WO2006068292A1 - 画像処理装置と画像処理方法および画像処理プログラム - Google Patents
画像処理装置と画像処理方法および画像処理プログラム Download PDFInfo
- Publication number
- WO2006068292A1 WO2006068292A1 PCT/JP2005/024000 JP2005024000W WO2006068292A1 WO 2006068292 A1 WO2006068292 A1 WO 2006068292A1 JP 2005024000 W JP2005024000 W JP 2005024000W WO 2006068292 A1 WO2006068292 A1 WO 2006068292A1
- Authority
- WO
- WIPO (PCT)
- Prior art keywords
- processing
- image
- motion
- pixel
- unit
- Prior art date
Links
Classifications
-
- 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
- H04N23/681—Motion detection
- H04N23/6812—Motion detection based on additional sensors, e.g. acceleration sensors
-
- 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
- H04N23/684—Vibration or motion blur correction performed by controlling the image sensor readout, e.g. by controlling the integration time
- H04N23/6845—Vibration or motion blur correction performed by controlling the image sensor readout, e.g. by controlling the integration time by combination of a plurality of images sequentially taken
Definitions
- Image processing apparatus image processing method, and image processing program
- the present invention relates to an image processing apparatus, an image processing method, and an image processing program. Specifically, the shooting information is detected from the imaging unit, and the full-screen movement direction of the image is detected based on the movement direction of the imaging unit in the shooting information. In addition, a processing region for at least one of the target image and the surrounding image corresponding to the target pixel in the target image to be predicted and a processing coefficient for motion blur removal corresponding to the detected motion direction are set. The pixel value corresponding to the target pixel is generated based on the pixel value of the pixel in the processing region and the processing coefficient.
- the data obtained using this sensor is information obtained by projecting information in the real world (for example, light) and space-time in a lower dimension than the real world.
- the information obtained by projection has distortion caused by projection.
- an object moving in front of a stationary background is imaged with a video camera and converted into an image signal, it is displayed based on the image signal because the real-world information is sampled and converted into data.
- a motion blur that causes a moving object to blur occurs as a distortion caused by projection.
- motion blur cannot be removed accurately unless motion vectors are detected accurately.
- information in the time direction is also used to improve processing accuracy, for example, if the shutter operation is performed and the exposure time is shorter than one frame period, the amount of movement of the object and the amount of blur of the input image Does not necessarily match. For this reason, unless the exposure time ratio, which is the ratio of the exposure time, is detected with high accuracy, the amount of blur in the input image cannot be detected correctly, and motion blur cannot be removed with high accuracy.
- the present invention provides an image processing apparatus, an image processing method, and an image processing program that can perform motion blur removal processing with high accuracy.
- the image processing apparatus is predicted to include an imaging information detection unit that detects imaging information from the imaging unit, and a motion detection unit that detects a full-screen motion direction of the image based on the motion direction of the imaging unit in the imaging information.
- a processing region setting unit that sets at least one processing region of the target image or peripheral image corresponding to the target pixel in the target image, and a processing coefficient for removing motion blur corresponding to the moving direction.
- a pixel value generation unit that generates a pixel value corresponding to the target pixel based on the pixel value of the pixel in the processing region and the processing coefficient.
- An image processing method includes a shooting information detection step of detecting shooting information from an imaging unit, a motion detection step of detecting a full screen movement direction of an image based on a movement direction of the imaging unit in the shooting information, and a prediction
- a processing region setting step for setting at least one of the target image and the peripheral image corresponding to the target pixel in the target image to be processed, and a processing coefficient for removing motion blur corresponding to the motion direction.
- An image processing program includes: a computer that detects shooting information from an imaging unit; and a motion detection step that detects a full-screen motion direction of an image based on a motion direction of the imaging unit in the shooting information.
- a processing region setting step for setting a processing region of at least one of the target image and the peripheral image corresponding to the target pixel in the predicted target image, and for removing motion blur corresponding to the motion direction.
- a processing coefficient setting step for setting a processing coefficient, and a pixel value generation step for generating a pixel value corresponding to the target pixel based on the pixel value of the pixel in the processing area and the processing coefficient.
- the exposure time, the movement of the imaging unit, the gamma value, etc. are detected as shooting information, and the inverse gamma processing of the image supplied from the imaging unit is performed based on the gamma value.
- the motion vector of the image and the exposure time ratio are detected based on the photographing information, and the motion blur amount is detected based on the motion vector and the exposure time ratio. Based on this amount of motion blur, a processing coefficient for motion blur removal is set.
- a processing region is set in at least one of the target image and peripheral image corresponding to the target pixel in the predicted target image, and motion blur removal is performed.
- a pixel value corresponding to the pixel of interest is generated from the processing coefficient for use and the pixel value of the pixel in the processing area.
- the generated pixel value is output after being subjected to gamma processing based on the gamma value of the shooting information.
- shooting information is detected from the imaging unit, and the full-screen movement direction of the image is detected based on the movement direction of the imaging unit in the shooting information.
- a processing region is set in at least one of the target image and the peripheral image corresponding to the target pixel in the predicted target image.
- a processing coefficient for removing motion blur corresponding to the motion direction is set, and a pixel value corresponding to the target pixel is generated based on the pixel value of the pixel in the processing region and the processing coefficient.
- a real world signal can be constructed with high accuracy.
- motion blur can be removed using time-direction information, and real-world signals can be constructed with higher accuracy.
- the size of the processing area is set according to the amount of motion blur
- the direction of the processing area is set according to the direction of motion
- the processing coefficient is set according to the amount of motion blur. It can be performed.
- motion blur removal is performed using an image that has been subjected to inverse gamma processing based on the shooting information, and gamma processing is performed on the generated pixel values. Therefore, even if a gamma-processed image is output from the imaging unit, motion blur can be correctly removed.
- Figure 1 shows the system configuration
- FIG. 2 is a diagram for explaining imaging by the imaging unit.
- 3A and 3B are diagrams for explaining the captured image.
- FIG. 4 is a diagram for explaining the temporal division operation of pixel values.
- FIG. 5 is a diagram for explaining the time direction division operation of the pixel value when the shutter operation is performed.
- FIG. 6 is a functional block diagram of the image processing apparatus.
- FIG. 7 is a diagram showing the configuration of the image processing apparatus when software is used.
- FIG. 8 is a diagram for explaining the processing area.
- FIG. 9 is a diagram for explaining a pixel value calculation method.
- FIG. 10 is a flowchart showing image processing.
- FIG. 11 is a diagram showing a configuration when the processing coefficient is obtained by learning and the blur removal process is performed.
- Figure 12 is a functional block diagram of the learning unit.
- Fig. 13 is a diagram for explaining the operation when discretely selecting prediction taps. o
- Fig. 14 is a diagram for explaining the calculation of bitty.
- FIG. 15 is a flowchart showing the processing coefficient learning process.
- Figure 16 is a flowchart of image processing using class classification.
- Figure 17 is a functional block diagram of the learning unit when classifying.
- Figure 18 is a flowchart showing the processing coefficient learning process using class classification.
- FIG. 19 is a diagram for explaining the processing area.
- FIG. 2 O A and FIG. 20 B are diagrams showing examples of setting processing areas.
- Figure 22 is a diagram for explaining the main terms and processing coefficients.
- FIG. 23 is another functional block diagram of the image processing apparatus.
- Figure 24 is a flowchart showing the image processing when the shirter operation is not performed. ⁇
- Figure 25 is a functional block diagram of the coefficient generator.
- Figure 27 is a diagram for explaining the main terms and processing coefficients when the phase in the spatial direction is matched.
- Figures 28A and 28B are diagrams for explaining the operation when the processing coefficient is obtained by learning.
- FIG. 30 is a functional block diagram of an image processing apparatus that performs motion blur processing of a still image.
- FIG. 1 is a block diagram showing a configuration of a system to which the present invention is applied.
- the imaging unit 10 generates an image signal D Va obtained by imaging the real society and supplies it to the image processing device 20.
- the image processing apparatus 20 extracts information embedded in the supplied image signal D Va of the input image, and generates and outputs an image signal D Vout from which the embedded information is extracted. Note that the image processing apparatus 20 can extract information embedded in the image signal D Va by using various information ET supplied from the outside.
- the imaging unit 10 is composed of a CCD (Charge-Coupled Device) area sensor, which is a solid-state imaging device, or a video power camera equipped with a MOS area sensor, and images the real world.
- CCD Charge-Coupled Device
- the motion object O Bf corresponding to the foreground is in the direction of arrow A between the imaging unit 10 and the object O Bb corresponding to the background.
- the imaging unit 10 images the moving object OB f corresponding to the foreground together with the object OBb corresponding to the background.
- the detection element of the imaging unit 10 converts the input light into charges for a period corresponding to the exposure time, and accumulates the photoelectrically converted charges.
- the amount of charge is roughly proportional to the intensity of the input light and the time that the light is input.
- the detection element adds the electric charge converted from the input light to the already accumulated electric charge. That is, the detection element integrates the input light for a period corresponding to the exposure time, and accumulates an amount of charge corresponding to the integrated light. It can be said that the detection element has an integral effect with respect to time. In this way, photoelectric conversion is performed by the image sensor, and the input light is converted into charges in units of pixels and accumulated in units of exposure time.
- a pixel signal is generated according to the accumulated charge amount, and an image signal having a desired frame rate is generated using the pixel signal and supplied to the image processing apparatus.
- the exposure time of the image sensor is a period in which the light input by the image sensor is converted into electric charge and the electric charge is accumulated in the detection element as described above.
- the image time is It is equal to the interval (for example, one frame period).
- the shatter movement is performed, it is equal to the shatter opening time.
- FIG. 3A and FIG. 3B are diagrams for explaining the captured image indicated by the image signal.
- FIG. 3A shows an image obtained by imaging the moving object OB f corresponding to the moving foreground and the object OBb corresponding to the stationary background. It is assumed that the moving object O B f corresponding to the foreground has moved horizontally in the direction of arrow A.
- Fig. 3B shows the relationship between the image and time at the position of line L (shown by a broken line) extending in the direction of arrow A as shown in Fig. 3A.
- the length of the moving object O Bf in the line L in the moving direction is, for example, 9 pixels, and if 5 pixels move during one exposure time, the front edge and pixel position P 13 that were at pixel position P 21 at the start of the frame period At the rear end, the exposure time ends at pixel positions P 25 and P 17, respectively.
- the exposure time in one frame is equal to one frame period, so the front end is the pixel position P 26 and the rear end is the pixel position P 18 at the start of the next frame period. .
- the background area includes only the background component.
- pixel positions P17 to P21 are foreground regions that include only foreground components.
- the pixel positions P13 to P16 and the pixel positions P22 to P25 are mixed regions in which the background component and the foreground component are mixed.
- the mixed region is classified into a force pad pack ground region where the background component is covered with the foreground as time passes, and an uncovered background region where the background component appears as time passes.
- the mixed area located on the front end side in the advancing direction of the foreground object OBf is the covered ground area
- the mixed area located on the rear end side is the uncovered background area.
- the image signal includes an image including the foreground area, the background area, or the covered pack ground area or the uncovered pack ground area.
- the pixel value on line L is divided in the time direction. Perform as shown in 4.
- pixel values are expanded in the time direction and divided into equal time intervals by the number of virtual divisions.
- the vertical direction corresponds to time, indicating that time passes from top to bottom in the figure.
- the number of virtual divisions is set according to the amount of motion V in the image object time interval. For example, if the amount of motion V in one frame period is 5 pixels as described above, set the virtual division number to “5” corresponding to the amount of motion V. Divide into 5 parts.
- the pixel value Px of the pixel position Px obtained when the object OBb corresponding to the background is imaged is Bx, and the motion object OBf corresponding to the foreground whose length in the line L is 9 pixels is stopped.
- the pixel values obtained for each pixel are F09 (front end side) to F01 (rear end side).
- the pixel value DP 14 of pixel locations P 14 is represented by the formula (1).
- the background component includes 3 virtual division times (frame period / V), Since the foreground component includes two virtual division times, the mixing ratio of the background component to the pixel value is (3/5) Similarly, for example, in pixel f placement P22, the background component includes one virtual division time and the foreground component includes four virtual division times, so the mixture ratio ⁇ is (1/5).
- the exposure time ratio indicating the ratio of the exposure time to one frame period is changed, for example, when the exposure time ratio is “3 da 5”, as shown in FIG.
- the number of virtual divisions is set according to the amount of movement V of the moving object within the image time interval. For example, when the amount of motion V in one frame period is 5 pixels as described above, the number of virtual divisions is set to '5' corresponding to the amount of motion V, and one frame period is set at equal time intervals. Divide into 5 parts. .
- the pixel value DP 14 at the pixel position ⁇ 14 is expressed by Expression (2).
- the background component includes one virtual division time (frame period / V) and the foreground component includes two virtual division times, so the mixing ratio is (1/3).
- the background component includes one virtual division time and the foreground component includes two virtual division times, so the mixing ratio ⁇ is (1Z3).
- the amount of movement within exposure time vs can be calculated by multiplying the exposure time ratio by the amount of movement V within one frame period. .
- the image processing apparatus 20 extracts significant information buried in the image signal and generates an image signal D Vout in which the motion blur of the moving object OBf corresponding to the foreground is reduced.
- FIG. 6 is a functional block diagram of the image processing apparatus. It does not matter whether each function of the image processing device is realized by hardware or software. In other words, the functional blocks in FIG. 6 may be realized by hardware or may be realized by software.
- FIG. 7 shows the configuration of the image processing apparatus 20 when software is used.
- a CPU Central Processing Unit
- ROM Read Only Memory
- a storage unit 208 A program for realizing each function of the processing device is stored.
- a RAM Random Access Memory
- a CPU Random Access Memory
- the CPU 201, ROM 202, and RAM 203 are connected to each other by a bus 204.
- an input / output interface 205 is connected to the CPU 201 via the bus 204.
- the input / output interface 205 is connected to an input unit 206 composed of a keyboard, a mouse, a microphone, and the like, and an output unit 207 composed of a display, a speaker, and the like.
- the CPU 201 executes various processes in response to commands input from the input unit 206. Then, the CPU 201 outputs an image, sound, or the like obtained as a result of the processing to the output unit 207.
- the storage unit 208 connected to the input / output interface 205 is composed of, for example, a hard disk, and stores programs executed by the CPU 201 and various data.
- a communication unit 209 communicates with an external device via the Internet or other networks.
- the communication unit 209 functions as an acquisition unit that captures the output of the sensor.
- the program may be acquired via the communication unit 209 and stored in the storage unit 208.
- the drive 210 connected to the input / output interface 205 is a program recorded on a recording medium when a recording medium such as a magnetic disk, an optical disk, a magneto-optical disk, or a semiconductor memory is loaded. Get data and data. The acquired program and data are transferred to and stored in the storage unit 208 as necessary.
- the shooting information detection unit 31 of the image processing device 20 detects shooting information when the imaging unit 10 generates the image signal DVa.
- the exposure time (shutter speed) when the image capturing unit 10 performs a shot motion is detected, and information (exposure time information) Sh indicating the detected exposure time is used as the shooting information.
- Sh Exposure time information
- gamma processing is performed when the image signal is generated by the image capture unit 10.
- the gamma value used is detected, and information indicating the detected gamma value (gamma value information) s v is used as shooting information.
- an angular velocity sensor using a three-dimensional gyroscope or a three-dimensional acceleration sensor is provided in the imaging unit 10, and The movement of the imaging unit is detected based on the sensor signal from the provided sensor, and information indicating the detected movement of the imaging unit (imaging unit movement information) S cm is used as shooting information. Further, when the shooting direction of the imaging unit 10 is remotely controlled, the imaging unit motion information S cm can be generated based on a control signal supplied from the outside.
- the shooting information detection unit 3 1 converts the generated gamma value information S y into an inverse gamma processing unit 3 2 and a gamma processing unit 40, exposure time information Sh as a motion blur amount detection unit 3 4, an imaging unit as motion information S cm Supply to motion detector 3 3.
- the inverse gamma processing unit 32 performs inverse gamma processing on the image signal D Va supplied from the imaging unit 10 based on the gamma value information S ⁇ supplied from the imaging information detection unit 31.
- a display device for example, a cathode ray tube
- the imaging unit 10 cancels the nonlinear characteristics in advance.
- gamma processing is performed so that the overall characteristics from the imaging unit 10 to the display device are linear. For this reason, the image signal D Va supplied from the imaging unit 10 is a signal corrected so as to cancel the non-linear characteristic?
- Inverse gamma processing of the image signal D Va is performed based on the gamma value used in this correction, and the image signal D Vb before the gamma processing is generated.
- This image signal D Vb is a signal having a linear characteristic with respect to a real-world image. By performing motion blur removal using this image signal D Vb, a real-world image can be estimated more accurately. .
- the motion vector estimation unit 3 3 1 of the motion detection unit 3 3 estimates the motion direction or motion vector of the entire screen based on the image signal D Vb and uses the image motion selection unit 3 3 2 as image motion information v pa. To supply.
- This motion vector estimation unit 3 3 1 finds the motion direction and motion vector for each pixel using, for example, a block matching method and a gradient method, and all the motion directions and motion vectors with the highest frequency are obtained. Use the direction of motion and the motion vector of the screen.
- the image motion selection unit 3 3 2 generates image motion information v pb indicating the full screen motion of the image based on the imaging unit motion information S cm supplied from the imaging information detection unit 3 1.
- the imaging unit movement information S cm indicates the movement of the imaging unit 10, and the image moves in the opposite direction to the movement of the imaging unit 10.
- the motion direction of the image motion information v pb can be determined.
- the motion amount of the image motion information vpb can be determined based on the motion amount of the imaging unit motion information S cm and the angle of view of the imaging unit 10.
- the image motion selection unit 3 3 2 selects either the image motion information v pa or the image motion information vpb, and uses the motion blur amount detection unit 3 4 and the target pixel setting unit 3 5 or processing as the image motion information vp. This is supplied to the region setting unit 3 6 and the pixel value generation unit 3 8.
- the selection of the image motion information v pa, v pb may be made by the user or may be made automatically by the image processing device 20.
- the image motion information v pa and v pb are selected according to the captured image, it is possible to improve the detection accuracy of the full screen motion of the image and remove motion blur well. Become. For example, when the image signal D Va is generated by capturing only a stationary subject while moving the image capturing direction of the image capturing unit 10, the full screen motion can be accurately performed by selecting the image motion information v pb. It can be detected.
- the image motion selection unit 3 3 2 can also select image motion information for each pixel. For example, when a moving subject and a stationary subject are mixed, and the stationary subject part is large, the motion direction estimated by the motion vector estimation unit 3 3 1 and the frequency of motion vector generation The image motion information v pb is selected at a pixel position having a high frequency, and the image motion information v pa is selected at a pixel position having a low occurrence frequency. In this case, it is possible to accurately obtain the movement of the image portion of the stationary subject.
- the image motion selection unit 3 3 2 detects the full screen motion by selecting only one of the image motion information v pa and the image motion information v pb for one image. Further, if the image motion information v pa and the image motion information v pb are selected on a pixel basis, motion detection can be performed on a pixel basis.
- the motion blur amount detection unit 3 4 detects an exposure time ratio, which is a ratio of the exposure time to the image interval, based on the exposure time information Sh from the imaging information detection unit 31. Multiply processing is performed using the image motion information vp supplied from the detection unit 33, and the motion blur amount MB within the exposure time is detected for each pixel, and is sent to the processing region setting unit 36 and the processing coefficient setting unit 37. Supply.
- the target pixel setting unit 35 sets a plurality of processing target images including the target image, and based on the image motion information vp set by the motion detection unit 33, the target processing corresponding to the target pixel of the target image A pixel on the image is detected, and the detected image is set as a target pixel of each processing target image.
- the processing region setting unit 36 supplies the processing region setting unit 36 with image signals of a plurality of processing target images in which the target pixel is set.
- the attention image indicates an image on which motion blur processing is performed in the input image supplied to the image processing apparatus 20.
- the attention pixel indicates a pixel on which motion blur processing is performed in the attention image. Yes.
- an image of (t) frame is a target image
- an image of (t 1 1) frame, (t) frame, and (t + 1) frame is set as a processing target image.
- the pixel on the image of the (t 1 1) frame or the (t + 1) frame corresponding to the target pixel of the target image is set as the target pixel.
- the image signal DVb (t) of the processing target image of the frame is processed by the processing region setting unit 36 (t) Frame tap selection unit 36 2
- the image signal D Vb (t-1) of the target image is set to (t— 1) Frame tap selector 3 6 1, and the image signal D Vb (t + 1) is supplied to the (t + 1) frame tap selection unit 3 6 3 respectively.
- the target image of the target image will be noticed if the (t + 1) frame image is moved back by the image motion information vp.
- the position of the target pixel in the pixel and the (t + 1) frame after the movement is the same, and the processing area setting unit 36 can easily set the processing area.
- the reference image of interest is not limited to the (t) frame image, and a (t ⁇ l) frame or (t + 1) frame image can be used as the image of interest.
- the target pixel is designated from the outside, or the pixels in the target image are automatically designated as the target pixel sequentially. In this way, it is possible to perform motion blur removal processing only for desired pixels, or to automatically perform motion blur removal processing on the entire image of interest.
- a range is specified from the outside and pixels within this range are automatically specified as the pixel of interest, it is possible to perform motion blur removal processing for the desired range. Become. .
- the processing region setting unit 36 sets at least one processing region of the target image and the peripheral image corresponding to the target pixel in the predicted target image.
- FIG. 6 shows a case where a processing area is set for a moving image of 3 ′ frames including the target image and surrounding images. As will be described later, when the image is a still image, a processing region is set for the target image that is a still image.
- the (t 1 1) frame tap selection unit 3 61 of the processing region setting unit 3 6 is based on the target pixel in the processed image supplied from the target pixel setting unit 35, for example, a motion blur amount detection unit 3 4 Set the processing area of the size in the spatial direction according to the motion blur amount MB set in.
- the processing area is set in the motion direction indicated by the image motion information v p supplied from the motion detection unit 33 with a size corresponding to the motion blur amount MB set by the motion blur detection unit 34.
- the pixel value of the processing area is supplied to the blur removal processing unit 3 8 1 of the pixel value generation unit 3 • 8.
- FIG. 8 shows the processing area, and the processing area WA is set to each processing target image with reference to the target pixel Pna indicated by a circle.
- the processing region is set according to the motion blur amount MB set by the »blur amount detection unit 34 using the target pixel Pna in the processing target image supplied from the above as a reference. Further, the pixel value of the processing area is supplied to the blur removal processing units 3 8 2 and 3 8 3. .
- the processing coefficient setting unit 37 is configured by using a memory, and a processing coefficient corresponding to the motion blur amount or the motion blur amount and the motion direction is stored in advance in the memory.
- the processing coefficient setting unit 3 7 sets the motion direction to a specific direction and stores a specific processing coefficient generated according to the motion blur amount, and stores the motion coefficient supplied from the motion blur amount detection unit 3 4.
- the processing coefficient corresponding to the amount of blur MB is read from the memory and supplied to the blur removal processing units 381, 382, 383 to set the processing coefficient.
- the processing coefficient setting unit 37 stores the processing coefficient corresponding to the motion blur amount and the motion direction
- the processing coefficient corresponding to the direction of motion indicated by the image motion information V p supplied from the memory is read from the memory and supplied to the blur removal processing units 3 8 1, 3 8 2, 3 8 3.
- the processing coefficient setting unit 37 when not shown, supplies adjustment information that enables adjustment of motion blur from the outside, based on this adjustment information, processing to be supplied to the pixel value generation unit 38.
- the effect of removing motion blur may be adjusted by switching the coefficient. In this case, for example, even when the motion blur is not optimally performed with the processing coefficient supplied first, it is possible to optimally remove the motion blur by switching the processing coefficient. It is also possible to leave motion blur intentionally by switching the processing coefficient.
- the blur removal processing unit 3 8 1 of the pixel value generation unit 3 8 includes (t 1 1) the pixel value of the processing area supplied from the frame tap selection unit 3 61 and the processing supplied from the processing coefficient setting unit 3 7. Linear combination is performed using the coefficients, and the pixel value after the blur removal processing is generated and supplied to the integration unit 39.
- the blur removal processing unit 3 8 1 is (t 1 1) from the frame type selection unit 3 6 1.
- a new pixel value for processing is generated from the pixel value of the supplied processing region and the image motion information vd, and the pixel value is generated by multiplying the generated pixel value for processing and a specific processing coefficient. To do.
- the processing pixel value is generated as follows. That is, when generating a processing coefficient used for blur removal processing, the motion direction is set to a specific direction, and the processing coefficient obtained at this time is stored in the processing coefficient setting unit 37 as a specific processing coefficient.
- the motion direction indicated by the image motion information vd has an angle difference ⁇ with respect to a specific direction when the processing coefficient is generated, for example, As shown in Fig. 9, when the motion direction set when generating the processing coefficient is the horizontal direction and the motion direction indicated by the image motion information vd has an angle difference ⁇ , the image motion information from the target pixel P na
- the pixel positions corresponding to the processing coefficient in the motion direction indicated by Vd are the positions Pwa and Pwb indicated by triangles.
- the pixel value DPw at the position Pwa is calculated using the pixel values at the peripheral pixel positions Pwl, Pw2, Pw3, and Pw4 from Equation (3). Similarly, the pixel value D Pwb of the position Pwb that is opposite to the position Pwa is calculated. Accordingly, (t 1 1) The frame tap selection unit 36 1 sets the processing area so that a pixel value for processing can be generated. 0
- the blur removal processing unit 3 8 1 is (t 1 1)
- the frame tap selection unit 3 6 1 and this motion blur amount and motion The pixel value is generated by performing a product-sum operation between the pixel value of the processing region set according to the direction and the processing coefficient supplied from the processing coefficient setting unit 37.
- the blur removal processing unit 38 1 uses the pixel value calculated in this way as a pixel value for processing, and the product sum of the pixel value for processing and the processing coefficient supplied from the processing coefficient setting unit 37 An operation is performed to generate a pixel value.
- the blur removal processing unit 38 2 (t) The pixel value of the processing region supplied from the 362 frame tap selection unit and the processing coefficient setting unit 3 7 Linear combination is performed using the processing coefficients, and the pixel value after blur removal processing is generated and supplied to the integration unit 39.
- the blur removal processing unit 3 8 3 performs linear combination using the pixel value of the processing area supplied from the (t + 1) frame tap selection unit 363 and the processing coefficient supplied from the processing coefficient setting unit 3 7. The pixel value after the blur removal process is generated and supplied to the integration unit 39.
- Equation (4) shows the case where the product-sum operation is performed as an example of the linear combination of the pixel value of the processing region and the processing coefficient.
- Equation (4) q ′ represents the pixel value of the pixel from which blur removal has been performed.
- Ci (i represents each pixel in the processing range with an integer value from 1 to n) represents the pixel value of the processing area.
- Di represents a processing coefficient.
- the integration unit 3 9 displays the image after the deblurring supplied from the deblurring processing unit 3 8:!
- the pixel values of the target pixel in the predicted image which is an image from which motion blur is removed, are generated by integrating the prime values and output as an image signal DVout.
- This integration of pixel values shall use statistical processing. For example, simple average, weighted average, median filter, etc. can be used.
- the gamma processing unit 40 performs gamma processing on the image signal DVc supplied from the integration unit 39 using the gamma value information S ⁇ supplied from the imaging information detection unit 31 and outputs the image signal DVout.
- FIG. 10 shows a flowchart when image processing is performed by software.
- the CPU 201 sets a target pixel for removing motion blur in the target image, and proceeds to step ST2.
- the CPU 201 performs inverse gamma processing on the pixel of interest based on the gamma value indicated by the shooting information, and generates an image signal when a real-world image is projected.
- the CPU 201 detects the motion vector of the pixel of interest and proceeds to step ST4.
- the CPU 201 detects the amount of motion blur of the pixel of interest. That is, based on the motion vector detected in step ST3 and the exposure time information indicated by the shooting information, the amount of motion blur at the exposure time of the target pixel is detected.
- step ST5 the CPU 201 sets a processing area. That is, the CPU 201 sets a processing target image based on the target pixel of the target image, and sets the target pixel in the processing target image. Furthermore, using the target pixel of each image to be processed as a reference, the processing region corresponding to the motion blur amount obtained in step ST 4 or the motion direction corresponding to the motion vector detected in step ST 3 and the motion direction obtained in step ST 4 are obtained. Set the processing area according to the amount of motion blur and proceed to step ST6.
- step ST6 the CPU 201 sets a processing coefficient and proceeds to step ST7.
- this processing coefficient setting a specific processing coefficient based on the motion blur amount of the target pixel obtained in step ST4 or the motion blur amount of the target pixel determined in step ST4 and the motion base detected in step ST3. Set the processing coefficient based on the direction of motion according to the kuttle. 4000
- step ST7 the CPU 201 performs blur removal processing on each processing target image using the processing coefficient set in step ST6. That is, the CPU 201 generates a pixel value corresponding to the target pixel based on the pixel value of the processing area set in step ST5 and the processing coefficient set according to the amount of motion and the direction of motion in step ST6. .
- a new pixel value for processing is generated from the pixel value of the processing area, and the pixel of interest is supported based on the pixel value for processing and the specific processing coefficient set according to the amount of motion in step ST6.
- the pixel value to be calculated is calculated.
- step ST8 the CPU 201 calculates the pixel value for each processing target image in step ST7. Pixel value integration processing is performed, and the obtained pixel value is output as the pixel value after the motion blur removal processing of the target pixel, and the process proceeds to step ST9.
- step ST9 the CPU 201 performs gamma processing on the pixel of interest using the gamma value indicated by the shooting information, and outputs the pixel value after the gamma processing as an image signal DVout.
- step ST 10 the CPU 201 determines whether or not the blur removal processing has been completed for all pixels in the processing range. If there is a pixel that has not been subjected to the blur removal processing, the process returns to step ST 1 and the processing range. When the deblurring is completed for all the pixels, the image processing for the target image is terminated.
- the processing accuracy of the motion blur removal process can be increased.
- Figure 11 shows the configuration when the blurring removal process is performed by determining the processing coefficients.
- the learning unit 51 executes a learning process using a still image that is a teacher image and a student image obtained by adding a motion pocket to the still image, and processing coefficients obtained by the learning are processed by the image processing device 20. It is stored in the coefficient setting unit 37.
- the image processing device 20 performs processing on the pixel of interest from an image including motion blur as an input image. An area is set, and calculation processing is performed using the pixel value of this processing area and the processing coefficient corresponding to the motion blur amount stored in the processing coefficient setting unit 37 to generate a pixel value after removing the blur. .
- FIG. 12 is a functional block diagram of the learning unit 51.
- the student image generation unit 5 1 1 of the learning unit 5 1 generates a student image by adding motion blur according to the motion amount V to the teacher image that is an input image, and supplies the student image to the prediction tap extraction unit 5 1 2 To do.
- the prediction tap extraction unit 5 1 2 sets a plurality of pixels as prediction taps for the pixel of interest to be deblurred, extracts pixel values of the prediction taps from the student image, and generates a normal equation expression unit 5 1 Output to 3. Note that the prediction tap corresponds to the above-described processing region when generating a coefficient according to the motion direction and the motion amount.
- the normal equation generation unit 5 1 3 generates a normal equation from the prediction tap pixel value supplied from the prediction tap extraction unit 5 1 2 and the pixel value of the teacher image, and supplies the normal equation to the coefficient determination unit 5 1 4 .
- the coefficient determination unit 5 14 calculates the processing coefficient based on the normal equation supplied from the normal equation generation unit 5 13 and stores it in the processing coefficient setting unit 37.
- the regular equation generation unit 5 1 3 and the coefficient determination unit 5 1 4 will be further described.
- each processing coefficient c is an undetermined coefficient before learning. Learning is performed by inputting pixels of multiple teacher images (still images).
- the pixel value of the m pixels is described as “qk (k is an integer value from 1 to m)”. the following equation (5) Is set.
- Equation (5) can obtain the pixel value q k ′ after blur removal that is substantially equal to the actual pixel value q k without motion blur by performing the calculation on the right side.
- the reason why it is not equal but near is because it includes an error. That is, the pixel value after blur removal, which is the calculation result on the right side, does not exactly match the pixel value of the target pixel in the actual image without motion blur, and includes a predetermined error.
- the processing coefficient di uses the pixel value q k after blur removal as the pixel value without motion blur. It is thought that it is the optimal coefficient for getting close to. Therefore, for example, the optimal processing coefficient di is determined by the method of least squares using m pixel values q k (m is an integer greater than n) collected by learning.
- Equation (6) The normal equation when the processing coefficient di on the right side of Equation (5) is obtained by the least square method can be expressed as Equation (6).
- Equation (6) the processing coefficient di can be determined by solving the normal equation shown in Equation (6). Specifically, if each matrix of the normal equation expressed by Equation (6) is defined as the following Equations (7) to (9), the normal equation is expressed by the following Equation (1 0), It is expressed as follows. m m
- each component of the matrix DMAT is the processing coefficient di to be obtained. Therefore, if the left-side matrix CMAT and the right-side matrix QMAT are determined in Eq. (1 0), the matrix DMAT (that is, the processing coefficient di) can be calculated by matrix solution. Specifically, as shown in Equation (7), each component of the matrix CMAT can be computed if the prediction tap c ik is known. Since the prediction tap c ik is extracted by the prediction tap extraction unit 5 1 2, the normal equation generation unit 5 1 3 uses the prediction taps c ik supplied from the prediction tap extraction unit 5 1 2 to form a matrix. Each component of CMAT can be calculated.
- each component of the matrix QMAT can be calculated if the prediction tap Ciu and the pixel value q k of the still image are known.
- the prediction tap Cik is the same as that included in each component of the matrix CMA ⁇ , and is a pixel of the teacher image with respect to the pixel value (is the pixel of interest (student image pixel) included in the prediction tap c ik.
- the normal equation generation unit 5 1 3 can calculate each component of the matrix QMAT using the prediction tap c ik supplied from the prediction tap extraction unit 5 1 2 and the teacher image. In this way, the normal equation generation unit 5 1 3 calculates the matrices C MAT and tl [J QMAT swords] and supplies the calculation results to the coefficient determination unit 5 1'4.
- the coefficient determination unit 5 14 calculates the processing coefficient di, which is each component of the matrix DMAT of the above equation (8). Specifically, the normal equation of the above equation (1 0) can be transformed into the following equation (1 1).
- each component of the left side matrix DMAT is the processing coefficient di to be obtained.
- each component of the matrix CMAT and the matrix QMAT is supplied from the normal equation generator 5 1 3. Therefore, when each component of the matrix CMAT and the matrix QMAT is supplied from the normal equation generator 5 1 3 to the coefficient determination unit 5 1 4, the coefficient determination unit 5 1 4 performs the matrix operation on the right side of Equation (1 1).
- the matrix DMAT is calculated with, and the calculation result (processing coefficient di) is stored in the processing coefficient setting unit 37.
- the above learning is performed by changing the amount of motion with the direction of motion as a constant direction
- a specific processing coefficient corresponding to the amount of motion blur can be stored.
- the processing coefficient corresponding to the motion blur amount and the motion direction can be stored.
- the prediction tap is selected by using a length proportional to the amount of motion. For example, a pixel included in the range of “3 X motion amount + 9 pixels” in the motion direction with the target pixel as a reference is set as a prediction tap. Also, if the pixels included in the range of “2 X motion amount + 3 pixels” are used as prediction taps, the number of taps can be reduced and the configuration can be simplified. Further, in selecting the prediction tap, the prediction tap may be selected discretely according to the motion amount V.
- the prediction tap is discretely selected according to the motion amount V
- the pixel position P 47 shown in FIG. 13 is used as the target pixel position * and the pixel value F 19 of the eye pixel is obtained
- equation (12) is established.
- the pixel values at pixel positions P44 and P45 are DP 44 and DP 45 , respectively.
- the level of the pixel value F14 is approximately obtained using spatial correlation. For example, if the spatial correlation between pixel position P42 and surrounding pixels is high, the approximation shown in Eq. (14) is established.
- FIG. 14 is a diagram for explaining the calculation of activity.
- the difference between 9 pixels in total of 3 pixels ⁇ 3 pixels centering on the pixel position P42 is the activity AC (P42) at the pixel position P42, and is expressed by the following equation (15).
- pixel position P42-U is a pixel on the line one line above pixel position P42
- pixel position P42-L is a pixel on the line one line below pixel position P42. The other is the same. .
- the pixel position AC42 (P4 2) represented by the pixel position P42 shown in the equation (15)
- the spatial correlation between the pixel P42 and the surrounding pixels is high, and it is shown that the pixel value at the pixel position P42 approximates the pixel value of the surrounding pixels. Conversely, when the activity is high, the spatial correlation between the pixel at pixel position P42 and the surrounding pixels is low, and the pixel value at pixel position P42 is likely to be different from the pixel values at the surrounding pixels. 4000
- Pixels P42, P44, P45, P49, P50, and P52 can also be used as prediction taps.
- FIG. 15 is a flowchart showing the processing coefficient learning process.
- step ST 11 a student image is generated from the teacher image. That is, motion blur is added to the teacher image to generate a student image that is an image with motion blur added.
- step ST12 a prediction tap is extracted from the student image.
- step S T 13 a normal equation is generated from the prediction tap and the pixels of the teacher image.
- step ST 14 the processing coefficient is determined by solving the normal equation generated in step ST 13 and stored in the processing coefficient setting unit 37.
- step ST 15 it is determined whether or not processing has been performed for all pixels. If processing has not been performed for all pixels, the processing from step ST 12 is repeated for new pixels, and all the pixels are processed. When the pixel processing is completed, the learning process is ended. ⁇ JP2005 / 024000
- motion blur removal processing can be executed.
- class classification is performed based on the size of the activity, and the processing area is changed according to the class code that is the class classification result, thereby blurring.
- the accuracy of removal can also be improved.
- the processing area setting unit 36 extracts the class tap for the target pixel, determines the class code from the extracted class tap, sets the processing area for the target pixel according to the class code, and sets the processing area.
- the determined class code is supplied from the processing area setting unit 36 to the processing coefficient setting unit 37, and the processing coefficient corresponding to the motion blur amount MB and the class code is read from the memory and supplied to the pixel value generation unit 38.
- class tap extraction is performed using the pixels used for activity calculation as class taps, as described with reference to FIGS. 13 and 14. Extract. Using the extracted class tap, the activity is obtained as described above, and the class code is detected based on the obtained activity. For example, the activities AC (P42) and AC (P52) are obtained, and when the activity AC (P42) is smaller than the activity AC (P52), the class code is set to ⁇ 2 '' and the class code is set to the processing coefficient. Further, the processing region setting unit 36 sets the processing region corresponding to the target pixel corresponding to the class code.
- the spatial correlation of the pixel position P 42 is higher than the pixel position P 52, so the pixel values of the pixel positions P 42, P 44, P45.
- the spatial correlation of the pixel position P 52 is higher than that of the pixel position P 42, so the pixel values of the pixel positions P 49, P 50, and P 52 are used as the pixel values of the processing region.
- the pixel value generation unit 3 8 is supplied.
- the processing coefficient setting unit 37 stores in advance processing coefficients obtained by learning by a learning unit, which will be described later, according to the class code, and supplies the processing coefficients corresponding to the class code and the motion amount to the pixel value generation unit 38. To do.
- the blur removal processing units 381 to 383 of the pixel value generation unit 38 perform arithmetic processing using the pixel values in the processing area and the processing coefficients supplied from the processing coefficient setting unit 37, and obtain the pixel values after the blur removal processing. Generated and supplied to the integration unit 39.
- the integration unit 39 integrates the pixel values supplied from the blur removal processing units 381 to 383, and outputs the pixel value of the pixel of interest in the predicted image that is an image from which motion blur removal has been performed.
- FIG. 16 shows a flowchart of image processing using class classification.
- the CPU 201 sets a target pixel for removing motion blur in the target image, and proceeds to step ST22.
- the CPU 201 performs inverse gamma processing on the pixel of interest based on the gamma value indicated by the shooting information to generate an image signal when a real-world image is projected.
- the CPU 201 detects a motion vector of the target pixel and proceeds to step ST 24.
- step ST24 the CPU 201 detects the amount of motion blur of the pixel of interest. That is, calculation processing is performed using the exposure time information indicated by the imaging information and the motion vector detected in step ST23, and the amount of motion blur within the exposure time of the pixel of interest is detected.
- step ST25 the CPU 201 extracts class taps according to the motion blur amount obtained in step ST24, and proceeds to step ST26.
- step ST26 the CPU 201 classifies and determines the class code from the class tap.
- step ST27 the CPU 201 sets a processing area. That is, the CPU 201 sets a processing target image based on the target pixel of the target image, and sets the target pixel in the processing target image. Further, a processing area corresponding to the class code, the motion blur amount and the motion direction of the target pixel is set with reference to the target pixel of each processing target image, and the process proceeds to step ST 28.
- step ST 28 the CPU 201 sets the processing coefficient according to the motion blur amount and class code of the target pixel, or the motion direction according to the motion vector detected in step ST 22 and the motion blur amount and class code of the target pixel. Set the corresponding processing coefficient and proceed to step ST29.
- step ST29 the CPU 201 performs blur removal processing on each processing target image. That is, the CPU 201 performs a calculation process of the pixel value of the processing region set in step ST 27 and the processing coefficient set according to the motion amount, the motion direction, and the class code in step ST 28, and the pixel corresponding to the target pixel Generate a value.
- a new pixel value for processing is generated from the pixel value in the processing area, and calculation of this processing pixel value and the specific processing coefficient set according to the motion amount and the class code in step ST 28 is performed.
- the processing is performed, the pixel value from which the blur is removed is calculated for each target pixel of each processing target image, and the process proceeds to Step ST30.
- step ST 30 the CPU 201 performs integration processing of the pixel values calculated for each processing target image in step ST 29, and outputs the obtained pixel value as a pixel value after the motion blur removal processing of the target pixel. Go to ST 3 1.
- step ST31 the CPU 201 performs gamma processing on the pixel of interest using the gamma value indicated by the shooting information, and outputs the pixel value after the gamma processing as an image signal DVout.
- step ST32 the CPU 201 determines whether or not the blur removal process has been completed for all pixels in the processing range. If there is a pixel that has not been subjected to the blur removal process, the process returns to step ST21, and the entire process range is processed. When the blur removal for the pixel is completed, the image processing for the focused image is terminated.
- FIG. 17 is a functional block diagram of the learning unit 52 when classifying.
- the class tap extraction unit 521 of the learning unit 52 selects a class tap as described above, and supplies the selected class tap to the class classification unit 522.
- the class classification unit 522 performs class classification as described above, determines a class code, and supplies it to the prediction tap extraction unit 523 and the normal equation generation unit 524.
- the prediction type extraction unit 523 switches and extracts prediction taps according to the class code, and supplies the extracted taps to the normal equation generation unit 524.
- the normal equation generation unit 524 is basically the same as the normal equation generation unit 513 of FIG. 12, but generates a normal equation for each class code supplied from the class classification unit 522 and generates a coefficient determination unit. Output to 525. Therefore, the coefficient determination unit 525 The coefficient is determined for each code, and the coefficient determined for each class code is stored in the processing coefficient setting unit 37.
- a specific processing coefficient corresponding to the movement amount and the class code is stored in the processing coefficient setting unit 37.
- processing coefficients corresponding to the amount of motion and the direction of motion and the class code are stored in the processing coefficient setting unit 37.
- FIG. 18 is a flowchart showing processing coefficient learning processing using class classification.
- step ST41 a student image is generated from the teacher image.
- motion blur is added to the teacher image to generate a student image that is an image with motion blur added.
- step ST42 class taps are extracted from student images. This cluster extraction is performed in the same manner as step ST 25 described above.
- step ST43 class classification is performed by classifying from the extracted class tap.
- step ST 44 a prediction tap is extracted from the student image corresponding to the class code determined in step ST 43.
- step ST 45 a normal equation is generated for each class code from the prediction tap and the pixels of the teacher image.
- step ST46 the processing coefficient is determined by solving the normal equation generated in step ST45, and stored in the processing coefficient setting unit 37 for each class code.
- step ST47 it is determined whether or not processing has been performed for all pixels. If processing has not been performed for all pixels, the processing from step ST42 is repeated for new pixels, and all pixels are processed. When the above process ends, the learning process ends. In this way, a processing coefficient for removing motion blur according to the amount of motion, the direction of motion, and the class can be generated. In addition, by making the direction of motion constant, specific processing coefficients can be generated according to the amount of motion and class.
- FIG. 19 shows the processing area in the spatial direction.
- the processing area for (2N + 1) pixels is set in the movement direction around the target pixel Pna.
- Fig. 20 B shows an example of setting the processing area. If the direction of the motion vector is horizontal as shown by the arrow B, for example, for the pixel of the motion object OBf that reduces motion blur, Set the processing area WB in the horizontal direction as shown in Fig. 20A. When the direction of the motion vector is diagonal, the processing area WB is set in the corresponding angular direction as shown in FIG. 20B. However, when setting the processing area in an oblique direction, the pixel value corresponding to the pixel position in the processing area is obtained by interpolation or the like.
- Equation (21) is a constraint equation using spatial correlation
- the number of equations is increased compared to the real world variable, and the value of the real world variable is obtained using the least squares method.
- Equation (22) shows the case where the processing area is set as shown in FIG. 21, and is obtained by adding an error generated in each equation to Equation (20) and Equation (21).
- T indicates a transposed matrix
- Equation (25) ( ⁇ ⁇ ⁇ ) A'X (25) where the error sum of squares is given by equation (26), and this error sum of squares is partially differentiated to obtain a partial difference as shown in equation (27). If the differential value is set to 0, then Equation (25) that minimizes the sum of squared errors can be obtained.
- the above-described configuration processing can be applied when the shutter operation is performed or when the shutter operation is not performed, but when the image capturing unit 10 does not perform the shutter operation.
- the pixel value of the target pixel can be obtained from the image of two frames, and the configuration of the image processing apparatus can be further simplified.
- the processing coefficient used for each pixel is determined by the above model formula, etc.
- the processing coefficient has the shape shown in Fig. 22 and a part where the absolute value of the processing coefficient becomes large is generated.
- the part where the absolute value of the processing coefficient becomes large is the main term.
- the position of this main term is the pixel position corresponding to the amount of movement and the amount of movement with reference to the pixel of interest P na.
- the main term MC al is the main term closest to the pixel of interest P na in the direction of motion
- the main term MCbl is the main term closest to the direction of motion in the opposite direction. Yes.
- the main terms M Cal and M Cbl are generated at a position away from the target pixel P na by the movement amount V (1/2) of the movement amount V or in the direction opposite to the movement direction. Also, the main term adjacent to the pixel of interest Pna is generated at a position separated by the amount of motion or in the direction opposite to the direction of motion.
- the position of the main term overlaps the same pixel position. That is, when focusing on the above-mentioned main terms M Cal and M Cbl, the position of the main term MCbl on the (t-1) frame image is the position of the main term MCbl on the (t) frame image. Therefore, the pixel position in the middle of the (t-1) frame and the (t) frame, which is the same as the main terms MCal and Mcbl, corresponds to the target pixel Pna.
- the pixel value of the target pixel P na from which motion blur has been removed is output correctly. it can.
- FIG. 23 a functional block diagram of an image processing apparatus that performs motion blur removal using the main term in the spatio-temporal direction as the output phase of the pixel of interest P na after motion blur removal is shown in the middle of the frame. 2 Shown in 3.
- parts corresponding to those in FIG. 6 are denoted by the same reference numerals, and detailed description thereof is omitted.
- the value generation unit 3 8 a uses the pixel values extracted from two consecutive frames and the processing coefficient supplied from the processing coefficient setting unit 37 to obtain the pixel value from which the motion blur of the target pixel Pna has been removed.
- the pixel value is generated using the middle of the two frames as the output phase.
- the frame tap selection unit 3 6 1 a uses the processing target image supplied from the target pixel setting unit 3 5 to determine the target pixel in the motion object in which motion blur occurs in the processing target image. In order to extract the main terms that mainly contain the components, (t 1
- Frame tap selection unit 3 6 2 a also (t 1 1) Similar to the frame tap selection unit 3 6 la, the motion in the processing target image is obtained from the processing target image supplied from the target pixel setting unit 35. A pixel of interest in a moving object with blurring. Pixels whose spatial position is approximately the same as the pixel of interest in the phase in the middle of the two frames to extract the main term that mainly contains elementary components. At least the value is extracted and supplied to the pixel value generation unit 38.
- the processing coefficient setting unit 3 7 a stores a specific processing coefficient based on the motion blur amount or a processing coefficient based on the motion blur amount and the motion direction, and the image motion supplied from the motion detection unit 3 3. Based on the information vp, a specific processing coefficient based on the motion blur amount indicated by the image motion information vp is supplied to the pixel value generation unit 38a. Alternatively, the processing coefficient based on the motion blur amount and the motion direction indicated by the image motion information V p is supplied to the pixel value generation unit 38 a.
- the pixel value generation unit 3 8 a receives (t 1 1) frame tap selection unit 3 of the processing area setting unit 3 6 6 Pixel value extracted from la and image motion information According to the motion direction indicated by vp 5024000
- a pixel value for processing is generated using the above equation (3).
- a product-sum operation is performed on the generated pixel value for processing and a specific processing coefficient to generate a pixel value.
- (t) a new pixel value for processing is generated according to the pixel value extracted from the frame tap selection unit 362a and the motion direction indicated by the image motion information vp.
- the pixel value is generated by performing a product-sum operation on the generated pixel value for processing and a specific processing coefficient.
- the pixel value generation unit 38 a selects the (t _ 1) frame tap of the processing area setting unit 36 when a specific processing coefficient based on the motion blur amount and the movement direction is supplied from the processing coefficient setting unit 37 a.
- the pixel value is generated by performing a product-sum operation on the pixel value extracted from the unit 361a and the processing coefficient.
- (t) a product-sum operation is performed on the pixel value extracted from the frame tap selection unit 362a and the processing coefficient to generate a pixel value.
- the pixel value generation unit 38a performs simple averaging, weighted addition, median filter processing, and the like using the pixel values of the target pixels of the two frames to generate a pixel value of one target pixel, and the (t) frame and ( t-1)
- the image signal DVc is supplied to the gamma processing unit 40 at an intermediate phase of the frame.
- FIG. 24 is a flowchart showing image processing when the shotter operation is not performed.
- the CPU 201 sets a pixel of interest for which motion blur is to be removed from the image of interest and proceeds to step ST52.
- the CPU 201 performs inverse gamma processing on the pixel of interest based on the gamma value indicated by the shooting information, and generates an image signal when a real-world image is projected.
- the CPU 201 detects the motion vector of the target pixel and proceeds to step ST54. 5024000
- 0 to 11201 sets a processing area. That is, the CPU 201 sets a processing target image based on the target pixel of the target image, and sets the target pixel in this processing target image. Further, using the target pixel of each processing target image as a reference, a processing region corresponding to the motion vector obtained in step ST 53 is set, and the process proceeds to step ST 55. In step ST55, the CPU 201 sets a processing coefficient according to the motion direction or the motion direction and the motion amount of the motion vector obtained in step ST53, and proceeds to step ST56.
- step ST 56 the CPU 201 performs blurring / removal processing on each processing target image using the processing coefficient set in step ST 55. That is, for each image to be processed, a product-sum operation is performed on the pixel value in the processing range set in step ST 54 and the processing coefficient set in step ST 55, and the pixel value after the motion blur removal processing for the target pixel is performed. Generate and proceed to step ST 57.
- step ST57 the CPU 201 performs gamma processing on the target pixel using the gamma value indicated by the shooting information, and outputs the pixel value after the gamma processing as an image signal DVout.
- step ST58 the CPU 201 determines whether or not the blur removal process has been completed for all pixels in the processing range. If there is a pixel that has not been subjected to the blur removal process, the process returns to step ST51. When the blur removal is completed for all pixels, the image processing for the focused image is terminated. In this way, detection of the motion blur amount in step ST4 in FIG. 10 described above is not required, and motion blur removal can be performed with simpler processing.
- the processing coefficient to be stored in the processing coefficient setting unit 37a can be obtained using the above model formula.
- the target pixel in the still image corresponding to the real world variable Pixels in the first area including (Y shown in Equation (22)
- Figure 25 is a functional block diagram of a coefficient generator that generates processing coefficients using model equations.
- the motion setting unit 71 sets the motion direction and the motion amount, and supplies motion information MH indicating the set motion direction and motion amount to the determinant generation unit 7 51.
- the determinant generator 7 51 of the processing coefficient generator 7 5 constructs a model composed of at least two images with motion blur added to the image, and generates two images, for example (t 1 1) frames.
- the model shown in Fig. 19 is constructed and the determinant is generated.
- the constraint formula is added so that Y and the number of formulas match. However, as shown in formula (22), a constraint formula is added and the least square method You can use it to find the processing coefficient using the least squares method.
- the determinant generation unit 751 generates the determinant as described above for each motion amount, and the coefficient determination unit 752 can determine the processing coefficient for each motion amount.
- the central pixel variable Yo is temporally centered in phase as shown in FIG.
- the central pixel variable Yo does not center in time as shown in FIG. Therefore, the number of virtual divisions is doubled, and one frame period is divided into 8 at equal time intervals as shown in Fig. 26B.
- Y 0a and Y lb a value whose phase is temporally centered can be obtained in the same manner as the central pixel variable Yo when the motion amount V is an odd number of pixels.
- the coefficient for the main term when the amount of motion is an odd number of pixels changes greatly in units of two consecutive pixels. For this reason, when the main pixel is used to determine the target pixel Pna, the phase of the target pixel Pna is the center of the two pixels. In other words, the image after the motion blur removal has a phase shifted by 1Z2 pixels from the image of (t 1 1) frame and (t) frame.
- variable Y. a is 1/2 times the variable Y.
- variable Y lb is 1 Roh twice the variable Yi
- the main term is converted into units of three pixels and the processing coefficient is set and stored.
- the pixel value of the target pixel from which motion blur is removed can be output at the same pixel position as the pixels of the two frames.
- the teacher image is moved to generate student images of (t-1) frames and (t) frames for the teacher images.
- a prediction tap based on the main term is constructed for the student image and learning is performed, and a processing coefficient for the prediction tap is obtained.
- the teacher image is an intermediate phase between the student images of the (t 1 1) frame and the (t) frame.
- the position on the student image corresponding to the pixel of interest on the teacher image corresponds to the position of the main term, as is clear from Fig. 22.
- the prediction tap is set with a spread in the horizontal and vertical directions based on the position on the student image corresponding to the target pixel on the teacher image.
- Fig. 28A shows the case of progressive format, and the shaded pixels indicate the pixel of interest.
- FIG. 28B shows the case of an interlace format image signal, as shown in FIG. 28B, for example, a prediction tap is constructed based on the position between (t 1 1) field lines. In this way, the prediction tap is constructed, and the processing coefficient is generated by calculating the normal equation as described above.
- Figures 29A and 29B show specific examples of processing coefficients obtained by learning.
- Figure 29A shows the (t 1 1) field processing coefficient and (t) field processing coefficient when the motion amount is 20 and the motion direction is 0 degrees.
- Figure 29B shows the (t 1 1) field processing coefficient and (t) field processing coefficient when the amount of movement is 20 and the direction of movement is 45 degrees.
- the frame tap selection unit 3 62 2 a extracts at least the pixel value of the pixel at the substantially same position as the spatial position of the pixel of interest. Then, extract the pixel value of the prediction tap shown in Fig. 29A or Fig.
- the pixel value generation unit 38a performs a product-sum operation using the processing coefficient in the 0 degree direction and 45 degree direction as shown by the broken line, and the pixel value of the prediction tap for this processing coefficient, and the pixel value of the target pixel. Is calculated.
- the frame tap selection unit 3 6 2 a uses the pixel values of the prediction taps shown in FIGS.
- pixel values of prediction taps corresponding to two lines straddling the target pixel in the 0 degree direction and 45 degree direction are extracted.
- the pixel value generation unit 3 8 a has a 0 degree direction or 45 degree direction as shown by the broken line.
- the product-sum operation is performed using the processing values equivalent to two lines across the target pixel and the pixel value of the prediction tap. Furthermore, since product-sum operation is performed in two lines, the product-sum operation result is multiplied by 1 Z to obtain the pixel value of the target pixel.
- the processing coefficient used for the product-sum operation is the same as the shape of the processing coefficient of the main term shown in Fig. 22.
- the motion blur removal can be performed with a simple configuration without providing the motion blur amount detection unit 3 4 and the integration unit 3 9.
- the main term in the time direction for example, for multiple frames or multiple fields, it is possible to perform a robust motion blur removal process.
- FIG. 30 shows the configuration of an image processing apparatus that removes motion blur from a still image.
- FIG. 30 as well, parts corresponding to those in FIG. 6 are given the same reference numerals.
- the processing region setting unit 3 6 b sets the processing region based on the target pixel or main term. To do. Further, the processing coefficient setting unit 37 b sets the processing coefficient based on the imaging unit motion information S cm.
- the pixel value generation unit 3 8 b performs a product-sum operation process based on the pixel value of the processing region and the processing coefficient, generates a pixel value after the motion blur removal process, and supplies the pixel value to the gamma processing unit 40.
- motion blur removal processing can be performed by using the imaging unit motion information S cm indicating the motion of the imaging unit 10.
- the generation of processing coefficients when using a still image can be generated by using a tap in the spatial direction and using the above equation (2 2).
- the main term is the position of the pixel of interest by setting the middle of the two frames as the output phase.
- the target pixel does not become the position of the main term.
- the processing coefficient corresponding to the amount of movement and the direction of movement is generated, and a processing coefficient corresponding to the amount of movement and the direction of movement is selected and used.
- the image processing apparatus, the image processing method, and the image processing program according to the present invention are useful for extracting information buried in an image signal obtained by imaging the real society using the imaging unit. It is suitable for obtaining a high-resolution image from which motion blur is removed.
Landscapes
- Engineering & Computer Science (AREA)
- Multimedia (AREA)
- Signal Processing (AREA)
- Studio Devices (AREA)
- Picture Signal Circuits (AREA)
- Image Processing (AREA)
Abstract
Description
Claims
Priority Applications (2)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
JP2006549086A JP4766334B2 (ja) | 2004-12-21 | 2005-12-21 | 画像処理装置と画像処理方法および画像処理プログラム |
US11/793,701 US7750943B2 (en) | 2004-12-21 | 2005-12-21 | Image processing device that removes motion blur from an image and method of removing motion blur from an image |
Applications Claiming Priority (2)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
JP2004369265 | 2004-12-21 | ||
JP2004-369265 | 2004-12-21 |
Publications (1)
Publication Number | Publication Date |
---|---|
WO2006068292A1 true WO2006068292A1 (ja) | 2006-06-29 |
Family
ID=36601877
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
PCT/JP2005/024000 WO2006068292A1 (ja) | 2004-12-21 | 2005-12-21 | 画像処理装置と画像処理方法および画像処理プログラム |
Country Status (3)
Country | Link |
---|---|
US (1) | US7750943B2 (ja) |
JP (1) | JP4766334B2 (ja) |
WO (1) | WO2006068292A1 (ja) |
Cited By (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
EP1883224A2 (en) | 2006-07-28 | 2008-01-30 | Sony Corporation | Image processing apparatus, image processing method, and program |
JP2009194896A (ja) * | 2008-01-18 | 2009-08-27 | Sanyo Electric Co Ltd | 画像処理装置及び方法並びに撮像装置 |
WO2011024832A1 (ja) * | 2009-08-28 | 2011-03-03 | 日本ビクター株式会社 | 映像信号処理装置 |
Families Citing this family (10)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JP4835949B2 (ja) * | 2007-12-21 | 2011-12-14 | ソニー株式会社 | 画像処理装置および方法、学習装置および方法、プログラム、並びに記録媒体 |
US8452124B2 (en) | 2010-04-30 | 2013-05-28 | Honeywell International Inc. | Method and system for detecting motion blur |
US8472735B2 (en) | 2010-09-30 | 2013-06-25 | The Charles Stark Draper Laboratory, Inc. | Attitude estimation with compressive sampling of starfield data |
US8472737B2 (en) | 2010-09-30 | 2013-06-25 | The Charles Stark Draper Laboratory, Inc. | Attitude estimation in compressed domain |
US8472736B2 (en) * | 2010-09-30 | 2013-06-25 | The Charles Stark Draper Laboratory, Inc. | Attitude estimation by reducing noise with dragback |
JP2013021636A (ja) * | 2011-07-14 | 2013-01-31 | Sony Corp | 画像処理装置および方法、学習装置および方法、プログラム、並びに記録媒体 |
US9635276B2 (en) | 2015-06-10 | 2017-04-25 | Microsoft Technology Licensing, Llc | Determination of exposure time for an image frame |
KR102486398B1 (ko) * | 2015-10-14 | 2023-01-10 | 삼성디스플레이 주식회사 | 영상 신호 처리 회로 및 이를 포함하는 표시 장치 |
JP6522539B2 (ja) * | 2016-03-18 | 2019-05-29 | 富士フイルム株式会社 | 内視鏡システム及びその作動方法 |
KR102314647B1 (ko) * | 2017-11-23 | 2021-10-19 | 삼성전자주식회사 | 이미지를 처리하는 방법, 디바이스 및 기록매체 |
Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JPH0963314A (ja) * | 1995-08-28 | 1997-03-07 | Matsushita Electric Works Ltd | 自動追尾照明装置 |
JP2001250119A (ja) * | 1999-12-28 | 2001-09-14 | Sony Corp | 信号処理装置および方法、並びに記録媒体 |
JP2002373336A (ja) * | 2001-06-15 | 2002-12-26 | Sony Corp | 画像処理装置および方法、記録媒体、並びにプログラム |
JP2003233817A (ja) * | 2002-02-08 | 2003-08-22 | Sony Corp | 画像処理装置および方法、記録媒体、並びにプログラム |
Family Cites Families (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US6987530B2 (en) * | 2001-05-29 | 2006-01-17 | Hewlett-Packard Development Company, L.P. | Method for reducing motion blur in a digital image |
JP4596216B2 (ja) * | 2001-06-20 | 2010-12-08 | ソニー株式会社 | 画像処理装置および方法、記録媒体、並びにプログラム |
JP4596224B2 (ja) * | 2001-06-27 | 2010-12-08 | ソニー株式会社 | 画像処理装置および方法、記録媒体、並びにプログラム |
JP3800192B2 (ja) * | 2002-05-21 | 2006-07-26 | カシオ計算機株式会社 | 撮像装置、画像合成方法及び画像合成プログラム |
JP2006086933A (ja) * | 2004-09-17 | 2006-03-30 | Canon Inc | 撮像装置及び制御方法 |
JP2006129236A (ja) * | 2004-10-29 | 2006-05-18 | Sanyo Electric Co Ltd | リンギング除去装置およびリンギング除去プログラムを記録したコンピュータ読み取り可能な記録媒体 |
-
2005
- 2005-12-21 JP JP2006549086A patent/JP4766334B2/ja not_active Expired - Fee Related
- 2005-12-21 WO PCT/JP2005/024000 patent/WO2006068292A1/ja not_active Application Discontinuation
- 2005-12-21 US US11/793,701 patent/US7750943B2/en not_active Expired - Fee Related
Patent Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JPH0963314A (ja) * | 1995-08-28 | 1997-03-07 | Matsushita Electric Works Ltd | 自動追尾照明装置 |
JP2001250119A (ja) * | 1999-12-28 | 2001-09-14 | Sony Corp | 信号処理装置および方法、並びに記録媒体 |
JP2002373336A (ja) * | 2001-06-15 | 2002-12-26 | Sony Corp | 画像処理装置および方法、記録媒体、並びにプログラム |
JP2003233817A (ja) * | 2002-02-08 | 2003-08-22 | Sony Corp | 画像処理装置および方法、記録媒体、並びにプログラム |
Cited By (8)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
EP1883224A2 (en) | 2006-07-28 | 2008-01-30 | Sony Corporation | Image processing apparatus, image processing method, and program |
JP2008033592A (ja) * | 2006-07-28 | 2008-02-14 | Sony Corp | 画像処理装置および画像処理方法、並びにプログラム |
US7932939B2 (en) | 2006-07-28 | 2011-04-26 | Sony Corporation | Apparatus and method for correcting blurred images |
EP1883224A3 (en) * | 2006-07-28 | 2011-12-28 | Sony Corporation | Image processing apparatus, image processing method, and program |
JP2009194896A (ja) * | 2008-01-18 | 2009-08-27 | Sanyo Electric Co Ltd | 画像処理装置及び方法並びに撮像装置 |
WO2011024832A1 (ja) * | 2009-08-28 | 2011-03-03 | 日本ビクター株式会社 | 映像信号処理装置 |
JP2011049941A (ja) * | 2009-08-28 | 2011-03-10 | Victor Co Of Japan Ltd | 映像信号処理装置 |
CN102577346A (zh) * | 2009-08-28 | 2012-07-11 | Jvc建伍株式会社 | 影像信号处理装置 |
Also Published As
Publication number | Publication date |
---|---|
JPWO2006068292A1 (ja) | 2008-06-12 |
JP4766334B2 (ja) | 2011-09-07 |
US20080095400A1 (en) | 2008-04-24 |
US7750943B2 (en) | 2010-07-06 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
WO2006068292A1 (ja) | 画像処理装置と画像処理方法および画像処理プログラム | |
JP4868236B2 (ja) | 画像処理装置と画像処理方法および画像処理プログラム | |
EP2088787B1 (en) | Image picking-up processing device, image picking-up device, image processing method and computer program | |
EP2088788B1 (en) | Image picking-up processing device, image picking-up device, image processing method and computer program | |
US20060192857A1 (en) | Image processing device, image processing method, and program | |
WO2005079062A1 (ja) | 画像処理装置と画像処理方法およびプログラム | |
US7130464B2 (en) | Image processing device | |
JP4867659B2 (ja) | 画像処理装置と学習装置と係数生成装置および方法 | |
WO2006068282A1 (ja) | 画像処理装置と画像処理方法および画像処理プログラム | |
EP3001669B1 (en) | Image processing apparatus, image processing method and program | |
JP4872672B2 (ja) | 学習装置と学習方法および学習プログラム | |
US7409074B2 (en) | Information processing apparatus and system using pixel data |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
AK | Designated states |
Kind code of ref document: A1 Designated state(s): AE AG AL AM AT AU AZ BA BB BG BR BW BY BZ CA CH CN CO CR CU CZ DE DK DM DZ EC EE EG ES FI GB GD GE GH GM HR HU ID IL IN IS JP KE KG KM KN KP KR KZ LC LK LR LS LT LU LV LY MA MD MG MK MN MW MX MZ NA NG NI NO NZ OM PG PH PL PT RO RU SC SD SE SG SK SL SM SY TJ TM TN TR TT TZ UA UG US UZ VC VN YU ZA ZM ZW |
|
AL | Designated countries for regional patents |
Kind code of ref document: A1 Designated state(s): GM KE LS MW MZ NA SD SL SZ TZ UG ZM ZW AM AZ BY KG KZ MD RU TJ TM AT BE BG CH CY CZ DE DK EE ES FI FR GB GR HU IE IS IT LT LU LV MC NL PL PT RO SE SI SK TR BF BJ CF CG CI CM GA GN GQ GW ML MR NE SN TD TG |
|
121 | Ep: the epo has been informed by wipo that ep was designated in this application | ||
WWE | Wipo information: entry into national phase |
Ref document number: 2006549086 Country of ref document: JP |
|
WWE | Wipo information: entry into national phase |
Ref document number: 11793701 Country of ref document: US |
|
NENP | Non-entry into the national phase |
Ref country code: DE |
|
122 | Ep: pct application non-entry in european phase |
Ref document number: 05822274 Country of ref document: EP Kind code of ref document: A1 |
|
WWW | Wipo information: withdrawn in national office |
Ref document number: 5822274 Country of ref document: EP |
|
WWP | Wipo information: published in national office |
Ref document number: 11793701 Country of ref document: US |