WO2010098054A1 - 画像補正装置及び画像補正方法 - Google Patents
画像補正装置及び画像補正方法 Download PDFInfo
- Publication number
- WO2010098054A1 WO2010098054A1 PCT/JP2010/001115 JP2010001115W WO2010098054A1 WO 2010098054 A1 WO2010098054 A1 WO 2010098054A1 JP 2010001115 W JP2010001115 W JP 2010001115W WO 2010098054 A1 WO2010098054 A1 WO 2010098054A1
- Authority
- WO
- WIPO (PCT)
- Prior art keywords
- region
- point spread
- spread function
- image
- adaptive
- Prior art date
Links
- 238000003702 image correction Methods 0.000 title claims abstract description 75
- 238000000034 method Methods 0.000 title claims description 43
- 230000003044 adaptive effect Effects 0.000 claims abstract description 176
- 230000011218 segmentation Effects 0.000 claims abstract description 3
- 230000006870 function Effects 0.000 claims description 120
- 238000004364 calculation method Methods 0.000 claims description 60
- 238000011156 evaluation Methods 0.000 claims description 38
- 229920013655 poly(bisphenol-A sulfone) Polymers 0.000 description 46
- 238000012545 processing Methods 0.000 description 21
- 238000010586 diagram Methods 0.000 description 19
- 230000033001 locomotion Effects 0.000 description 16
- 238000012937 correction Methods 0.000 description 9
- 238000013500 data storage Methods 0.000 description 8
- 230000008569 process Effects 0.000 description 8
- 230000006866 deterioration Effects 0.000 description 7
- 238000004891 communication Methods 0.000 description 4
- 230000003287 optical effect Effects 0.000 description 3
- 239000000470 constituent Substances 0.000 description 2
- 230000007423 decrease Effects 0.000 description 2
- 238000005286 illumination Methods 0.000 description 2
- 238000003384 imaging method Methods 0.000 description 2
- 230000000717 retained effect Effects 0.000 description 2
- 230000035945 sensitivity Effects 0.000 description 2
- 230000009466 transformation Effects 0.000 description 2
- 230000005540 biological transmission Effects 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 238000006243 chemical reaction Methods 0.000 description 1
- 230000005484 gravity Effects 0.000 description 1
- 230000010354 integration Effects 0.000 description 1
- 239000004973 liquid crystal related substance Substances 0.000 description 1
- 230000007246 mechanism Effects 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
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
- H04N23/00—Cameras or camera modules comprising electronic image sensors; Control thereof
- H04N23/60—Control of cameras or camera modules
- H04N23/68—Control of cameras or camera modules for stable pick-up of the scene, e.g. compensating for camera body vibrations
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
- H04N23/00—Cameras or camera modules comprising electronic image sensors; Control thereof
- H04N23/60—Control of cameras or camera modules
- H04N23/68—Control of cameras or camera modules for stable pick-up of the scene, e.g. compensating for camera body vibrations
- H04N23/681—Motion detection
- H04N23/6811—Motion detection based on the image signal
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
- H04N23/00—Cameras or camera modules comprising electronic image sensors; Control thereof
- H04N23/60—Control of cameras or camera modules
- H04N23/68—Control of cameras or camera modules for stable pick-up of the scene, e.g. compensating for camera body vibrations
- H04N23/682—Vibration or motion blur correction
- H04N23/683—Vibration or motion blur correction performed by a processor, e.g. controlling the readout of an image memory
-
- 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
- G06T2207/20172—Image enhancement details
- G06T2207/20201—Motion blur correction
Definitions
- the present invention relates to an image correction apparatus that corrects an image captured using a digital still camera, a security camera, or the like, and more particularly to an image correction apparatus that corrects image blurring.
- the captured image may be blurred.
- the blur generated in the image is roughly classified into two types, a hand shake blur and a subject blur.
- Camera shake is image blur caused by the camera moving when the shutter is pressed.
- Subject blur is image blur caused by the subject moving during shooting.
- PSF Point Spread Function
- FIG. 14A a PSF representing deterioration from a non-blurred image 1001 to a defocused image 1002 is represented by an image 1003.
- FIG. 14B a PSF representing deterioration from a non-blurred image 1004 to an image 1005 blurred in the x direction is represented by an image 1006.
- Non Patent Document 1 a method using a low-resolution image sequence photographed using a high-speed shutter and a high-resolution image photographed using a low-speed shutter has been proposed (for example, non Patent Document 1).
- the motion of a subject image is estimated by associating pixels among a plurality of images included in an image sequence of a low-resolution high-speed shutter with less blur. Then, using the estimated movement of the subject image, the image of the low-speed shutter having a large blur is corrected, thereby obtaining a high-resolution image in which the subject blur is corrected.
- Patent Document 1 a method of correcting using a plurality of continuously shot images has been proposed (for example, see Patent Document 1).
- the movement (trajectory) of a subject is obtained based on a plurality of images that are continuously captured, so that all pixel positions of the target image are obtained.
- One deterioration function for example, PSF
- PSF deterioration function
- Non-Patent Document 1 has a problem that subject blur cannot be corrected in an environment where a high-speed shutter cannot be used. For example, when an image is taken with a high-speed shutter in a low illumination environment, only a dark image can be obtained because the amount of light is insufficient. When such a dark image is used, the accuracy of pixel association between images decreases. In addition, when the sensitivity is increased in order to eliminate the shortage of light in a low illumination environment, noise signals caused by the imaging device such as dark current noise and thermal noise are amplified, so the accuracy of pixel correspondence between images is increased. descend. As described above, in the method described in Non-Patent Document 1, in a low-light environment where there are many opportunities for subject blurring, the accuracy of pixel correspondence between images decreases, so subject blurring can be corrected with high accuracy. Can not.
- Patent Document 1 has a description that the deterioration function at all pixel positions of the target image is calculated based on the deterioration function calculated based on a plurality of regions of the target image. Is not listed.
- the present invention has been made in view of the above problems, and provides an image correction apparatus capable of correcting an input image including a plurality of blurs to a target image with less blurs than the input image with high accuracy. For the purpose.
- an image correction apparatus is an image correction apparatus that generates a target image with less blur than the input image by correcting the input image, and configures the input image.
- the input image is divided into a plurality of adaptive regions by determining a region where blurring is common as one adaptive region based on pixel values of pixels to be processed, and image blurring is performed for each of the divided adaptive regions.
- an adaptive region dividing unit that calculates a point spread function indicating the characteristics of the pixel, and using the calculated point spread function, a pixel located between representative pixels that are pixels representing each of the plurality of adaptive regions
- a point spread function interpolation unit that interpolates a point spread function
- an image correction unit that generates the target image by correcting the input image using the point spread function after interpolation.
- the input image can be corrected using a point spread function for each area where blurring is common, so an input image containing multiple blurrings can be accurately corrected to a target image with less blurring than the input image. It becomes possible to do. Furthermore, since the point spread function of pixels located between the representative pixels can be interpolated using the point spread function calculated for each region, the point spread function becomes discontinuous at the boundary of the divided regions. It is possible to suppress a sense of incongruity in the target image.
- the adaptive region dividing unit determines a region having a common blur as one adaptive region based on the similarity of the point spread function.
- This configuration makes it possible to determine the commonness of blur based on the similarity of the point spread function, so that a region with common blur can be determined as one adaptive region with high accuracy.
- the adaptive region dividing unit includes a first point spread function calculating unit that calculates a point spread function of an initial region that is a partial region of the input image as a first point spread function, and the initial region is a holding region.
- An area holding unit to hold as a second point spread function calculation unit that calculates a point spread function of an evaluation area that includes the initial area and is larger than the initial area as a second point spread function;
- a similarity determination unit that determines whether the first point spread function and the second point spread function are similar, and the similarity determination unit determines that the first point spread function and the second point spread function are When it is determined that they are not similar, an adaptive region determination unit that determines a region having a common blur as one adaptive region by determining a holding region last held in the region holding unit as the adaptive region
- the region holding unit holds the evaluation region as a holding region when the similarity determination unit determines that the first point spread function and the second point spread function are similar to each other, and
- the point spread function calculation unit is a region that
- the input image can be adaptively divided into adaptive regions in which a region having a similar point spread function becomes a single region, and thus can be obtained by photographing a plurality of subjects moving in different directions. Further, it is possible to divide the input image into adaptive areas corresponding to the subject. That is, the input image can be corrected using a point spread function that matches the blur direction for each subject image, so that a target image with less blur can be generated.
- the first point spread function calculation unit includes a candidate region selection unit that selects all or a part of the input image as a candidate region and a blur direction of the candidate region selected by the candidate region selection unit.
- An initial region determination that determines the candidate region as the initial region when the blur direction determination unit determines whether the blur direction is a single blur direction and the blur direction determination unit determines that the single blur direction is the single blur direction.
- the candidate area selection unit selects an area smaller than the candidate area as a new candidate area when the blur direction determination unit determines that the direction is not a single blur direction.
- the initial region when determining the adaptive region can be determined as a region with a single blur direction, it is possible to reduce the possibility that subject images having different blur directions are included in the adaptive region. it can.
- the input image can be corrected using the PSF that matches the blur direction for each subject image, so that a target image with less blur can be generated.
- the adaptive region dividing unit determines a region having a common blur as one adaptive region based on whether or not the direction is a single blur.
- This configuration makes it possible to determine the commonality of blurring based on whether or not the direction is a single blurring direction, so that a region where blurring is common can be determined as one adaptive region with high accuracy.
- the adaptive region dividing unit includes a candidate region selecting unit that selects all or a part of the input image as a candidate region, and a blur direction of the candidate region selected by the candidate region selecting unit is single.
- a candidate region selecting unit that selects all or a part of the input image as a candidate region
- a blur direction of the candidate region selected by the candidate region selecting unit is single.
- the point spread function interpolation unit divides each of the at least two point spread functions into two or more when the at least two point spread functions of the calculated point spread functions are expressed by a straight line. It is preferable to interpolate the point spread functions of the pixels located between the representative pixels of the adaptive area corresponding to the at least two point spread functions based on the correspondence relationship of the point spread functions.
- This configuration makes it possible to easily interpolate the point spread function when the point spread function is represented by a straight line.
- the point spread function interpolation unit is configured to correspond to each of the point spread functions based on a correspondence obtained by associating the calculated point spread functions with each other using a matching method based on dynamic programming. It is preferable to interpolate the point spread function of the pixels located between the representative pixels in the target area.
- This configuration makes it possible to interpolate the point spread function even when the point spread function is not represented by a straight line.
- the point spread function interpolation unit determines whether or not the calculated point spread functions are similar to each other, and is positioned between representative pixels in the adaptive area corresponding to the point spread functions determined to be similar to each other. It is preferable to interpolate the point spread function of the pixel to be processed.
- the present invention can be realized not only as such an image correction apparatus, but also as an integrated circuit including characteristic components included in such an image correction apparatus.
- the present invention can be realized not only as such an image correction apparatus, but also as an image correction method in which the operation of a characteristic component included in such an image correction apparatus is used as a step.
- the present invention can also be realized as a program that causes a computer to execute each step included in the image correction method.
- a program can be distributed via a recording medium such as a CD-ROM (Compact Disc-Read Only Memory) or a transmission medium such as the Internet.
- the image correction apparatus divides an input image including a plurality of blurs into a plurality of adaptive regions so that a region where blurs are common becomes one adaptive region. be able to.
- the image correction apparatus can correct the input image for each region where blurring is common, the input image including a plurality of blurrings is corrected to a target image with less blurring than the input image with high accuracy. It becomes possible.
- FIG. 1 is a block diagram showing a functional configuration of an image correction apparatus according to Embodiments 1 and 2 of the present invention.
- FIG. 2 is a block diagram showing a functional configuration of the program storage unit according to Embodiment 1 of the present invention.
- FIG. 3 is a flowchart showing the overall processing flow of the image correction apparatus according to Embodiments 1 and 2 of the present invention.
- FIG. 4 is a flowchart showing a flow of processing by the adaptive region dividing unit according to Embodiment 1 of the present invention.
- FIG. 5 is a diagram for explaining a flow of region division processing by the adaptive region division unit according to Embodiment 1 of the present invention.
- FIG. 1 is a block diagram showing a functional configuration of an image correction apparatus according to Embodiments 1 and 2 of the present invention.
- FIG. 2 is a block diagram showing a functional configuration of the program storage unit according to Embodiment 1 of the present invention.
- FIG. 3 is a flowchart showing the overall processing flow of
- FIG. 6 is a flowchart showing a flow of processing by the PSF interpolation unit according to Embodiments 1 and 2 of the present invention.
- FIG. 7 is a diagram for explaining PSF interpolation by the PSF interpolation unit according to Embodiments 1 and 2 of the present invention.
- FIG. 8 is a diagram for explaining PSF interpolation by the PSF interpolation unit according to Embodiments 1 and 2 of the present invention.
- FIG. 9 is a flowchart showing a flow of processing relating to determination of an initial region by the first PSF calculation unit according to Embodiment 1 of the present invention.
- FIG. 10 is a diagram for explaining a flow of processing relating to determination of an initial region by the first PSF calculation unit according to Embodiment 1 of the present invention.
- FIG. 10 is a diagram for explaining a flow of processing relating to determination of an initial region by the first PSF calculation unit according to Embodiment 1 of the present invention.
- FIG. 11 is a block diagram showing a functional configuration of the program storage unit according to Embodiment 2 of the present invention.
- FIG. 12 is a flowchart showing a process flow of the adaptive region dividing unit according to Embodiment 2 of the present invention.
- FIG. 13 is a block diagram showing a functional configuration of the system LSI according to the present invention.
- FIG. 14A is a diagram for explaining the PSF.
- FIG. 14B is a diagram for explaining the PSF.
- FIG. 1 is a block diagram showing a functional configuration of an image correction apparatus 100 according to Embodiment 1 of the present invention.
- the image correction apparatus 100 includes an arithmetic control unit 101, a memory unit 102, a display unit 103, an input unit 104, a communication I / F (interface) unit 105, a data storage unit 106, and a program storage unit 107. Is provided. These components 101 to 107 are connected via a bus 108.
- the arithmetic control unit 101 is a CPU (Central Processing Unit), a numerical processor, etc., and loads and executes a necessary program from the program storage unit 107 to the memory unit 102 in accordance with an instruction from the user.
- the constituent units 102 to 107 are controlled.
- the memory unit 102 is a RAM (Random Access Memory) that provides a work area for the arithmetic control unit 101.
- the display unit 103 is a CRT (Cathode-Ray Tube), an LCD (Liquid Crystal Display), or the like.
- the input unit 104 is a keyboard, a mouse, or the like. The display unit 103 and the input unit 104 are used for dialogue between the image correction apparatus 100 and the user under the control of the arithmetic control unit 101.
- the communication I / F unit 105 is a LAN adapter or the like, and is used for communication between the image correction apparatus 100 and the camera 400 or the like.
- the data storage unit 106 is a hard disk, a flash memory, or the like that stores an input image acquired by the camera 400 or the like and a target image obtained by correcting the input image.
- the program storage unit 107 is a ROM (Read Only Memory) that stores various programs for realizing the functions of the image correction apparatus 100.
- FIG. 2 is a block diagram showing a functional configuration of the program storage unit 107 according to Embodiment 1 of the present invention.
- the program storage unit 107 is functionally (as a processing unit that functions when executed by the arithmetic control unit 101), an adaptive region dividing unit 110, a PSF interpolation unit 120, and an image correction unit 130.
- the program storage unit 107 is functionally (as a processing unit that functions when executed by the arithmetic control unit 101), an adaptive region dividing unit 110, a PSF interpolation unit 120, and an image correction unit 130.
- the adaptive region dividing unit 110 divides the input image into a plurality of adaptive regions by determining a region with common blur as one adaptive region based on the pixel values of the pixels constituting the input image. Furthermore, the adaptive area dividing unit 110 calculates a PSF for each divided adaptive area. Specifically, the adaptive region dividing unit 110 determines a region where blurring is common as one adaptive region based on the similarity of PSFs.
- the adaptive region dividing unit 110 includes a first PSF calculation unit 111, a region holding unit 112, a second PSF calculation unit 113, a similarity determination unit 114, and an adaptive region determination unit 115. Prepare.
- the first PSF calculation unit 111 calculates the PSF of the initial area, which is a partial area of the input image, as the first PSF. Specifically, the first PSF calculation unit 111 includes a candidate region selection unit 111a, a blur direction determination unit 111b, and an initial region determination unit 111c. Then, the first PSF calculation unit 111 calculates the PSF of the initial region determined by the initial region determination unit 111c as the first PSF.
- the candidate area selection unit 111a selects all or a part of the input image as a candidate area. In addition, when the blur direction determination unit 111b determines that the single blur direction is not determined, the candidate region selection unit 111a selects a region smaller than the candidate region as a new candidate region.
- the blur direction determination unit 111b determines whether the blur direction of the candidate area selected by the candidate area selection unit 111a is a single blur direction.
- the initial region determination unit 111c determines a candidate region as an initial region when the blur direction determination unit 111b determines that a single blur direction is present.
- the area holding unit 112 holds the initial area as a holding area. In addition, the area holding unit 112 holds the evaluation area as a holding area when the similarity determination unit 114 described later determines that the first PSF and the second PSF are similar.
- the second PSF calculation unit 113 calculates the PSF of the evaluation area that is an area including the initial area and larger than the initial area as the second PSF. Further, when the similarity determination unit 114 determines that the first PSF and the second PSF are similar, the second PSF calculation unit 113 includes a PSF of a new evaluation region that includes the evaluation region and is larger than the evaluation region. Is calculated as the second PSF.
- the similarity determination unit 114 determines whether the first PSF and the second PSF are similar.
- the adaptive region determination unit 115 determines the holding region held last in the region holding unit 112 as the adaptive region. That is, the adaptive area determination unit 115 determines an area where blurring is common as one adaptive area based on the similarity of PSFs.
- the PSF interpolation unit 120 uses the calculated PSF to interpolate the PSF of the pixels located between the representative pixels.
- the representative pixel is a pixel that represents each of the plurality of regions.
- the representative pixel is a pixel such as the center or the center of gravity of each region.
- the PSF interpolation unit 120 associates PSFs calculated for each adaptive region with each other using a matching method (hereinafter referred to as “DP matching”) by dynamic programming (Dynamic Programming), for example. Then, the PSF interpolation unit 120 interpolates the PSFs of the pixels located between the representative pixels of the adaptive area corresponding to the PSF, based on the correspondence obtained by associating the PSFs in this way.
- DP matching a matching method by dynamic programming
- the PSF interpolation unit 120 may divide and associate the at least two PSFs. Specifically, the PSF interpolation unit 120 calculates PSFs of pixels located between the representative pixels in the adaptive area corresponding to the at least two PSFs based on the correspondence obtained by associating the PSFs in this way. Interpolation may be performed.
- the PSF interpolation unit 120 determines whether the PSFs are similar to each other, and interpolates the PSFs of pixels located between the representative pixels in the adaptive area corresponding to the PSFs determined to be similar to each other.
- the image correction unit 130 generates a target image with less blur than the input image by correcting the input image using PSF.
- FIG. 3 is a flowchart showing an overall processing flow of the image correction apparatus 100 according to Embodiment 1 of the present invention.
- the adaptive region dividing unit 110 divides the input image into a plurality of adaptive regions of an adaptive size according to the pixel values of the pixels constituting the input image stored in the data storage unit 106. Further, the adaptive region dividing unit 110 calculates a PSF for each adaptive region (S101). Subsequently, the PSF interpolation unit 120 uses the PSF calculated for each adaptive region to interpolate the PSFs of pixels located between the representative pixels in each adaptive region (S102).
- the image correcting unit 130 corrects the input image using the PSF for each pixel after interpolation, thereby generating a target image with less blur than the input image (S103).
- the image correction unit 130 corrects the input image using the PSF for each pixel obtained based on the processing in step S101 and step S102, using the Richardson-Lucy method represented by the following equation (1).
- I indicates an image after correction (target image).
- K represents PSF.
- B represents an input image.
- the image correction unit 130 corrects the input image using the Richardson-Lucy method, but corrects the input image using other methods such as the Fourier log method and the maximum entropy method. May be.
- FIG. 4 is a flowchart showing a flow of processing (S101) by the adaptive region dividing unit 110 according to Embodiment 1 of the present invention.
- the first PSF calculation unit 111 determines an initial region that is a partial region of the input image (S201). Details of the process for determining the initial area will be described later.
- the first PSF calculation unit 111 calculates the PSF of the determined initial region as the first PSF (S202). Specifically, the first PSF calculating unit 111 calculates a PSF from one image (for example, see Non-Patent Document 2, “Removing Camera Shake from a single image (Rob Fergus et.al, SIGGRAPH 2006)”). To calculate the PSF.
- the data storage unit 106 stores in advance an image gradient distribution that appears in a general natural image without blurring. Then, the first PSF calculation unit 111 repeatedly compares the image gradient distribution obtained when the initial region is corrected using a given PSF and the image gradient distribution stored in the data storage unit 106. Then, a PSF whose image gradient distribution matches or is similar is searched. Then, the first PSF calculation unit 111 calculates the PSF obtained as a result of the search in this way as the first PSF.
- the area holding unit 112 holds the initial area determined by the first PSF calculation unit 111 as a holding area (S203).
- the second PSF calculation unit 113 selects an evaluation area (S204).
- This evaluation area is an area including the initial area and is larger than the initial area. For example, when the initial area is a rectangular area, the second PSF calculation unit 113 selects an area that is one pixel larger in the x direction and one pixel larger in the y direction than the initial area as the evaluation area. Then, the second PSF calculation unit 113 calculates the PSF of the selected evaluation area as the second PSF (S205).
- the second PSF calculation unit 113 does not need to determine an area larger than the initial area in both the x direction and the y direction as the evaluation area.
- the second PSF calculation unit 113 may determine an area expanded from the initial area as an evaluation area for only one of the x direction and the y direction.
- the second PSF calculation unit 113 may determine, for example, an area expanded in the x direction or the y direction only as a part of the outer edge of the initial area as the evaluation area.
- the second PSF calculation unit 113 can select an evaluation area flexibly, and thus can increase the possibility of selecting an image of a subject that moves in the same manner as one area.
- the similarity determination unit 114 determines whether or not the first PSF and the second PSF are similar (S206). Specifically, the similarity determination unit 114 calculates the inter-image distance when the first PSF and the second PSF are expressed as images as the similarity indicating the PSF similarity. For example, the similarity determination unit 114 calculates the L1 norm calculated by the following equation (2) as the similarity. The L1 norm indicates that the smaller the value, the higher the degree of similarity.
- P1ij indicates a PSF value of a pixel specified by coordinates (i, j) when the first PSF is expressed as an image.
- P2ij indicates the PSF value of the pixel specified by the coordinates (i, j) when the second PSF is expressed as an image.
- the similarity determination unit 114 determines that the first PSF and the second PSF are similar when the L1 norm of the first PSF and the second PSF is smaller than a predetermined threshold. Conversely, the similarity determination unit 114 determines that the first PSF and the second PSF are not similar when the L1 norm of the first PSF and the second PSF is equal to or greater than a predetermined threshold.
- the similarity determination unit 114 calculates the L1 norm as the similarity, but the similarity may be calculated using a similarity determination method between other images. For example, the similarity determination unit 114 may calculate the L2 norm as the similarity.
- the area holding unit 112 holds the evaluation area as a holding area (S207). Further, the second PSF calculation unit 113 selects a new evaluation area (S204). This new evaluation area is an area including the holding area held in step S207 and is larger than the holding area held in step S207. And the process of step S205 and step S206 is performed using a new evaluation area
- the adaptive region determination unit 115 determines the retained region retained last in step S207 as the adaptive region ( S208).
- the first PSF calculation unit 111 determines whether or not all the pixels of the input image are included in the already determined adaptive region (S209).
- the first PSF calculation unit 111 when all the pixels of the input image are not included in the already determined adaptive area (No in S209), the first PSF calculation unit 111 includes an area that includes only pixels that are not yet included in the adaptive area. Is determined as the initial region (S201). For example, the first PSF calculation unit 111 determines the initial region so as to include pixels adjacent to the right end or the lower end of the determined adaptive region. Note that the first PSF calculation unit 111 may determine the initial region so as not to include adjacent pixels but to include pixels separated by the number of pixels determined by a priori knowledge. Further, the first PSF calculation unit 111 may determine the initial region so as to include pixels separated by the number of pixels indicated by the input value received from the user received by the input unit 104.
- step S102 in FIG. 3 is executed.
- the adaptive region dividing unit 110 divides the input image into a plurality of adaptive regions by repeating the processing from step S201 to step S209.
- FIG. 5 is a diagram for explaining the flow of region division processing by the adaptive region dividing unit 110 according to Embodiment 1 of the present invention.
- the first PSF calculation unit 111 determines an initial region 501. Then, the first PSF calculation unit 111 calculates the first PSF 502 (P1) that is the PSF of the initial region 501. Note that although the initial area 501 is rectangular here, the initial area 501 does not necessarily have to be rectangular. For example, the initial region 501 may have an arbitrary shape such as a rhombus, a parallelogram, or a circle.
- the second PSF calculation unit 113 selects an area including the initial area 501 and larger than the initial area 501 as the evaluation area 503. Then, the second PSF calculation unit 113 calculates the PSF of the evaluation area 503 as the second PSF 504 (P2).
- the evaluation area 503 is rectangular, but the evaluation area 503 is not necessarily rectangular.
- the evaluation region 503 may have an arbitrary shape such as a rhombus, a parallelogram, or a circle.
- the similarity determination unit 114 calculates the L1 norm of the first PSF 502 and the second PSF 504. Since the calculated L1 norm is smaller than the threshold value TH, the similarity determination unit 114 determines that the first PSF 502 and the second PSF 504 are similar. Therefore, the area holding unit 112 holds the evaluation area 503.
- the second PSF calculation unit 113 selects an area including the evaluation area 503 and larger than the evaluation area 503 as the evaluation area 505. Then, the second PSF calculation unit 113 calculates the PSF of the evaluation area 505 as the second PSF 506 (P3).
- the similarity determination unit 114 calculates the L1 norm of the first PSF 502 and the second PSF 506. Since the calculated L1 norm is equal to or greater than the threshold value TH, the similarity determination unit 114 determines that the first PSF 502 and the second PSF 504 are not similar. Therefore, the adaptive area determination unit 115 determines the evaluation area 503 last held by the area holding unit 112 as an adaptive area.
- the first PSF calculation unit 111 determines an area adjacent to the determined adaptive area as the initial area 507.
- the adaptive region dividing unit 110 divides the input image into a plurality of adaptive regions as shown in FIG. To do.
- the PSF interpolation unit 120 interpolates the PSF as described below in order to reduce the uncomfortable feeling of the target image after correction.
- FIG. 6 is a flowchart showing a flow of processing (S102) by the PSF interpolation unit 120 according to Embodiment 1 of the present invention.
- the PSF interpolation unit 120 selects a PSF group including at least two PSFs from a plurality of PSFs calculated for each of the plurality of adaptive regions divided by the adaptive region dividing unit 110 (S301). For example, the PSF interpolation unit 120 selects PSFs in two adaptive regions adjacent to each other.
- the PSF interpolation unit 120 searches for corresponding points that are corresponding points between the PSFs included in the selected PSF group (S302). For example, as shown in FIG. 7, when each of the two PSFs is represented by one straight line, the PSF interpolation unit 120 divides each of the two PSFs into N equal parts (N is a positive integer). Then, the PSF interpolation unit 120 searches for corresponding points by associating N equally divided PSFs between two PSFs. Further, for example, as shown in FIG. 8, when at least one of the two PSFs is not represented by one straight line, a corresponding point may be searched by DP matching. The combination of corresponding points searched here is called a correspondence relationship.
- the PSF interpolation unit 120 calculates the distance between the searched corresponding points (S303). Specifically, the PSF interpolation unit 120 calculates the L1 norm, for example, as the distance, as in step S206 in FIG. When the L1 norm is calculated as a distance, the smaller the L1 norm is, the higher the degree of similarity is.
- the PSF interpolation unit 120 determines whether or not the calculated distance is smaller than a predetermined threshold (S304). That is, the PSF interpolation unit 120 determines whether the PSFs included in the PSF group are similar to each other.
- the PSF interpolation unit 120 executes the process of Step S306. That is, when it is determined that the PSFs are not similar to each other, the PSF interpolation unit 120 does not interpolate the PSFs in the adaptive region corresponding to the PSFs. That is, the PSF of each pixel in the adaptive region is uniform with the PSF calculated for the adaptive region.
- the PSF interpolation unit 120 does not interpolate the PSFs because the subject images included in the adaptive regions corresponding to the PSFs that are not similar to each other move differently. This is because there is a high possibility that the image is an image of a subject that is playing. That is, when the PSF is interpolated in a plurality of adaptive regions including images of different subjects as described above, the PSF corresponding to the motion of the subject which is not possible may be interpolated.
- the PSF interpolation unit 120 interpolates the PSF in the adaptive region corresponding to the PSF (S305). That is, when it is determined that the PSFs are similar to each other, the PSF interpolation unit 120 interpolates the PSFs of pixels located between the representative pixels in the adaptive region. That is, the PSF interpolation unit 120 interpolates PSFs of pixels that are inside and outside each adaptive region and are pixels other than the representative pixel based on the correspondence relationship between the distance and the PSF from the representative pixel.
- the PSF interpolation unit 120 calculates a new PSF based on the correspondence relationship of the PSF according to the distance between the pixel position to be interpolated and each representative pixel, and uses the calculated PSF as the PSF of the pixel. Interpolate.
- the PSF interpolation unit 120 interpolates the PSF in the adaptive region corresponding to the PSF only when it is determined that the PSFs are similar to each other. This is because if the PSFs are similar, the subject image included in each of the adaptive regions corresponding to the PSF is highly likely to be an image of one subject that is moving in the same way. As described above, the PSF interpolation unit 120 interpolates the PSF between a plurality of adaptive regions including an image of one subject, so that it is possible to reduce a sense of incongruity that occurs at the boundary of the adaptive region in the target image. .
- the PSF interpolation unit 120 may interpolate the PSF by any interpolation method such as linear interpolation, polynomial interpolation, or spline interpolation.
- the PSF interpolation unit 120 determines whether all the PSFs have already been selected (S306).
- the PSF interpolation unit 120 selects a PSF group including PSFs that are not yet selected (S301).
- the process of step S103 in FIG. 3 is executed.
- the PSF interpolation unit 120 uses the PSF calculated for each adaptive region, and interpolates the PSFs of pixels located between pixels representing each adaptive region. As a result, the PSF interpolation unit 120 can suppress an unnaturally large change in the PSF at the boundary between adjacent adaptive regions. That is, the image correction unit 130 can generate a target image that does not feel strange at the boundary of the adaptive region.
- the first PSF calculation unit 111 determines whether an initial region is too small, it is difficult for the first PSF calculation unit 111 to distinguish between texture and blur, and thus it is difficult to calculate the PSF.
- the initial region is too large, there is a high possibility that an image of a subject that moves differently is included in the initial region. Therefore, it is difficult for the first PSF calculation unit 111 to calculate a PSF according to the motion of the subject. Therefore, a method for determining an initial region while reducing the size from a large region in order so that images of a plurality of subjects that move differently are not included will be described below.
- FIG. 9 is a flowchart showing a flow of processing (S201) related to determination of the initial region by the first PSF calculation unit 111 according to Embodiment 1 of the present invention.
- the candidate area selection unit 111a selects all or a part of the input image as a candidate area (S401). Subsequently, the blur direction determination unit 111b converts the pixel value of the selected candidate region into a frequency space using a discrete Fourier transform (DFT: Discrete Fourier Transform) or the like (S402).
- DFT discrete Fourier transform
- the blur direction determination unit 111b determines whether or not the frequency distribution obtained by converting into the frequency space can be expressed by a sinc function (S403). What was a rectangular wave in the image space is represented by a sinc function (formula (3)) in the frequency space.
- the blur direction determination unit 111b determines whether the frequency distribution obtained by converting to the frequency space can be expressed using the sinc function, so that the blur direction of the candidate region is a single blur direction. It is determined whether or not there is.
- the blur direction determination unit 111b calculates a correlation value between the frequency distribution obtained by converting into the frequency space and each of the plurality of sinc functions.
- the plurality of sinc functions are sinc functions corresponding to combinations of a plurality of amplitudes and a plurality of phases.
- the blur direction determination unit 111b calculates a correlation value using a method of calculating similarity between images such as L1 norm, L2 norm, or normalized correlation.
- the blur direction determination unit 111b determines whether or not there is a sinc function whose correlation value exceeds a predetermined threshold value.
- the blur direction determination unit 111b determines that the frequency distribution can be expressed using the sinc function.
- the blur direction determination unit 111b determines that the frequency distribution cannot be expressed by the sinc function.
- the initial area determination unit 111c sets the candidate area to the initial position. The area is determined (S405).
- the initial area determination unit 111c sets the candidate area as the initial area. Not determined as. Therefore, the blur direction determination unit 111b determines whether or not the size of the candidate area is smaller than a predetermined threshold value (S404).
- the predetermined threshold value is a value indicating the size of an area in which a PSF capable of correcting blurring of an input image with high accuracy can be calculated. The size of this area differs depending on the exposure time when the image is taken. For example, when the exposure time is about 1 second, the threshold value may be a value of about 100 pixels (10 ⁇ 10 pixels).
- the candidate area selection unit 111a selects an area smaller than the current candidate area as a new candidate area. (S401).
- the initial area determination unit 111c determines the currently selected candidate area as the initial area (S405). This is because since the candidate area is small, the subject image included in the candidate area is considered to be a single subject image. That is, since the subject image included in the candidate region is considered to be an image of a subject that has moved in a plurality of directions, the initial region determination unit 111c determines the currently selected candidate region as the initial region without further reducing the candidate region. Determine as.
- the first PSF calculation unit 111 determines the initial region while reducing the candidate region until it is determined that the frequency distribution can be expressed by the sinc function or until a predetermined size is obtained. Therefore, the first PSF calculation unit 111 can determine a small region that does not include PSFs with different blur directions as the initial region.
- FIG. 10 is a diagram for explaining a flow of processing related to determination of an initial region by the first PSF calculation unit 111 according to Embodiment 1 of the present invention.
- the candidate area selection unit 111a selects a relatively large area as a candidate area 601 that has not yet been determined as an adaptive area. Note that the candidate area selection unit 111a may select the entire image as a candidate area.
- the blur direction determination unit 111b converts the pixel value of the candidate region 601 into a frequency component by discrete Fourier transform or the like.
- FIG. 10B is a diagram showing an image 602 representing the converted frequency component.
- the blur direction determination unit 111b includes a plurality of images corresponding to a plurality of amplitude and phase sinc functions, and an image 602 representing the frequency component obtained by the conversion. A correlation value is calculated.
- the image 602 shown in FIG. 10B is an image in which points indicating frequency components spread vertically and horizontally. Therefore, the correlation value between the image 602 and the image corresponding to the sinc function is small. That is, there is no image similar to the image 602 corresponding to the sinc function.
- the blur direction determination unit 111b determines that the blur direction of the candidate area 601 is not a single blur direction. Then, the candidate area selection unit 111a selects an area smaller than the candidate area 601 as a new candidate area 603 as illustrated in FIG. Then, the blur direction determination unit 111b converts the pixel value of the candidate region 603 into a frequency component using a discrete Fourier transform or the like.
- FIG. 10E is a diagram illustrating an image 604 representing the converted frequency component. Then, as illustrated in FIG. 10F, the blur direction determination unit 111b includes a plurality of images corresponding to each of a plurality of amplitude and phase sinc functions, and an image 604 representing a frequency component obtained by using the transformation. The correlation value is calculated.
- the image 604 shown in FIG. 10E is an image in which points indicating frequency components are arranged in one straight line. Therefore, the correlation value between the image 604 and the image corresponding to the sinc function is large. That is, an image corresponding to a sinc function similar to the image 604 exists. Therefore, the blur direction determination unit 111b determines that the blur direction of the candidate area 603 is a single blur direction. Then, the initial region determination unit 111c determines the candidate region 603 as the initial region.
- the first PSF calculation unit 111 can determine whether the gradation is due to texture or the gradation due to blur by determining the initial region based on whether or not it can be expressed by a sinc function. It becomes possible.
- the image correction apparatus 100 can divide an input image into a plurality of adaptive regions so that a region where blurring is common becomes one adaptive region.
- the image correction apparatus 100 can correct the input image using the PSF for each region where shake is common, the input image including a plurality of shakes is changed to a target image with less shake than the input image. It becomes possible to correct with high accuracy.
- the image correction apparatus 100 can interpolate the PSF of the pixels located between the representative pixels using the PSF calculated for each region, the PSF becomes discontinuous at the boundary between the divided regions. It is possible to suppress a sense of incongruity in the target image. Further, the image correction apparatus 100 can correct an input image without using an image other than the input image.
- the image correction apparatus 100 can divide the input image into adaptive regions so that a region with a similar PSF becomes one adaptive region. Therefore, the image correction apparatus 100 can divide a single input image obtained by photographing a plurality of subjects moving in different directions into regions corresponding to the subjects. That is, since the image correction unit 130 can correct the input image using the PSF that matches the blur direction for each subject image, the image correction apparatus 100 can generate a target image with less blur. .
- the image correction apparatus 100 can determine the initial region when determining the adaptive region as a region of a single blur direction, there is a possibility that subject images having different blur directions are included in the adaptive region. Can be reduced. That is, the image correction apparatus 100 can correct the input image using the PSF that matches the blur direction of the subject image, and thus can generate a target image with less blur.
- the PSF interpolation unit 120 does not interpolate PSFs for adaptive regions where PSFs are not similar to each other. Therefore, the PSF interpolation unit 120 can reduce the possibility of interpolating the PSF with respect to the boundary between subject images that are moving differently. That is, the PSF interpolation unit 120 can reduce the possibility of interpolating a PSF corresponding to a motion different from the actual subject motion.
- adaptive area dividing section 110 determines an adaptive area so that areas with similar PSFs become one area, but image correction apparatus 100 to which the present invention is applied is
- the adaptive region segmentation unit 110 divides an input image into a plurality of adaptive regions, each of which is a region where blurring is common, using a priori knowledge that similar textures are likely to move the same. May be.
- the adaptive area dividing unit 110 also uses, for example, an area dividing method (for example, a graph cut) used in the computer vision field, and inputs an input image to a plurality of adaptive areas, each of which is a region where blurring is common. May be divided. Thereby, the adaptive area dividing unit 110 can perform flexible texture division.
- the image correction apparatus 100 can capture a plurality of images taken with a long exposure time. Only a region with motion may be extracted from the image using the inter-image difference, and the above-described PSF interpolation and image correction may be performed only on the extracted region. Accordingly, the image correction apparatus 100 can reduce the total calculation cost, memory cost, and the like.
- the first PSF calculation unit 111 determines the initial area while reducing the candidate area. For example, the first PSF calculation unit 111 determines an area having a predetermined size corresponding to the exposure time of the input image. The initial region may be determined. In this case, the data storage unit 106 preferably stores a table in which the exposure time is associated with the size of the area. Then, the first PSF calculation unit 111 acquires the size of the area corresponding to the exposure time of the input image by referring to the table stored in the data storage unit 106, and uses the acquired size area as the initial area. It is preferable to determine.
- the image correction apparatus 200 according to the second embodiment and the image correction apparatus 100 according to the first embodiment are different in the configuration of the adaptive area dividing unit included in the program storage unit 107, but the other components are the same. . Therefore, in the following, illustration and description of a block diagram having the same configuration as in the first embodiment and a flowchart having the same processing flow are omitted. Moreover, the same code
- FIG. 11 is a block diagram showing a functional configuration of the program storage unit 107 according to Embodiment 2 of the present invention.
- the adaptive region dividing unit 210 determines a region where blurring is common as one adaptive region based on whether or not there is a single blur direction. As shown in FIG. 11, the adaptive region division unit 210 includes a candidate region selection unit 211, a blur direction determination unit 212, an adaptive region determination unit 213, and a PSF calculation unit 214.
- the candidate area selection unit 211 determines candidate areas that are all or part of the input image. In addition, when the blur direction determination unit 212 determines that the single blur direction is not the single blur direction, the candidate region selection unit 211 selects a region smaller than the candidate region as a new candidate region.
- the blur direction determination unit 212 determines whether or not the blur direction of the candidate area selected by the candidate area selection unit 211 is a single blur direction.
- the adaptive region determination unit 213 determines a candidate region as an adaptive region when the blur direction determination unit 212 determines that the single blur direction is present. That is, the adaptive region determination unit 213 determines a region with common blur as one adaptive region based on whether or not the direction is a single blur direction.
- the PSF calculation unit 214 calculates the PSF of the adaptive region determined by the adaptive region determination unit 213.
- FIG. 12 is a flowchart showing a flow of processing (S101) of the adaptive region dividing unit 210 according to Embodiment 2 of the present invention.
- the candidate area selection unit 211 selects all or a part of the input image as a candidate area (S501). Subsequently, the blur direction determination unit 212 converts the pixel value of the selected candidate region into a frequency space by a discrete Fourier transform or the like (S502).
- the blur direction determination unit 212 determines whether or not the frequency distribution obtained by converting to the frequency space can be expressed by a sinc function (S503). That is, the blur direction determination unit 212 determines whether or not there is a single blur direction in the candidate area.
- the blur direction determination unit 212 calculates a correlation value between the frequency distribution obtained by converting into the frequency space and each of the plurality of sinc functions.
- the plurality of sinc functions are sinc functions corresponding to combinations of a plurality of amplitudes and a plurality of phases.
- the blur direction determination unit 212 calculates a correlation value using a method of calculating similarity between images such as L1 norm, L2 norm, or normalized correlation.
- the blur direction determination unit 212 determines whether there is a sinc function having a correlation value exceeding a predetermined threshold among a plurality of sinc functions.
- the blur direction determination unit 212 determines that the frequency distribution can be expressed using the sinc function.
- the blur direction determination unit 212 determines that the frequency distribution cannot be expressed using the sinc function.
- the adaptive region determination unit 213 selects the candidate region.
- the adaptive area is determined (S505).
- the blur direction determination unit 212 determines whether the size of the candidate area is smaller than a predetermined threshold (S504).
- the predetermined threshold value is a value indicating the size of an area in which a PSF capable of correcting blurring of an input image with high accuracy can be calculated. The size of this area differs depending on the exposure time when the image is taken. For example, when the exposure time is about 1 second, the threshold value may be a value of about 100 pixels (10 ⁇ 10 pixels).
- the candidate area selection unit 211 selects an area smaller than the current candidate area as a new candidate area. (S501).
- the adaptive region determination unit 213 determines the currently selected candidate region as the adaptive region (S505). ). This is because since the candidate area is small, the subject image included in the candidate area is considered to be a single subject image. That is, since the subject image included in the candidate region is considered to be an image of a single subject moved in a plurality of directions, the adaptive region determination unit 213 can select the currently selected candidate without further reducing the candidate region. The region is determined as an adaptive region.
- the first PSF calculation unit 111 determines the adaptive region while reducing the candidate region until it is determined that the frequency distribution can be expressed using the sinc function or until a predetermined size is obtained. Therefore, the first PSF calculation unit 111 can determine an area that does not include a PSF of a different blur direction as an adaptive area.
- the image correction apparatus 200 can determine a region in a single blur direction as an adaptive region, so that one region including images of a plurality of subjects moving in different directions is used. Dividing the input image as one adaptive region can be suppressed. In other words, the image correction apparatus 200 can correct the input image using the PSF corresponding to the blur direction of the subject image, so that a target image with less blur can be generated.
- the image correction apparatus has been described based on the embodiments.
- the present invention is not limited to these embodiments. Unless it deviates from the meaning of this invention, the form which carried out various deformation
- the constituent elements of the image correction apparatus may be configured by one system LSI (Large Scale Integration).
- the system LSI is an ultra-multifunctional LSI manufactured by integrating a plurality of components on a single chip.
- the system LSI is a computer system including a microprocessor, a ROM (Read Only Memory), a RAM (Random Access Memory), and the like. More specifically, for example, as shown in FIG. 13, the adaptive area dividing unit 110, the PSF interpolation unit 120, and the image correction unit 130 may be configured by one system LSI 300.
- the present invention may be realized as an image photographing device including the image correction device according to the above-described embodiment and a photographing unit including an optical system and an imaging element.
- the image correction apparatus has a high-precision subject blur correction function, and can realize high sensitivity of a digital still camera or a security camera, so that it is useful as nighttime shooting, security by a camera in a dark environment, etc. It is.
Landscapes
- Engineering & Computer Science (AREA)
- Multimedia (AREA)
- Signal Processing (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Image Processing (AREA)
- Studio Devices (AREA)
- Facsimile Image Signal Circuits (AREA)
Abstract
Description
図1は、本発明の実施の形態1に係る画像補正装置100の機能構成を示すブロック図である。
次に、本発明の実施の形態2について、図面を参照しながら説明する。
101 演算制御部
102 メモリ部
103 表示部
104 入力部
105 通信I/F部
106 データ記憶部
107 プログラム格納部
108 バス
110、210 適応的領域分割部
111 第1PSF算出部
111a、211 候補領域選択部
111b、212 ブレ方向判定部
111c 初期領域決定部
112 領域保持部
113 第2PSF算出部
114 類似度判定部
115、213 適応的領域決定部
120 PSF補間部
130 画像補正部
214 PSF算出部
300 システムLSI
400 カメラ
Claims (13)
- 入力画像を補正することにより、前記入力画像よりもブレの少ない目的画像を生成する画像補正装置であって、
前記入力画像を構成する画素の画素値に基づいて、ブレが共通する領域を1つの適応的領域として決定することにより、前記入力画像を複数の適応的領域に分割し、分割した適応的領域ごとに、画像のブレの特徴を示す点広がり関数を算出する適応的領域分割部と、
算出された前記点広がり関数を用いて、前記複数の適応的領域のそれぞれを代表する画素である代表画素間に位置する画素の点広がり関数を補間する点広がり関数補間部と、
補間後の前記点広がり関数を用いて前記入力画像を補正することにより、前記目的画像を生成する画像補正部とを備える
画像補正装置。 - 前記適応的領域分割部は、点広がり関数の類似性に基づいて、ブレが共通する領域を1つの適応的領域として決定する
請求項1に記載の画像補正装置。 - 前記適応的領域分割部は、
前記入力画像の一部の領域である初期領域の点広がり関数を第1点広がり関数として算出する第1点広がり関数算出部と、
前記初期領域を保持領域として保持する領域保持部と、
前記初期領域を含む領域であって前記初期領域よりも大きな領域である評価領域の点広がり関数を第2点広がり関数として算出する第2点広がり関数算出部と、
前記第1点広がり関数と前記第2点広がり関数とが類似するか否かを判定する類似度判定部と、
前記類似度判定部により前記第1点広がり関数と前記第2点広がり関数とが類似しないと判定された場合、前記領域保持部に最後に保持された保持領域を前記適応的領域として決定することにより、ブレが共通する領域を1つの適応的領域として決定する適応的領域決定部とを備え、
前記領域保持部は、前記類似度判定部により前記第1点広がり関数と前記第2点広がり関数とが類似すると判定された場合、前記評価領域を保持領域として保持し、
前記第2点広がり関数算出部は、前記類似度判定部により前記第1点広がり関数と前記第2点広がり関数とが類似すると判定された場合、前記評価領域を含む領域であって前記評価領域より大きな領域である新たな評価領域の点広がり関数を前記第2点広がり関数として算出する
請求項2に記載の画像補正装置。 - 前記第1点広がり関数算出部は、
前記入力画像の全部又は一部の領域を候補領域として選択する候補領域選択部と、
前記候補領域選択部により選択された候補領域のブレ方向が単一のブレ方向であるか否かを判定するブレ方向判定部と、
前記ブレ方向判定部により単一のブレ方向であると判定された場合、前記候補領域を前記初期領域として決定する初期領域決定部とを備え、
前記候補領域選択部は、前記ブレ方向判定部により単一のブレ方向でないと判定された場合、前記候補領域よりも小さな領域を新たな候補領域として選択する
請求項3に記載の画像補正装置。 - 前記適応的領域分割部は、単一のブレ方向であるか否かに基づいて、ブレが共通する領域を1つの適応的領域として決定する
請求項1に記載の画像補正装置。 - 前記適応的領域分割部は、
前記入力画像の全部又は一部の領域を候補領域として選択する候補領域選択部と、
前記候補領域選択部により選択された候補領域のブレ方向が単一のブレ方向であるか否かを判定するブレ方向判定部と、
前記ブレ方向判定部により単一のブレ方向であると判定された場合、前記候補領域を前記適応的領域として決定することにより、ブレが共通する領域を1つの適応的領域として決定する適応的領域決定部とを備え、
前記候補領域選択部は、前記ブレ方向判定部により単一のブレ方向でないと判定された場合、前記候補領域よりも小さな領域を新たな候補領域として選択する
請求項5に記載の画像補正装置。 - 前記点広がり関数補間部は、算出された前記点広がり関数のうち少なくとも2つの点広がり関数が直線で表現される場合、当該少なくとも2つの点広がり関数のそれぞれを2以上に分割し、分割した点広がり関数を用いて当該少なくとも2つの点広がり関数を互いに対応付けることにより得られる対応関係に基づいて、当該少なくとも2つの点広がり関数にそれぞれ対応する前記適応的領域の代表画素間に位置する画素の点広がり関数を補間する
請求項1に記載の画像補正装置。 - 前記点広がり関数補間部は、算出された前記点広がり関数が互いに類似するか否かを判定し、互いに類似すると判定された点広がり関数にそれぞれ対応する適応的領域の代表画素間に位置する画素の点広がり関数を補間する
請求項7に記載の画像補正装置。 - 前記点広がり関数補間部は、算出された前記点広がり関数を動的計画法によるマッチング手法を用いて互いに対応付けることにより得られる対応関係に基づいて、当該点広がり関数にそれぞれ対応する前記適応的領域の代表画素間に位置する画素の点広がり関数を補間する
請求項1に記載の画像補正装置。 - 前記点広がり関数補間部は、算出された前記点広がり関数が互いに類似するか否かを判定し、互いに類似すると判定された点広がり関数にそれぞれ対応する適応的領域の代表画素間に位置する画素の点広がり関数を補間する
請求項9に記載の画像補正装置。 - 入力画像を補正することにより、前記入力画像よりもブレの少ない目的画像を生成する画像補正方法であって、
前記入力画像を構成する画素の画素値に基づいて、ブレが共通する領域を1つの適応的領域として決定することにより、前記入力画像を適応的領域に分割し、分割した適応的領域ごとに、画像のブレの特徴を示す点広がり関数を算出する適応的領域分割ステップと、
算出された前記点広がり関数を用いて、前記複数の適応的領域のそれぞれを代表する画素である代表画素間に位置する画素の点広がり関数を補間する点広がり関数補間ステップと、
補間後の前記点広がり関数を用いて前記入力画像を補正することにより、前記目的画像を生成する画像補正ステップとを含む
画像補正方法。 - 入力画像を補正することにより、前記入力画像よりもブレの少ない目的画像を生成する集積回路であって、
前記入力画像を構成する画素の画素値に基づいて、ブレが共通する領域を1つの適応的領域として決定することにより、前記入力画像を複数の適応的領域に分割し、分割した適応的領域ごとに、画像のブレの特徴を示す点広がり関数を算出する適応的領域分割部と、
算出された前記点広がり関数を用いて、前記複数の適応的領域のそれぞれを代表する画素である代表画素間に位置する画素の点広がり関数を補間する点広がり関数補間部と、
補間後の前記点広がり関数を用いて前記入力画像を補正することにより、前記目的画像を生成する画像補正部とを備える
集積回路。 - 請求項11に記載の画像補正方法をコンピュータに実行させるためのプログラム。
Priority Applications (4)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
US12/988,890 US8422827B2 (en) | 2009-02-25 | 2010-02-22 | Image correction apparatus and image correction method |
JP2010533363A JP5331816B2 (ja) | 2009-02-25 | 2010-02-22 | 画像補正装置及び画像補正方法 |
CN2010800014220A CN102017607B (zh) | 2009-02-25 | 2010-02-22 | 图像校正装置以及图像校正方法 |
EP10745941.4A EP2403235B1 (en) | 2009-02-25 | 2010-02-22 | Image correction device and image correction method |
Applications Claiming Priority (2)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
JP2009042741 | 2009-02-25 | ||
JP2009-042741 | 2009-02-25 |
Publications (1)
Publication Number | Publication Date |
---|---|
WO2010098054A1 true WO2010098054A1 (ja) | 2010-09-02 |
Family
ID=42665272
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
PCT/JP2010/001115 WO2010098054A1 (ja) | 2009-02-25 | 2010-02-22 | 画像補正装置及び画像補正方法 |
Country Status (5)
Country | Link |
---|---|
US (1) | US8422827B2 (ja) |
EP (1) | EP2403235B1 (ja) |
JP (1) | JP5331816B2 (ja) |
CN (1) | CN102017607B (ja) |
WO (1) | WO2010098054A1 (ja) |
Cited By (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
WO2012101995A1 (en) * | 2011-01-25 | 2012-08-02 | Panasonic Corporation | Object separating apparatus, image restoration apparatus, object separating method and image restoration method |
JP2012216947A (ja) * | 2011-03-31 | 2012-11-08 | Fujifilm Corp | 撮像装置及び画像処理方法 |
WO2013061664A1 (ja) * | 2011-10-28 | 2013-05-02 | 日立アロカメディカル株式会社 | 超音波イメージング装置、超音波イメージング方法および超音波イメージング用プログラム |
WO2014010726A1 (ja) * | 2012-07-12 | 2014-01-16 | 株式会社ニコン | 画像処理装置及び画像処理プログラム |
KR101527656B1 (ko) * | 2013-12-12 | 2015-06-10 | 한국과학기술원 | 의료 영상에서 비강체 영상 정합을 이용한 점상 강도 분포 함수를 보간하는 방법 및 시스템 |
Families Citing this family (13)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US8624986B2 (en) * | 2011-03-31 | 2014-01-07 | Sony Corporation | Motion robust depth estimation using convolution and wavelet transforms |
JP2013005258A (ja) * | 2011-06-17 | 2013-01-07 | Panasonic Corp | ブレ補正装置、ブレ補正方法及び帳票 |
US9124797B2 (en) | 2011-06-28 | 2015-09-01 | Microsoft Technology Licensing, Llc | Image enhancement via lens simulation |
US9137526B2 (en) * | 2012-05-07 | 2015-09-15 | Microsoft Technology Licensing, Llc | Image enhancement via calibrated lens simulation |
CN102789574B (zh) * | 2012-07-02 | 2015-07-29 | 湖北微驾技术有限公司 | 基于三棱镜光学传感器的指纹图像几何矫正算法 |
WO2014074138A1 (en) * | 2012-11-09 | 2014-05-15 | Nikon Corporation | Globally dominant point spread function estimation |
AU2012258467A1 (en) * | 2012-12-03 | 2014-06-19 | Canon Kabushiki Kaisha | Bokeh amplification |
US10165263B2 (en) * | 2013-09-30 | 2018-12-25 | Nikon Corporation | Point spread function estimation of optics blur |
US9684970B2 (en) * | 2015-02-27 | 2017-06-20 | Qualcomm Incorporated | Fast adaptive estimation of motion blur for coherent rendering |
WO2018129692A1 (en) * | 2017-01-12 | 2018-07-19 | Intel Corporation | Image refocusing |
CN110991550B (zh) * | 2019-12-13 | 2023-10-17 | 歌尔科技有限公司 | 一种视频监控方法、装置、电子设备及存储介质 |
CN111476746A (zh) * | 2020-03-19 | 2020-07-31 | 航天信德智图(北京)科技有限公司 | 一种基于ihs变换与自适应区域特征的遥感图像融合方法 |
US20230319401A1 (en) * | 2022-03-31 | 2023-10-05 | Qualcomm Incorporated | Image capture using dynamic lens positions |
Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JP2002112099A (ja) * | 2000-09-28 | 2002-04-12 | Nikon Corp | 画像修復装置および画像修復方法 |
JP2006221347A (ja) * | 2005-02-09 | 2006-08-24 | Tokyo Institute Of Technology | ぶれ情報検出方法 |
JP2008176735A (ja) * | 2007-01-22 | 2008-07-31 | Toshiba Corp | 画像処理装置及び方法 |
Family Cites Families (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JP3189870B2 (ja) * | 1996-12-24 | 2001-07-16 | シャープ株式会社 | 画像処理装置 |
US20010008418A1 (en) | 2000-01-13 | 2001-07-19 | Minolta Co., Ltd. | Image processing apparatus and method |
JP2001197355A (ja) | 2000-01-13 | 2001-07-19 | Minolta Co Ltd | デジタル撮像装置および画像復元方法 |
JPWO2005069216A1 (ja) | 2004-01-15 | 2010-02-25 | 松下電器産業株式会社 | 光学的伝達関数の測定方法、画像復元方法、およびデジタル撮像装置 |
JP2005309560A (ja) * | 2004-04-19 | 2005-11-04 | Fuji Photo Film Co Ltd | 画像処理方法および装置並びにプログラム |
US7991240B2 (en) * | 2007-09-17 | 2011-08-02 | Aptina Imaging Corporation | Methods, systems and apparatuses for modeling optical images |
US8040382B2 (en) * | 2008-01-07 | 2011-10-18 | Dp Technologies, Inc. | Method and apparatus for improving photo image quality |
-
2010
- 2010-02-22 EP EP10745941.4A patent/EP2403235B1/en active Active
- 2010-02-22 US US12/988,890 patent/US8422827B2/en active Active
- 2010-02-22 WO PCT/JP2010/001115 patent/WO2010098054A1/ja active Application Filing
- 2010-02-22 CN CN2010800014220A patent/CN102017607B/zh active Active
- 2010-02-22 JP JP2010533363A patent/JP5331816B2/ja active Active
Patent Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JP2002112099A (ja) * | 2000-09-28 | 2002-04-12 | Nikon Corp | 画像修復装置および画像修復方法 |
JP2006221347A (ja) * | 2005-02-09 | 2006-08-24 | Tokyo Institute Of Technology | ぶれ情報検出方法 |
JP2008176735A (ja) * | 2007-01-22 | 2008-07-31 | Toshiba Corp | 画像処理装置及び方法 |
Non-Patent Citations (1)
Title |
---|
See also references of EP2403235A4 * |
Cited By (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
WO2012101995A1 (en) * | 2011-01-25 | 2012-08-02 | Panasonic Corporation | Object separating apparatus, image restoration apparatus, object separating method and image restoration method |
JP2012216947A (ja) * | 2011-03-31 | 2012-11-08 | Fujifilm Corp | 撮像装置及び画像処理方法 |
WO2013061664A1 (ja) * | 2011-10-28 | 2013-05-02 | 日立アロカメディカル株式会社 | 超音波イメージング装置、超音波イメージング方法および超音波イメージング用プログラム |
JPWO2013061664A1 (ja) * | 2011-10-28 | 2015-04-02 | 日立アロカメディカル株式会社 | 超音波イメージング装置、超音波イメージング方法および超音波イメージング用プログラム |
WO2014010726A1 (ja) * | 2012-07-12 | 2014-01-16 | 株式会社ニコン | 画像処理装置及び画像処理プログラム |
JPWO2014010726A1 (ja) * | 2012-07-12 | 2016-06-23 | 株式会社ニコン | 画像処理装置及び画像処理プログラム |
KR101527656B1 (ko) * | 2013-12-12 | 2015-06-10 | 한국과학기술원 | 의료 영상에서 비강체 영상 정합을 이용한 점상 강도 분포 함수를 보간하는 방법 및 시스템 |
Also Published As
Publication number | Publication date |
---|---|
US20110033132A1 (en) | 2011-02-10 |
CN102017607B (zh) | 2013-12-25 |
EP2403235A4 (en) | 2013-01-16 |
JPWO2010098054A1 (ja) | 2012-08-30 |
CN102017607A (zh) | 2011-04-13 |
EP2403235B1 (en) | 2014-07-09 |
EP2403235A1 (en) | 2012-01-04 |
JP5331816B2 (ja) | 2013-10-30 |
US8422827B2 (en) | 2013-04-16 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
JP5331816B2 (ja) | 画像補正装置及び画像補正方法 | |
EP2574038B1 (en) | Image capturing apparatus, image processing apparatus, image processing method, and image processing program | |
JP5909540B2 (ja) | 画像処理表示装置 | |
JP6271990B2 (ja) | 画像処理装置、画像処理方法 | |
JP5906493B2 (ja) | 画像処理装置、画像処理方法、画像処理プログラム及び記録媒体 | |
JP5499050B2 (ja) | 画像処理装置、撮像装置、及び画像処理方法 | |
JP2008176735A (ja) | 画像処理装置及び方法 | |
JP2011119802A (ja) | 画像処理装置、画像処理方法 | |
JP5388072B2 (ja) | モーションブラー制御装置、方法、及びプログラム | |
JP5184574B2 (ja) | 撮像装置、画像処理装置、および画像処理方法 | |
JP5765893B2 (ja) | 画像処理装置、撮像装置および画像処理プログラム | |
KR20110075366A (ko) | 디지털 촬영 장치, 그 제어 방법, 및 컴퓨터 판독가능 매체 | |
JP7025237B2 (ja) | 画像処理装置およびその制御方法ならびにプログラム | |
JP6486453B2 (ja) | 画像処理装置、画像処理方法、プログラム | |
JP2016184888A (ja) | 画像処理装置、撮像装置、画像処理方法及びコンピュータプログラム | |
JP2007179211A (ja) | 画像処理装置、画像処理方法、およびそのプログラム | |
JP2018180964A (ja) | 画像処理装置及び画像処理方法 | |
JP6245847B2 (ja) | 画像処理装置および画像処理方法 | |
JP2017028583A (ja) | 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 | |
JP2017229025A (ja) | 画像処理装置、画像処理方法、及びプログラム | |
JP2015119428A (ja) | 画像処理方法、画像処理装置、撮像装置、画像処理プログラム、および、記憶媒体 | |
JP6381212B2 (ja) | 撮像装置及びその制御方法 | |
JP6566780B2 (ja) | 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 | |
JP6548409B2 (ja) | 画像処理装置、その制御方法、および制御プログラム、並びに撮像装置 | |
JP6194793B2 (ja) | 画像補正装置、画像補正方法及び画像補正プログラム |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
WWE | Wipo information: entry into national phase |
Ref document number: 201080001422.0 Country of ref document: CN |
|
WWE | Wipo information: entry into national phase |
Ref document number: 2010533363 Country of ref document: JP |
|
121 | Ep: the epo has been informed by wipo that ep was designated in this application |
Ref document number: 10745941 Country of ref document: EP Kind code of ref document: A1 |
|
WWE | Wipo information: entry into national phase |
Ref document number: 12988890 Country of ref document: US |
|
WWE | Wipo information: entry into national phase |
Ref document number: 2010745941 Country of ref document: EP |
|
NENP | Non-entry into the national phase |
Ref country code: DE |