップディスプレイ、映像補正方法及び映像補正プログラム |
CN114972333B (zh) * | 2022-07-19 | 2022-10-25 | 淄博市淄川区市政环卫服务中心 | 基于人工智能的道路裂缝检测方法及系统 |
CN115661187B (zh) * | 2022-12-12 | 2023-06-02 | 山东本草堂中药饮片有限公司 | 用于中药制剂分析的图像增强方法 |
CN116311262B (zh) * | 2023-05-23 | 2023-08-01 | 济南大陆机电股份有限公司 | 一种仪器仪表信息的识别方法、系统、设备和存储介质 |
Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US8249377B1 (en) * | 2008-06-08 | 2012-08-21 | Hewlett-Packard Development Company, L.P. | Blurred digital image deblurring |
CN103049906A (zh) * | 2012-12-07 | 2013-04-17 | 清华大学深圳研究生院 | 一种图像深度提取方法 |
CN103201765A (zh) * | 2010-09-28 | 2013-07-10 | 马普科技促进协会 | 用于从所观察到的数字图像序列恢复数字图像的方法与设备 |
CN103514582A (zh) * | 2012-06-27 | 2014-01-15 | 郑州大学 | 基于视觉显著的图像去模糊方法 |
US9208543B2 (en) * | 2013-10-23 | 2015-12-08 | Adobe Systems Incorporated | Deblurring images having spatially varying blur |
Family Cites Families (11)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US7561186B2 (en) * | 2004-04-19 | 2009-07-14 | Seiko Epson Corporation | Motion blur correction |
US7551772B2 (en) * | 2004-11-30 | 2009-06-23 | Hewlett-Packard Development Company, L.P. | Blur estimation in a digital image |
JP4799428B2 (ja) * | 2007-01-22 | 2011-10-26 | 株式会社東芝 | 画像処理装置及び方法 |
US8139886B2 (en) * | 2008-06-23 | 2012-03-20 | Microsoft Corporation | Blur estimation |
US8582910B2 (en) * | 2009-10-02 | 2013-11-12 | The Chinese University Of Hong Kong | Methods and apparatus for editing images |
US8885941B2 (en) * | 2011-09-16 | 2014-11-11 | Adobe Systems Incorporated | System and method for estimating spatially varying defocus blur in a digital image |
JP5911292B2 (ja) * | 2011-12-19 | 2016-04-27 | キヤノン株式会社 | 画像処理装置、撮像装置、画像処理方法、および、画像処理プログラム |
US9008453B2 (en) * | 2012-06-25 | 2015-04-14 | Yissum Research Development Company Of The Hebrew University Of Jerusalem Ltd. | Blur-kernel estimation from spectral irregularities |
AU2013206597A1 (en) * | 2013-06-28 | 2015-01-22 | Canon Kabushiki Kaisha | Depth constrained superpixel-based depth map refinement |
CN105469363B (zh) * | 2014-09-05 | 2018-05-18 | 华为技术有限公司 | 基于模糊图像生成清晰图像的方法和装置 |
JP2017010092A (ja) * | 2015-06-17 | 2017-01-12 | キヤノン株式会社 | 画像処理装置、撮像装置、画像処理方法、画像処理プログラム、および、記憶媒体 |
-
2015
- 2015-12-14 WO PCT/CN2015/097237 patent/WO2017100971A1/zh active Application Filing
- 2015-12-14 US US15/753,971 patent/US10628924B2/en active Active
Patent Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US8249377B1 (en) * | 2008-06-08 | 2012-08-21 | Hewlett-Packard Development Company, L.P. | Blurred digital image deblurring |
CN103201765A (zh) * | 2010-09-28 | 2013-07-10 | 马普科技促进协会 | 用于从所观察到的数字图像序列恢复数字图像的方法与设备 |
CN103514582A (zh) * | 2012-06-27 | 2014-01-15 | 郑州大学 | 基于视觉显著的图像去模糊方法 |
CN103049906A (zh) * | 2012-12-07 | 2013-04-17 | 清华大学深圳研究生院 | 一种图像深度提取方法 |
US9208543B2 (en) * | 2013-10-23 | 2015-12-08 | Adobe Systems Incorporated | Deblurring images having spatially varying blur |
Non-Patent Citations (1)
Title |
---|
ZHUO, SHAOJIE ET AL.: "Defocus Map Estimation from a Single Image", PATTERN RECOGNITION, vol. 44, no. 9, 21 March 2011 (2011-03-21), pages 1852 - 1858, XP028217036, ISSN: 0031-3203 * |
Cited By (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110135576A (zh) * | 2018-02-02 | 2019-08-16 | 辉达公司 | 一种用于视频去模糊的无监督学习方法 |
CN110135576B (zh) * | 2018-02-02 | 2023-07-04 | 辉达公司 | 一种用于视频去模糊的无监督学习方法 |
CN112766190A (zh) * | 2021-01-25 | 2021-05-07 | 北京有竹居网络技术有限公司 | 对抗样本生成方法、装置、存储介质及电子设备 |
Also Published As
Publication number | Publication date |
---|---|
US20180276796A1 (en) | 2018-09-27 |
US10628924B2 (en) | 2020-04-21 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
WO2017100971A1 (zh) | 一种失焦模糊图像的去模糊方法和装置 | |
Pan et al. | Soft-segmentation guided object motion deblurring | |
Hyun Kim et al. | Dynamic scene deblurring | |
US10963993B2 (en) | Image noise intensity estimation method, image noise intensity estimation device, and image recognition device | |
Zhong et al. | Handling noise in single image deblurring using directional filters | |
US9692939B2 (en) | Device, system, and method of blind deblurring and blind super-resolution utilizing internal patch recurrence | |
Deng et al. | A guided edge-aware smoothing-sharpening filter based on patch interpolation model and generalized gamma distribution | |
CN106934806B (zh) | 一种基于结构清晰度的无参考图失焦模糊区域分割方法 | |
CN105574823B (zh) | 一种失焦模糊图像的去模糊方法和装置 | |
US20140294299A1 (en) | Image processing apparatus and method | |
Zhang et al. | Improved image deblurring based on salient-region segmentation | |
CN105631854A (zh) | 一种基于fpga平台的自适应图像清晰度评价算法 | |
CN108234826B (zh) | 图像处理方法及装置 | |
CN113011433B (zh) | 一种滤波参数调整方法及装置 | |
CN110503611A (zh) | 图像处理的方法和装置 | |
Safonov et al. | Adaptive sharpening of photos | |
Junjie et al. | An image defocus deblurring method based on gradient difference of boundary neighborhood | |
Yu et al. | Continuous digital zooming of asymmetric dual camera images using registration and variational image restoration | |
CN114519675A (zh) | 图像处理方法、装置、电子设备及可读存储介质 | |
Elhefnawy et al. | Effective visibility restoration and enhancement of air polluted images with high information fidelity | |
Zhu et al. | Efficient perceptual-based spatially varying out-of-focus blur detection | |
Horkaew et al. | Structural adaptive anisotropic NAS-RIF for biomedical image restoration | |
Zhu | Measuring spatially varying blur and its application in digital image restoration | |
Latha et al. | Simultaneous denoising and deblurring by non-convex regularization: alternating minimization framework | |
Li et al. | Hybrid synthesis for exposure fusion from hand-held camera inputs |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
121 | Ep: the epo has been informed by wipo that ep was designated in this application |
Ref document number: 15910445 Country of ref document: EP Kind code of ref document: A1 |
|
WWE | Wipo information: entry into national phase |
Ref document number: 15753971 Country of ref document: US |
|
NENP | Non-entry into the national phase |
Ref country code: DE |
|
32PN | Ep: public notification in the ep bulletin as address of the adressee cannot be established |
Free format text: NOTING OF LOSS OF RIGHTS PURSUANT TO RULE 112(1) EPC (EPO FORM 1205A DATED 08.10.2018) |
|
122 | Ep: pct application non-entry in european phase |
Ref document number: 15910445 Country of ref document: EP Kind code of ref document: A1 |