WO2012086173A1 - ステレオ画像処理装置及びステレオ画像処理方法 - Google Patents
ステレオ画像処理装置及びステレオ画像処理方法 Download PDFInfo
- Publication number
- WO2012086173A1 WO2012086173A1 PCT/JP2011/007084 JP2011007084W WO2012086173A1 WO 2012086173 A1 WO2012086173 A1 WO 2012086173A1 JP 2011007084 W JP2011007084 W JP 2011007084W WO 2012086173 A1 WO2012086173 A1 WO 2012086173A1
- Authority
- WO
- WIPO (PCT)
- Prior art keywords
- unit
- image
- reference image
- pixel
- sub
- Prior art date
Links
Images
Classifications
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
- H04N13/00—Stereoscopic video systems; Multi-view video systems; Details thereof
- H04N13/20—Image signal generators
- H04N13/204—Image signal generators using stereoscopic image cameras
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01C—MEASURING DISTANCES, LEVELS OR BEARINGS; SURVEYING; NAVIGATION; GYROSCOPIC INSTRUMENTS; PHOTOGRAMMETRY OR VIDEOGRAMMETRY
- G01C11/00—Photogrammetry or videogrammetry, e.g. stereogrammetry; Photographic surveying
- G01C11/04—Interpretation of pictures
- G01C11/06—Interpretation of pictures by comparison of two or more pictures of the same area
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/50—Depth or shape recovery
- G06T7/55—Depth or shape recovery from multiple images
- G06T7/593—Depth or shape recovery from multiple images from stereo images
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/97—Determining parameters from multiple pictures
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/10—Image acquisition modality
- G06T2207/10004—Still image; Photographic image
- G06T2207/10012—Stereo images
-
- 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
- G06T2207/20021—Dividing image into blocks, subimages or windows
Definitions
- the present invention relates to a stereo image processing apparatus and a stereo image processing method for calculating an image shift caused by parallax from a stereo image (a base image and a reference image) obtained by photographing the same object.
- stereo image processing apparatuses are known. Application of this stereo image processing apparatus to, for example, an apparatus that measures a distance from a vehicle or pedestrian stereo image taken by a vehicle-mounted camera to the vehicle or pedestrian is being considered.
- the camera interval is also reduced, and as a result, the shift of the stereo image is also reduced. Therefore, a stereo image processing apparatus is required to have a highly accurate parallax calculation function.
- a one-dimensional POC (Phase Only Correlation) method has been proposed as a high-precision stereo matching (stereo image processing parallax calculation) method in a stereo image processing apparatus (see, for example, Patent Document 1).
- this one-dimensional POC method first, a partial image (one-dimensional image data string) is cut out from each of the standard image and the reference image using a Hanning window. Next, the cut-out partial standard image and partial reference image are subjected to a one-dimensional Fourier transform to be converted into Fourier image data and then synthesized. Next, the synthesized Fourier image data is subjected to a one-dimensional inverse Fourier transform after the amplitude component is normalized. Thereby, the phase only correlation coefficient is obtained. Then, an image shift amount (parallax) is calculated based on the correlation peak of the phase-only correlation coefficient.
- the conventional technique is highly accurate for an object having a small size of an image area in the baseline length direction (hereinafter referred to as “image area size in the baseline length direction”) occupied on a stereo image, such as a pedestrian located far away.
- image area size in the baseline length direction There is a problem that it is difficult to calculate parallax. This is because when the image region size in the baseline length direction is small, it is necessary to reduce the one-dimensional image data sequence in order to reduce the influence of the image around the background. This is because the accuracy is lowered.
- the present invention has been made in view of the above points, and provides a stereo image processing apparatus and a stereo image processing method capable of calculating parallax with high accuracy even for an object having a small image area size in the baseline length direction.
- the purpose is to provide.
- a stereo image processing apparatus is a stereo image processing apparatus that calculates a shift amount in units of subpixels between a reference image and a reference image that form a stereo image, and the first window function is extracted as a target position.
- Subpixel estimation unit reference image is cut out from the reference image, the first window function or the second window function to be set is applied, and the subpixel estimation unit is applied from the reference image.
- Said second window function formed by shifting the window function comprises a a window function setting means for setting the cut-out means.
- a stereo image processing method is a stereo image processing method for calculating a shift amount in units of subpixels between a reference image and a reference image constituting a stereo image, and the first window function is extracted as a target position.
- Subpixel estimation unit reference image is cut out from the reference image, the first window function or the second window function to be set is applied, and the subpixel estimation unit is applied from the reference image.
- a phase difference between a cutout step of cutting out a reference image, a data string composed of the luminance values of the cut out sub-pixel estimation unit reference image, and a data string consisting of the luminance values of the cut out sub-pixel estimation unit reference image Based on the calculation step for calculating the amount of deviation in subpixel units, and the amount of deviation in subpixel units calculated in the calculation step.
- the present invention it is possible to provide a stereo image processing apparatus and a stereo image processing method capable of calculating parallax with high accuracy even for an object having a small image region size in the baseline length direction.
- FIG. 1 is a block diagram showing a configuration of a stereo image processing apparatus according to Embodiment 1 of the present invention.
- Block diagram showing the configuration of the filter unit Block diagram showing the configuration of the high-precision filter section Flow diagram for explaining the operation of the stereo image processing apparatus Diagram for explaining the processing of the image matching unit Flow chart showing details of sub-pixel unit arithmetic processing Diagram for explaining the concept of subpixel unit arithmetic processing Diagram for explaining the concept of subpixel unit arithmetic processing Diagram used to explain the processing concept of the window shift unit Diagram used to explain the processing concept of the window shift unit
- FIG. 3 is a block diagram showing a configuration of a stereo image processing apparatus according to Embodiment 2 of the present invention. Flow diagram for explaining the operation of the stereo image processing apparatus
- one pixel is one coordinate point, where the horizontal direction of the image is the X axis and the vertical direction of the image is the Y axis.
- FIG. 1 shows a configuration of a stereo image processing apparatus 100 according to Embodiment 1 of the present invention.
- the stereo image processing apparatus 100 includes a stereo image acquisition unit 101, an image matching unit 102, a filter unit 103, peak position detection units 104 and 106, and a high-accuracy filter unit 105.
- the stereo image acquisition unit 101 acquires stereo images captured by two or more imaging systems (that is, cameras).
- a stereo image includes a standard image and a reference image in which the same object is captured by two different imaging systems. Then, the stereo image acquisition unit 101 outputs the acquired stereo image to the image matching unit 102, the filter unit 103, and the high-accuracy filter unit 105.
- the stereo image acquisition unit 101 acquires a stereo image captured by two cameras whose baseline length direction substantially matches the horizontal direction.
- the image matching unit 102 acquires corresponding points at the pixel level of the reference image with respect to the reference point of the reference image. Specifically, the image matching unit 102 performs image matching processing on the standard image and the reference image acquired by the stereo image acquisition unit 101, thereby corresponding points at the pixel level of the reference image with respect to the standard point of the standard image. To get. Then, the image matching unit 102 calculates a “pixel level deviation amount n” between the standard image and the reference image. On the other hand, the unit of “parallax” between the standard image and the reference image obtained by the peak position detection unit 104 and the peak position detection unit 106 described later is a subpixel.
- the deviation between the standard image and the reference image is roughly detected in “pixel units”, and thereafter, the deviation between the standard image and the reference image is detected by the peak position detection unit 104 and the peak position detection unit 106. Are detected in “sub-pixel units”.
- the image matching unit 102 sets a predetermined pixel included in the reference image as a “reference point”, and a peripheral partial image centered on the reference point (hereinafter referred to as “unit reference image”) as the reference image. Cut out from. Further, the image matching unit 102 cuts out a plurality of partial images (hereinafter referred to as “unit reference images”) having the same size as the unit standard image from the reference images. The plurality of unit reference images are cut out from different positions in the reference image. For example, a rectangular window having a predetermined size rectangular window (vertical size: wv pixel, horizontal size: wh pixel) is used to cut out the unit standard image and the standard reference image.
- the window function that defines the rectangular window is hereinafter referred to as a “pixel unit estimation window function”.
- the image matching unit 102 only has to cut out a plurality of unit reference images by changing the cutout position in the baseline length direction. Then, the amount of deviation between the position of the reference point in the reference image and the position of the corresponding point in the reference image is calculated as the above-described amount of deviation n in pixels.
- the image matching unit 102 identifies a unit reference image having the highest matching degree with the unit standard image among the plurality of cut unit reference images.
- One pixel corresponding to the “reference point” in the specified unit reference image becomes the “pixel level corresponding point” in the reference image.
- a SAD Sud of Absolute Differences
- a luminance difference degree is used as an index representing the degree of matching.
- the filter unit 103 receives the reference point and the shift amount n from the image matching unit 102 and the stereo image from the stereo image acquisition unit 101 as inputs.
- the filter unit 103 calculates an antiphase filter coefficient based on the position of the reference image and the reference point, and performs a filtering process on the periphery of the corresponding point at the pixel level of the reference image using the calculated antiphase filter coefficient. .
- FIG. 2 shows the configuration of the filter unit 103.
- the filter unit 103 includes an image cutout unit 402 and a filtering unit 403.
- the image cutout unit 402 cuts out a partial image from the reference image as a sub-pixel estimation unit reference image in the baseline length direction of the stereo image.
- a subpixel unit estimation window function is used to cut out the subpixel estimation unit reference image.
- the window function for subpixel unit estimation for example, a window function of a Hanning window can be used.
- the image cutout unit 402 cuts out a partial image having the same size as the subpixel estimation unit reference image from the reference image as a subpixel estimation unit reference image using the subpixel unit estimation window function.
- the image clipping unit 402 outputs the sub-pixel estimation unit reference image and the sub-pixel estimation unit reference image to the filtering unit 403.
- the image cutout unit 402 determines the cutout position of the image in the reference image so that the reference point is included in the unit reference image for subpixel estimation. In addition, the image cutout unit 402 determines the cutout position of the image so that the pixel-level corresponding point is included in the unit reference image for subpixel estimation.
- the filtering unit 403 calculates an anti-phase filter coefficient obtained by inverting the position of each pixel value in the sub-pixel estimation unit reference image cut out by the image cut-out unit 402 in the front-rear direction. Then, the filtering unit 403 performs a filtering process on the sub-pixel estimation unit reference image using the calculated anti-phase filter coefficient. Then, the filter generation unit 403 outputs the result of the filtering process (hereinafter referred to as “filtering result”) to the peak position detection unit 104.
- the peak position detection unit 104 acquires the relative positional relationship in which the degree of correlation between the sub-pixel estimation unit reference image and the sub-pixel estimation unit reference image is the highest in the filtering result received from the filter unit 103. Then, based on the acquired relative positional relationship, the peak position detection unit 104 has a parallax (deviation amount) between the sub-pixel estimation unit reference image and the corresponding point at the sub-pixel level in the sub-pixel estimation unit reference image with respect to the reference point. ) Is calculated.
- the high-precision filter unit 105 receives the reference point and the shift amount n from the image matching unit 102, the stereo image from the stereo image acquisition unit 101, and the shift amount in sub-pixel units from the peak position detection unit 104 as inputs.
- the high-accuracy filter unit 105 calculates an antiphase filter coefficient based on the subpixel estimation unit reference image, and uses the calculated antiphase filter coefficient to refer to a subpixel estimation unit. Filter the image. That is, as with the filter unit 103, the high-accuracy filter unit 105 first cuts out a sub-pixel estimation unit reference image from the reference image and calculates an antiphase filter coefficient from the sub-pixel estimation unit reference image. That is, the antiphase filter coefficient is calculated based on the unit reference image for subpixel estimation extracted from the reference image.
- the high-accuracy filter unit 105 inputs, from the peak position detection unit 104, the sub-pixel unit estimation window function used in the filter unit 103 to cut out the sub-pixel estimation unit reference image.
- the sub-pixel unit estimation shift window function used to cut out the sub-pixel estimation unit reference image is formed in the high-accuracy filter unit 105 by shifting by the amount of shift of the sub-pixel unit. Then, using the subpixel unit estimation shift window function, the high-accuracy filter unit 105 cuts out the subpixel estimation unit reference image from the reference image, and uses the calculated antiphase filter coefficient to refer to the subpixel estimation unit reference.
- a filtering process is performed on the image, and the filtering result is output to the peak position detection unit 106.
- FIG. 3 shows the configuration of the high-precision filter unit 105.
- the high accuracy filter unit 105 includes a window function shift unit 411, an image cutout unit 412, and a filtering unit 413.
- the window function shift unit 411 shifts the sub-pixel unit estimation window function used for cutting out the sub-pixel estimation unit reference image by the filter unit 103 by the shift amount of the sub-pixel unit calculated by the peak position detection unit 104.
- the sub-pixel unit estimation shift window function used to cut out the sub-pixel estimation unit reference image in the high-precision filter unit 105 is formed.
- the image cutout unit 412 cuts out the subpixel estimation unit reference image including the reference point from the reference image.
- the image cutout unit 412 uses the subpixel unit estimation shift window function formed in the window function shift unit 411 to use corresponding points (that is, points that are shifted by the shift amount n from the same coordinates as the reference point in the reference image). Is extracted from the reference image. That is, the image cutout unit 412 basically has the same function as the image cutout unit 402.
- the filtering unit 413 calculates an anti-phase filter coefficient from the unit reference image for subpixel estimation cut out by the image cutout unit 412. Then, the filtering unit 413 performs filtering processing on the sub-pixel estimation unit reference image using the calculated antiphase filter coefficient, and outputs the filtering result to the peak position detection unit 106. That is, the filtering unit 413 basically has the same function as the filtering unit 403.
- the peak position detection unit 106 detects the peak position in the filtering result received from the high-accuracy filter unit 105, thereby calculating a shift amount in subpixel units between the standard image and the reference image.
- the peak is a position where the filtering result is the maximum value.
- one pixel is one coordinate point with the horizontal direction of the image as the X axis and the vertical direction of the image as the Y axis.
- the epipolar line direction baseline length direction
- the epipolar line direction is assumed to be parallel to the X axis in the entire image.
- FIG. 4 is a flowchart for explaining the operation of the stereo image processing apparatus 100.
- the stereo image processing apparatus 100 sequentially moves the reference point in the entire reference image, and performs the following operations in steps S1 to S9. This is performed for all the pixels in the distance measurement target area.
- step S ⁇ b> 1 the image matching unit 102 determines the position of a reference point to be analyzed in the distance measurement target region (hereinafter referred to as “analysis reference position”).
- the image matching unit 102 cuts out a unit reference image from the reference image received from the stereo image acquisition unit 101.
- the unit reference image is an image (image region) of a partial region based on the analysis reference position (that is, the reference point) determined in step S1.
- the unit of the size of the unit reference image is a pixel. That is, the unit reference image is an image of a plurality of pixel rows (pixel rows) ⁇ a plurality of pixel columns (pixel columns).
- step S3 the image matching unit 102 determines a search range and a search start position in the reference image based on the analysis reference position determined in step S1.
- the parallax of the stereo image is determined from the baseline length, which is the distance between the cameras, the focal length of the lens, and the distance from the stereo camera to the object. Therefore, the image matching unit 102 may determine the search range based on the distance from the stereo camera to the object to be measured.
- the image matching unit 102 sets the same coordinates as the reference point in the reference image as the search start position in the reference image. do it.
- step S4 the image matching unit 102 determines a cutout reference position, and cuts out a unit reference image having the same size as the unit reference image from the reference image.
- the image matching unit 102 determines the search start position determined in step S3 as, for example, the first cutout reference position, and thereafter moves the cutout reference position.
- step S5 the image matching unit 102 calculates a matching degree between the unit standard image and the unit reference image.
- this matching degree for example, an SAD value or a luminance similarity indicating a luminance dissimilarity is used.
- step S6 the image matching unit 102 performs a process end determination process for the search range. That is, the image matching unit 102 determines whether or not the search range is covered by the movement of the cutout reference position. If it is determined here that the search range has not ended (step S6: No), the image matching unit 102 returns to step S4. As a result, the image matching unit 102 moves the cutout reference position within the search range so that the cutout region of the unit reference image in step S4 is shifted. In this way, the processing from steps S4 to S6 is repeated until the search range ends (step S6: Yes).
- step S7 the image matching unit 102 specifies the position of the unit reference image having the maximum matching degree based on the plurality of matching degrees obtained by the processes in steps S4 to S6.
- the image matching unit 102 detects a unit reference image having the minimum or minimum luminance difference.
- step S2 to step S7 will be specifically described with reference to FIG.
- the image matching unit 102 cuts out a peripheral partial image centered on the analysis reference position (reference point) (xa, ya) as a unit reference image in step S ⁇ b> 2 described above.
- a rectangular window having a predetermined size vertical size: wv pixel, horizontal size: wh pixel
- the center of the rectangular window defined by the pixel unit estimation window function and the analysis reference position are described as being coincident with each other. It suffices if a reference position exists.
- the image matching unit 102 determines a search range and a search start position in the reference image based on the analysis reference position determined in step S1. For example, the same coordinates (xa, ya) as the analysis reference position in the reference image are used as the search start position (initial coordinates for cutting out the unit reference image in the reference image).
- the image matching unit 102 cuts out a peripheral partial image centered on the search start position as a unit reference image from the reference image in step S4 while sequentially shifting the cutout reference position. The cutout reference position is shifted by one pixel here.
- the same rectangular window as the rectangular window used for extracting the unit standard image is used. That is, a rectangular window defined by the same pixel unit estimation window function is used to cut out the unit standard image and the unit reference image.
- the image matching unit 102 calculates the matching degree between the unit standard image and each unit reference image in step S5 described above.
- this matching degree for example, an SAD value indicating the luminance difference degree is used. This SAD value is calculated by the following equation (1).
- step S6 if it is determined in step S6 that the search range has not ended, the image matching unit 102 shifts the position to be extracted and returns to step S4 to refer to a new unit reference. Cut an image from a reference image.
- the position to be cut out is shifted by one pixel.
- the shifting direction is the direction of the right arrow extending from the coordinates (xa, ya) of the reference image in FIG.
- the image matching unit 102 calculates the SAD value of each of the plurality of unit reference images for one unit standard image. Then, in step S7, the image matching unit 102 specifies a unit reference image having the maximum matching degree based on the plurality of matching degrees obtained by the process of step S5. Specifically, the image matching unit 102 specifies, for example, a unit reference image corresponding to the smallest SAD value among the plurality of SAD values. Assuming that the coordinates of the cut-out reference position of the specified unit reference image are (xa + n, ya), n is a shift amount in pixel units.
- the image matching unit 102 sets the cutout reference position where the SAD value is minimum as the corresponding point at the pixel level of the analysis reference position (reference point).
- the SAD value is used as an index of matching degree.
- the present invention is not limited to this, and any one that can be used as an index of matching degree can be used.
- the image matching unit 102 may use SSD (Sum of Squared Differences) as an index of matching degree.
- step S8 of FIG. 4 the filter unit 103 and the peak position detection unit 104 perform sub-pixel units based on the corresponding points at the pixel level obtained in step S7 and the reference image and reference image received from the stereo image acquisition unit 101. Perform arithmetic processing.
- FIG. 6 is a flowchart showing details of the subpixel unit arithmetic processing.
- FIG. 7 is a diagram for explaining the concept of subpixel unit arithmetic processing.
- step S12 the image cutout unit 402 cuts out a unit reference image for subpixel estimation from the unit reference image.
- the unit of the size of the sub-pixel estimation unit reference image is a pixel.
- the position where the unit reference image for subpixel estimation is cut out is the direction in the X-axis direction parallel to the epipolar line from the position of the reference point.
- a subpixel unit estimation window function is used to cut out the subpixel estimation unit reference image.
- the window function for subpixel unit estimation for example, the window function w (m) of the Hanning window represented by the equation (2) is used.
- the window function includes a Hamming window, a Blackman window, A Kaiser window or the like may be used.
- These window functions are selected according to which of the characteristics of the unit reference image for subpixel estimation (for example, frequency power characteristics, phase characteristics, cut-out edge continuity) is regarded as important. For example, when importance is attached to the phase characteristics, the Kaiser window is suitable. However, using a Kaiser window complicates the computation considerably. On the other hand, the Hanning window is suitable from the viewpoint of reducing the amount of calculation.
- the cutout image does not include noise. This is for accurately obtaining the amount of deviation in sub-pixel units.
- the image cut-out process in the image matching unit 102 is performed in units of pixels, the emphasis is on reducing the number of operations rather than accuracy. For this reason, a window function that simply cuts out image data is used as the pixel unit estimation window function used in the image matching unit 102.
- the subpixel unit estimation window function used in the subpixel estimation unit reference image cut-out process emphasizes that there is less noise. It is preferably a function where the changes at both ends are continuous (ie a function where the first and last values of one period are zero).
- the continuity of the subpixel estimation unit reference image signal is maintained, and noise components due to clipping included in the characteristics of the antiphase filter described later can be reduced. it can.
- the pixel unit estimation window function is more likely to have a main lobe than the subpixel unit estimation window function. ) Is narrow and the amplitude of the side lobe is large.
- the sub-pixel unit estimation window function w (m) uses a Hanning window having a size of 1 pixel on the vertical axis and “KJ” pixels on the horizontal axis.
- M is an integer from J to K.
- the subpixel unit estimation window function w (m) is set around the reference point (xa, ya). As a result, an image having a vertical axis size of 1 pixel and a horizontal axis size of “KJ” pixels centered on the reference point (xa, ya) is cut out as a unit reference image for subpixel estimation.
- F ′ (m) represents the luminance value of the sub-pixel estimation unit reference image.
- step S13 the image cutout unit 402 cuts out a unit reference image for subpixel estimation from the unit reference image having the maximum matching degree with the unit reference image detected in step S7.
- the same subpixel unit estimation window function as that in the case of the subpixel estimation unit reference image is also used for the clipping process of the subpixel estimation unit reference image.
- the subpixel unit estimation window function is set to the corresponding point (xa + n, ya).
- g ′ (m) represents the luminance value of the unit reference image for subpixel estimation.
- the sub-pixel unit estimation window function w (m) having the size of 1 pixel on the vertical axis and “KJ” on the horizontal axis is used.
- This size is an example. Yes, it is not limited to this.
- the vertical size is 3 pixels
- the above processing may be performed for each pixel, and the obtained results may be averaged.
- the vertical size includes a plurality of pixels
- the above-described processing may be performed for each vertical pixel, and the results for each of the plurality of pixels included in the vertical size may be weighted and averaged.
- the weighting coefficient used in this case may be determined by a window function like a two-dimensional POC.
- the image cutout unit 402 When the vertical size of the sub-pixel estimation unit reference image and the sub-pixel estimation unit reference image is 2 pixels or more, the image cutout unit 402 performs the above-described averaging and the like, and performs the sub-pixel estimation unit standard. Each of the image and the unit reference image for subpixel estimation is converted into a one-dimensional data string.
- step S14 the filtering unit 403 calculates anti-phase filter coefficients from the sub-pixel estimation unit reference image. Specifically, the filter unit 103 rearranges the data sequence in the reverse order so as to invert the position of each pixel of the constituent data sequence of the sub-pixel estimation unit reference image.
- step S15 the filtering unit 403 filters the sub-pixel estimation unit reference image using the antiphase filter coefficient calculated in step S14, and outputs the filtering result to the peak position detection unit 104.
- the parallax when obtaining the parallax of an object of the same size in real space, the parallax is smaller when the object is far from the stereo camera than when it is nearby, and the shift amount n in units of pixels is also smaller. At the same time, the image area size in the baseline length direction of the object is also reduced.
- the tap length of the anti-phase filter is set according to the size of the shift amount n in units of pixels detected by the image matching unit 102. For example, when the shift amount n per pixel is small, the tap length of the antiphase filter is set to be short accordingly.
- the stereo image processing apparatus 100 adaptively changes the tap length of the antiphase filter by changing the sizes of the sub-pixel estimation unit reference image and the sub-pixel estimation unit reference image with respect to the shift amount n. can do. Thereby, the parallax calculation according to the size of the object can be performed.
- the filtering result is the output of the linear transition invariant system. Therefore, theoretically, the filtering result is the true deviation amount except for the distortion correction error of the lens, the error of gain noise caused by the image sensor such as CCD, and the error of the calculation accuracy of the image cutout by windowing. It is expressing. Therefore, the true peak position in the sub-pixel unit can be obtained by interpolating the numerical value between the pixels according to the sampling theorem for the output of the antiphase filter discretized in the pixel unit.
- step S ⁇ b> 16 the peak position detection unit 104 detects the peak position based on the filtering result, and calculates a shift amount (that is, displacement of the peak position in the X-axis direction) between the base image and the reference image in subpixel units. . Then, the peak position detection unit 104 outputs the calculation result to the high accuracy filter unit 105.
- a shift amount that is, displacement of the peak position in the X-axis direction
- step S ⁇ b> 17 the window function shift unit 411 uses the reference image and the reference image calculated by the peak position detection unit 104 to calculate the sub-pixel unit estimation window function used by the filter unit 103 to cut out the sub-pixel estimation unit reference image.
- a shift window function for subpixel unit estimation is formed by shifting the image by the amount of deviation in subpixel units from the image.
- the subpixel unit estimation shift window function is used by the image cutout unit 412 to cut out the subpixel estimation unit reference image.
- the window function shift unit 411 outputs the sub-pixel unit estimation shift window function to the image cutout unit 412. As a result, the shift window function for subpixel unit estimation is set in the image cutout unit 412.
- step S18 the image cutout unit 412 cuts out the sub-pixel estimation unit reference image from the unit reference image using the sub-pixel unit estimation window function.
- the cut-out process in step S18 is the same as the process in step S12.
- step S19 the image cutout unit 412 uses the sub-pixel unit estimation shift window function to determine the unit reference image for sub-pixel estimation based on the position of the unit reference image having the maximum matching degree specified in step S7. Cut out.
- the clipping process in step S19 is basically the same as the process in step S13, but the window function used when clipping the sub-pixel estimation unit reference image is different. That is, the window functions used in both steps (that is, the subpixel unit estimation window function and the subpixel unit estimation shift window function) are shifted by an interval corresponding to the shift amount detected in step S16.
- step S20 the filtering unit 413 calculates an antiphase filter coefficient from the unit reference image for subpixel estimation.
- the calculation process in step S20 is the same as the process in step S14.
- step S ⁇ b> 21 the filtering unit 413 filters the sub-pixel estimation unit reference image using the antiphase filter coefficient calculated in step S ⁇ b> 20 and outputs the filtering result to the peak position detection unit 106.
- step S ⁇ b> 22 the peak position detection unit 106 detects the peak position based on the filtering result, and calculates a shift amount (that is, displacement of the peak position in the X-axis direction) between the base image and the reference image in subpixel units. .
- FIG. 8 is a diagram for explaining the concept of processing of the high-precision filter unit.
- a Hanning window having a vertical axis of 1 pixel and a horizontal axis of “KJ” pixels is used as the sub-pixel unit estimation window function w (m) for the reference image.
- M is an integer from J to K.
- the subpixel unit estimation window function w (m) is set around the reference point (xa, ya).
- F ′ (m) represents the luminance value of the sub-pixel estimation unit reference image.
- the sub-pixel unit estimation window function w (m) for the reference image has a size of 1 pixel on the vertical axis and “KJ” pixels on the horizontal axis, and a sub-pixel unit estimation window function.
- a Hanning window is used in which w (m) is shifted by the amount of shift in sub-pixel units calculated by the peak position detector 104.
- the sub-pixel unit estimation shift window function w (m) is set around the corresponding point (xa + n, ya).
- an image having a vertical axis size of 1 pixel and a horizontal axis size of “KJ” pixels centered on the corresponding point (xa + n, ya) is extracted as a sub-pixel estimation unit reference image.
- FIG. 9 is a diagram for explaining the concept of processing of the window function shift unit 411.
- FIG. 9A shows the luminance signal of the standard image in the cutout range (for example, 15 pixels) and the luminance signal of the reference image in the cutout range. That is, in FIG. 9A, the luminance values at each pixel of the standard image and reference image for 15 pixels are plotted.
- FIG. 9B shows a sub-pixel unit estimation window function for cutting out the sub-pixel estimation unit reference image and the sub-pixel estimation unit reference image in the filter unit 103.
- FIG. 9C shows the result of multiplying the base image and reference image shown in FIG. 9A by the sub-pixel unit estimation window function shown in FIG. 9B. That is, FIG. 9C shows a luminance signal in which the influence of discontinuity at the end points is reduced by multiplying the sub-pixel unit estimation window function.
- FIG. 9D shows a subpixel unit estimation window function for subpixel estimation reference image segmentation and a subpixel unit estimation shift window function for subpixel estimation reference image segmentation in the high-precision filter unit 105. Indicates.
- the shift amount between the subpixel unit estimation window function and the subpixel unit estimation shift window function is as shown in FIG.
- the result obtained by multiplying the reference image by the subpixel unit estimation window function (subpixel estimation reference image output from the image cutout unit 402) and the reference image are multiplied by the subpixel unit estimation window function. This corresponds to the phase difference with the result (subpixel estimation reference image output by the image cutout unit 402).
- FIG. 9E shows a result of multiplying the base image and the reference image shown in FIG. 9A by the sub-pixel unit estimation window function and the sub-pixel unit estimation shift window function (image cutout unit 402). Are subpixel estimation reference images and subpixel estimation reference images output by the image cutout unit 412).
- FIG. 10 is a diagram for explaining the effect of the processing of the window function shift unit 411.
- the horizontal axis represents the period of the sin function used as the window function
- the vertical axis represents the difference value between the true value of the parallax and the calculated shift amount in subpixel units between the reference image and the reference image. Represents.
- a curve 1001 formed by “ ⁇ ” plotted in FIG. 10 is a subpixel estimation clipped from the reference image by multiplying the subpixel unit estimation window function shown in FIG. 9C. Peak position detection performed based on the pixel data sequence of the unit reference image and the pixel data sequence of the unit reference image for subpixel estimation cut out from the reference image by multiplying the window function for subpixel unit estimation Results are shown.
- a curve 1002 formed by “ ⁇ ” plotted in FIG. 10 is a subpixel cut out from the reference image by multiplying the subpixel unit estimation window function shown in FIG. Peaks based on the pixel data sequence of the unit reference image for estimation and the pixel data sequence of the unit reference image for subpixel estimation extracted from the reference image by multiplying the shift window function for subpixel unit estimation The result of position detection is shown.
- the difference value (error) between the calculated shift amount of the subpixel unit and the true value becomes large.
- the difference value (error) between the calculated shift amount of the sub-pixel unit and the true value is almost zero. In other words, the shift accuracy of the subpixel unit estimation window function can improve the accuracy of the calculated subpixel unit shift amount.
- the standard image and the reference image shown in FIG. 9A are image data sequences that are referred to based on the “deviation amount n” calculated by the image matching unit 102.
- the error is maximum (ie, when the error is 0.5 pixel)
- the same subpixel unit estimation window function as shown in FIG. 9B is used for both the base image and the reference image. (That is, if the shift interval between the window function used for the reference image and the window function used for the reference image is zero), the difference from the true value is 0.5 pixels.
- a shift in units of subpixels that is greatly affected by the phase difference of the window function used for both the image and the reference image being 0 pixels is calculated. For this reason, as shown in FIG. 10, a difference value (error) occurs between the calculated shift in subpixel units and the true value.
- a pair of subpixel unit estimation window functions and subpixel unit estimation shift window functions given a shift interval corresponding to an error between the “deviation amount n” and the true value (see FIG. 9D). ) Can be used to cut out an image corresponding to the error between the “deviation amount n” and the true value.
- the deviation in units of sub-pixels calculated using the image data thus cut out becomes a value closer to the true value, and the difference value (error) from the true value can be greatly reduced as shown in FIG. it can.
- the effect of improving the parallax calculation accuracy by shifting the sub-pixel unit estimation window function increases as the error between the “deviation amount n” in pixel units and the true value increases. Further, the error between the “deviation amount n” in pixel units and the true value depends on the base line length of the camera (distance between stereo cameras), the focal length of the lens, and the distance between the camera and the object.
- the window function shift unit 411 may normalize the coefficient of the sub-pixel unit estimation shift window function after shifting the sub-pixel unit estimation window function. The accuracy of parallax calculation can be further improved by normalizing the coefficients.
- the shift window function for subpixel unit estimation with normalized coefficients is expressed by the following equation, for example.
- step S18 and step S20 in FIG. 6 overlaps with step S12 and step S14, they may be omitted.
- the antiphase filter obtained in steps S12 and S14 may be stored in a memory (not shown), and the stored antiphase filter may be used in step S21. By doing so, the amount of calculation can be reduced.
- step S9 the distance measurement target region end determination process is performed. If there is an unprocessed area in which the processes from step S1 to step S8 have not yet been performed, steps S1 to S8 are performed for the unprocessed area. The process up to is performed.
- the image cutout unit 402 uses the subpixel unit estimation window function from the standard image and the reference image, and uses the subpixel estimation unit standard image and subframe.
- the pixel reference unit reference image is cut out, and the filtering unit 403 calculates the antiphase filter coefficient by inverting the data sequence of the data sequence composed of the luminance values of the sub-pixel estimation unit reference image cut out by the image cutout unit 402. Then, the data sequence composed of the luminance values of the sub-pixel estimation unit reference image is filtered using the anti-phase filter coefficient, and the peak position detection unit 104 determines the sub-pixel based on the peak position in the filtering result of the filtering means. The amount of deviation of the unit is calculated.
- the window function shift unit 411 forms the sub-pixel unit estimation shift window function by shifting the sub-pixel unit estimation window function based on the sub-pixel unit shift amount calculated by the peak position detection unit 104. And set in the image cutout unit 412.
- the image cutout unit 412 cuts out the subpixel estimation unit reference image from the reference image using the subpixel unit estimation window function, and calculates the subpixel unit estimation shift window function set by the window function shift unit 411.
- a unit reference image for subpixel estimation is cut out from the reference image, and the filtering unit 413 reverses the data order of the data sequence composed of the luminance values of the unit reference image for subpixel estimation cut out by the image cutout unit 412.
- the anti-phase filter coefficient is calculated by using the anti-phase filter coefficient to filter the data sequence composed of the luminance values of the sub-pixel estimation unit reference image, and the peak position detection unit 106 determines the peak in the filtering result of the filtering unit. Based on the position, the sub-pixel offset It is calculated.
- FIG. 11 shows a configuration of stereo image processing apparatus 900 according to Embodiment 2 of the present invention.
- the stereo image processing apparatus 900 includes a high-precision filter execution determination unit 901 and an output unit 902.
- the high-accuracy filter execution determination unit 901 determines whether or not to execute the processing of the high-accuracy filter unit 105 and the peak position detection unit 106 according to the amount of deviation in subpixel units calculated by the peak position detection unit 104. judge.
- the shift amount in subpixel units calculated by the peak position detection unit 104 takes any value between zero and the maximum value.
- the range from zero to the reference value is set as the “non-execution region”, and the range from the reference value to the maximum value is set as the “execution region”.
- the reference value may be determined by a trade-off between processing time and required parallax accuracy during operation. For example, when the parallax is not required to have a very high accuracy, but it is desired to shorten the processing time, the reference value may be set to a higher value.
- the high-accuracy filter execution determination unit 901 determines whether the value of the shift amount in subpixel units calculated by the peak position detection unit 104 belongs to “non-execution region” or “execution region”. If it is determined that the pixel belongs to the “non-execution region”, the high-accuracy filter execution determination unit 901 outputs the shift amount in subpixel units output from the peak position detection unit 104 to the output unit 902. To do. On the other hand, when it is determined that it belongs to the “execution region”, the high-accuracy filter execution determination unit 901 sends the sub-pixel unit shift amount output from the peak position detection unit 104 to the high-accuracy filter unit 105. Output.
- the output unit 902 outputs the shift amount in sub-pixel units output from the high-precision filter execution determination unit 901 or the peak position detection unit 106.
- FIG. 12 is a flowchart showing details of the subpixel unit arithmetic processing according to the second embodiment. Compared with the flowchart shown in FIG. 6, in the flowchart shown in FIG. 12, step S31 of high-precision filter execution determination is provided.
- step S ⁇ b> 31 the high-accuracy filter execution determination unit 901 causes the high-accuracy filter unit 105 and the peak position detection unit 106 to execute processing according to the amount of subpixel deviation calculated by the peak position detection unit 104. It is determined whether or not.
- step S31 Yes
- the processing from step S17 to step S22 is executed.
- step S31 if it is determined not to be executed (step S31: No), the flow shown in FIG. 12 ends.
- the high-accuracy filter execution determination unit 901 performs the determination based on the shift amount in subpixel units calculated by the peak position detection unit 104, and only when the effect of improving the parallax calculation accuracy is large, The processing of the peak position detection unit 106 can be executed. Thereby, when the improvement effect of the parallax calculation accuracy is small, the high-accuracy filter unit 105 and the peak position detection unit 106 can be omitted, and the amount of calculation can be reduced.
- the image clipping unit 402 and the image clipping unit 412, the filtering unit 403 and the filtering unit 413, and the peak position detection unit 104 and the peak position detection unit 106 are described as separate functional units. It was. However, the present invention is not limited to this, and the image clipping unit 402 and the image clipping 412 are a single functional unit, the filtering unit 403 and the filtering unit 413 are a single functional unit, and the peak position detection unit 104.
- the peak position detection unit 106 may be a single function unit.
- step S17, step S19, step S21, and step S22 is not limited to once, but may be simply repeated a plurality of times, or the subpixel detected in step S22. It may be repeated until the value of the unit deviation amount is equal to or less than the reference value. Thereby, the parallax detection accuracy can be further improved.
- the window function shifted by the shift amount in subpixel units detected in the immediately preceding step S22 is used as the subpixel unit estimation shift window function.
- the image matching unit 102 roughly detects the amount of deviation between the standard image and the reference image in “pixel units”, and then the peak position detection unit 104 sets the amount of deviation to “sub-pixel units”.
- the case where it is detected in detail is described. That is, the case where the amount of deviation is calculated in two stages has been described.
- the present invention is not limited to this, and can be applied to a case where the amount of deviation is directly calculated in “sub-pixel unit” without performing detection in “pixel unit”.
- Each functional block used in the description of each of the above embodiments is typically realized as an LSI which is an integrated circuit. These may be individually made into one chip, or may be made into one chip so as to include a part or all of them. Although referred to as LSI here, it may be referred to as IC, system LSI, super LSI, or ultra LSI depending on the degree of integration.
- the method of circuit integration is not limited to LSI, and implementation with a dedicated circuit or a general-purpose processor is also possible.
- An FPGA Field Programmable Gate Array
- a reconfigurable processor that can reconfigure the connection and setting of circuit cells inside the LSI may be used.
- the stereo image processing apparatus and stereo image processing method of the present invention are useful as those capable of calculating parallax with high accuracy even for an object having a small image area size in the baseline length direction.
- Stereo image processing apparatus 101 Stereo image acquisition part 102 Image matching part 103 Filter part 104,106 Peak position detection part 105 High precision filter part 402,412 Image clipping part 403,413 Filtering part 411 Window function shift part 901 High precision Filter execution determination unit 902 output unit
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Multimedia (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Theoretical Computer Science (AREA)
- Radar, Positioning & Navigation (AREA)
- Remote Sensing (AREA)
- Signal Processing (AREA)
- Measurement Of Optical Distance (AREA)
- Image Processing (AREA)
- Image Analysis (AREA)
Abstract
Description
[ステレオ画像処理装置100の構成]
図1は、本発明の実施の形態1に係るステレオ画像処理装置100の構成を示す。図1において、ステレオ画像処理装置100は、ステレオ画像取得部101と、画像マッチング部102と、フィルタ部103と、ピーク位置検出部104,106と、高精度フィルタ部105とを有する。
以上の構成を有するステレオ画像処理装置100の動作について説明する。
ステップS1で、画像マッチング部102は、測距対象領域において分析の対象となる基準点の位置(以下、「分析基準位置」という)を決定する。
ステップS2で、画像マッチング部102は、ステレオ画像取得部101から受け取る基準画像から、単位基準画像を切り出す。単位基準画像は、ステップS1において決定された分析基準位置(つまり、基準点)を基準とした部分領域の画像(画像領域)である。単位基準画像の大きさの単位はピクセルである。つまり、単位基準画像は、複数のピクセル行(画素行)×複数のピクセル列(画素列)の画像である。
ステップS3で、画像マッチング部102は、ステップS1で決定された分析基準位置に基づいて、参照画像におけるサーチ範囲及びサーチ開始位置を決定する。ステレオ画像の視差は、カメラ間の距離である基線長とレンズの焦点距離と、ステレオカメラから対象物までの距離とから、決定される。よって、画像マッチング部102は、ステレオカメラから測距しようとする対象物までの距離に基づいてサーチ範囲を決定すればよい。また、ステレオカメラから無限遠にある対象物は基準画像と参照画像の同じ位置に撮像されるため、画像マッチング部102は、基準画像における基準点と同じ座標を、参照画像におけるサーチ開始位置として設定すればよい。
ステップS4で、画像マッチング部102は、切り出し基準位置を決定し、単位基準画像と同一サイズの単位参照画像を参照画像から切り出す。画像マッチング部102は、ステップS3で決定されたサーチ開始位置を、例えば最初の切り出し基準位置に決定し、以降、切り出し基準位置を移動させていく。
ステップS5で、画像マッチング部102は、単位基準画像と、単位参照画像とのマッチング度を算出する。このマッチング度には、例えば、輝度相違度を示すSAD値又は輝度類似度が用いられる。
ステップS6で、画像マッチング部102は、サーチ範囲に対する処理の終了判定処理を行う。すなわち、画像マッチング部102は、切り出し基準位置の移動によりサーチ範囲を網羅したか否かを判定する。ここでサーチ範囲が終了していないと判定される場合(ステップS6:No)には、画像マッチング部102は、ステップS4へ戻る。この結果、画像マッチング部102は、ステップS4での単位参照画像の切り出し領域がずれるように、サーチ範囲内において切り出し基準位置を移動させる。このようにして、ステップS4~S6までの処理は、サーチ範囲が終了するまで(ステップS6:Yes)繰り返される。
ステップS7で、画像マッチング部102は、ステップS4~S6までの処理によって得られた複数のマッチング度に基づいて、マッチング度が最大となる単位参照画像の位置を特定する。マッチング度として輝度相違度が用いられている場合には、画像マッチング部102は、輝度相違度が極小ないし最小となる単位参照画像を検出する。
図4のステップS8で、フィルタ部103及びピーク位置検出部104は、ステップS7で得られたピクセルレベルの対応点と、ステレオ画像取得部101から受け取る基準画像及び参照画像に基づいて、サブピクセル単位演算処理を行う。
ステップS12で、画像切り出し部402は、単位基準画像からサブピクセル推定用単位基準画像を切り出す。サブピクセル推定用単位基準画像の大きさの単位は、ピクセルである。サブピクセル推定用単位基準画像を切り出す位置は、基準点の位置から、向きはエピポーラ線に沿って平行な、X軸方向の向きとする。
ステップS13で、画像切り出し部402は、ステップS7で検出された、単位基準画像とのマッチング度が最大となる単位参照画像から、サブピクセル推定用単位参照画像を切り出す。サブピクセル推定用単位参照画像の切り出し処理にも、サブピクセル推定用単位基準画像の場合と同じサブピクセル単位推定用窓関数が用いられる。ただし、サブピクセル単位推定用窓関数は、対応点(xa+n、ya)に設定される。これにより、サブピクセル推定用単位参照画像として、対応点(xa+n,ya)を中心に、縦軸サイズが1ピクセルであり横軸サイズが“K-J”ピクセルである画像が切り出される。図7において、g’(m)は、サブピクセル推定用単位参照画像の輝度値を表している。
ステップS14で、フィルタリング部403は、サブピクセル推定用単位基準画像から逆位相フィルタ係数を算出する。具体的には、フィルタ部103は、サブピクセル推定用単位基準画像の構成データ列の各画素の位置を反転させる形で、データ列を逆の順番に並べ換える。
ステップS15で、フィルタリング部403は、サブピクセル推定用単位参照画像をステップS14で算出された逆位相フィルタ係数を用いてフィルタリングし、フィルタリング結果をピーク位置検出部104へ出力する。
ステップS16で、ピーク位置検出部104は、フィルタリング結果に基づいてピーク位置を検出し、基準画像と参照画像とのサブピクセル単位のズレ量(つまり、ピーク位置のX軸方向の変位)を算出する。そして、ピーク位置検出部104は、その算出結果を高精度フィルタ部105へ出力する。
ステップS17で、窓関数シフト部411は、フィルタ部103においてサブピクセル推定用単位参照画像の切り出しに用いたサブピクセル単位推定用窓関数を、ピーク位置検出部104において算出された、基準画像と参照画像とのサブピクセル単位のズレ量分、シフトすることにより、サブピクセル単位推定用シフト窓関数を形成する。サブピクセル単位推定用シフト窓関数は、画像切り出し部412においてサブピクセル推定用単位参照画像の切り出しに用いられる。窓関数シフト部411は、サブピクセル単位推定用シフト窓関数を画像切り出し部412へ出力する。これにより、サブピクセル単位推定用シフト窓関数が画像切り出し部412に設定される。
ステップS18で、画像切り出し部412は、サブピクセル単位推定用窓関数を用いて、単位基準画像からサブピクセル推定用単位基準画像を切り出す。ステップS18での切り出し処理は、ステップS12での処理と同様である。
ステップS19で、画像切り出し部412は、サブピクセル単位推定用シフト窓関数を用いて、ステップS7で特定されたマッチング度が最大となる単位参照画像の位置を基準とするサブピクセル推定用単位参照画像を切り出す。ステップS19での切り出し処理は、ステップS13での処理と基本的には同じであるが、サブピクセル推定用単位参照画像を切り出す際に用いられる窓関数が異なる。すなわち、両ステップで用いられる窓関数(つまり、サブピクセル単位推定用窓関数とサブピクセル単位推定用シフト窓関数)は、ステップS16にて検出されたズレ量に対応する間隔だけずれている。
ステップS20で、フィルタリング部413は、サブピクセル推定用単位基準画像から逆位相フィルタ係数を算出する。ステップS20での算出処理は、ステップS14での処理と同じである。
ステップS21で、フィルタリング部413は、サブピクセル推定用単位参照画像をステップS20で算出された逆位相フィルタ係数を用いてフィルタリングし、フィルタリング結果をピーク位置検出部106へ出力する。
ステップS22で、ピーク位置検出部106は、フィルタリング結果に基づいてピーク位置を検出し、基準画像と参照画像とのサブピクセル単位のズレ量(つまり、ピーク位置のX軸方向の変位)を算出する。
ステップS9では、測距対象領域の終了判定処理が行われ、未だステップS1~ステップS8までの処理が行われていない未処理領域が存在する場合には、その未処理領域についてステップS1~ステップS8までの処理が行われる。
実施の形態2では、高精度フィルタ処理の前に、当該高精度フィルタ処理を実行するか否かの判定を行い、当該判定結果に応じて、高精度フィルタ処理を実行する。
図11は、本発明の実施の形態2に係るステレオ画像処理装置900の構成を示す。図11において、ステレオ画像処理装置900は、高精度フィルタ実行判定部901と、出力部902とを有する。
図12は、実施の形態2に係るサブピクセル単位演算処理の詳細を示すフロー図である。図6に示されるフロー図と比べて、図12に示されるフロー図では、高精度フィルタ実行判定のステップS31が設けられている。
(1)上記各実施の形態では、画像切り出し部402と画像切り出し部412、フィルタリング部403とフィルタリング部413、及びピーク位置検出部104とピーク位置検出部106を、別個の機能部として説明を行った。しかしながら、本発明はこれに限定されるものではなく、画像切り出し部402と画像切り出し412とを単一の機能部、フィルタリング部403とフィルタリング部413とを単一の機能部、ピーク位置検出部104とピーク位置検出部106とを単一の機能部としても良い。
101 ステレオ画像取得部
102 画像マッチング部
103 フィルタ部
104,106 ピーク位置検出部
105 高精度フィルタ部
402,412 画像切り出し部
403,413 フィルタリング部
411 窓関数シフト部
901 高精度フィルタ実行判定部
902 出力部
Claims (4)
- ステレオ画像を構成する基準画像と参照画像とのサブピクセル単位のズレ量を算出するステレオ画像処理装置であって、
第1の窓関数を切り出し対象位置に適用して、前記基準画像からサブピクセル推定用単位基準画像を切り出し、前記第1の窓関数、又は、設定される第2の窓関数を適用して、前記参照画像からサブピクセル推定用単位参照画像を切り出す切り出し手段と、
前記切り出されたサブピクセル推定用単位基準画像の輝度値からなるデータ列と、前記切り出されたサブピクセル推定用単位参照画像の輝度値からなるデータ列との位相差に基づいて、前記サブピクセル単位のズレ量を算出する算出手段と、
前記算出手段によって算出されたサブピクセル単位のズレ量に基づいて前記第1の窓関数をシフトすることにより形成した前記第2の窓関数を、前記切り出し手段に設定する窓関数設定手段と、
を具備するステレオ画像処理装置。 - 前記算出手段は、前記切り出されたサブピクセル推定用単位基準画像の輝度値からなるデータ列のデータ順序を反転させることにより逆位相フィルタ係数を算出し、前記サブピクセル推定用単位参照画像の輝度値からなるデータ列に対して前記逆位相フィルタ係数を用いてフィルタリングし、フィルタリング結果におけるピーク位置に基づいて前記サブピクセル単位のズレ量を算出する、
請求項1に記載のステレオ画像処理装置。 - 前記第1の窓関数によって切り出されたサブピクセル推定用単位基準画像及びサブピクセル推定用単位参照画像に基づいて算出された前記サブピクセル単位のズレ量に応じて、前記窓関数設定手段の処理を実行させるか否かを判定する判定手段、をさらに具備する請求項1に記載のステレオ画像処理装置。
- ステレオ画像を構成する基準画像と参照画像とのサブピクセル単位のズレ量を算出するステレオ画像処理方法であって、
第1の窓関数を切り出し対象位置に適用して、前記基準画像からサブピクセル推定用単位基準画像を切り出し、前記第1の窓関数、又は、設定される第2の窓関数を適用して、前記参照画像からサブピクセル推定用単位参照画像を切り出す切り出しステップと、
前記切り出されたサブピクセル推定用単位基準画像の輝度値からなるデータ列と、前記切り出されたサブピクセル推定用単位参照画像の輝度値からなるデータ列との位相差に基づいて、前記サブピクセル単位のズレ量を算出する算出ステップと、
前記算出ステップによって算出されたサブピクセル単位のズレ量に基づいて前記第1の窓関数をシフトすることにより前記第2の窓関数を形成し設定する窓関数設定ステップと、
を具備するステレオ画像処理方法。
Priority Applications (4)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
EP11851244.1A EP2657901B1 (en) | 2010-12-20 | 2011-12-19 | Stereo image processing apparatus and stereo image processing method |
JP2012549631A JP5891415B2 (ja) | 2010-12-20 | 2011-12-19 | ステレオ画像処理装置 |
CN201180040250.2A CN103080973B (zh) | 2010-12-20 | 2011-12-19 | 立体图像处理装置及立体图像处理方法 |
US13/818,473 US9055284B2 (en) | 2010-12-20 | 2011-12-19 | Stereo image processing apparatus and stereo image processing method |
Applications Claiming Priority (2)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
JP2010283622 | 2010-12-20 | ||
JP2010-283622 | 2010-12-20 |
Publications (1)
Publication Number | Publication Date |
---|---|
WO2012086173A1 true WO2012086173A1 (ja) | 2012-06-28 |
Family
ID=46313465
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
PCT/JP2011/007084 WO2012086173A1 (ja) | 2010-12-20 | 2011-12-19 | ステレオ画像処理装置及びステレオ画像処理方法 |
Country Status (5)
Country | Link |
---|---|
US (1) | US9055284B2 (ja) |
EP (1) | EP2657901B1 (ja) |
JP (1) | JP5891415B2 (ja) |
CN (1) | CN103080973B (ja) |
WO (1) | WO2012086173A1 (ja) |
Cited By (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
WO2014049919A1 (ja) * | 2012-09-27 | 2014-04-03 | パナソニック株式会社 | ステレオ画像処理装置およびステレオ画像処理方法 |
EP2803944A3 (en) * | 2013-05-14 | 2015-04-29 | Ricoh Company, Ltd. | Image Processing Apparatus, Distance Measurement Apparatus, Vehicle-Device Control System, Vehicle, and Image Processing Program |
CN104903915A (zh) * | 2013-01-14 | 2015-09-09 | 罗伯特·博世有限公司 | 用于监视车辆的周围环境的方法和设备以及用于实施紧急制动的方法 |
WO2021256020A1 (ja) * | 2020-06-19 | 2021-12-23 | 日立Astemo株式会社 | ステレオ画像処理装置及び距離測定方法 |
Families Citing this family (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JP6551743B2 (ja) * | 2013-06-05 | 2019-07-31 | ソニー株式会社 | 画像処理装置および画像処理方法 |
JP5999127B2 (ja) * | 2014-03-12 | 2016-09-28 | トヨタ自動車株式会社 | 画像処理装置 |
WO2015181811A1 (en) * | 2014-05-28 | 2015-12-03 | Inuitive Ltd. | A method for stereoscopic reconstruction of three dimensional images |
US9807313B2 (en) * | 2015-01-02 | 2017-10-31 | Intel Corporation | Method and system of increasing integer disparity accuracy for camera images with a diagonal layout |
JP2024011739A (ja) * | 2022-07-15 | 2024-01-25 | キヤノン株式会社 | 距離計測装置の校正方法、距離計測装置及びコンピュータプログラム |
Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JP2008123141A (ja) | 2006-11-09 | 2008-05-29 | Yamatake Corp | 対応点探索方法および3次元位置計測方法 |
WO2009096520A1 (ja) * | 2008-02-01 | 2009-08-06 | Konica Minolta Holdings, Inc. | 対応点探索装置および対応点探索方法 |
WO2010113389A1 (ja) * | 2009-03-31 | 2010-10-07 | パナソニック株式会社 | ステレオ画像処理装置およびステレオ画像処理方法 |
JP2010283622A (ja) | 2009-06-04 | 2010-12-16 | Ricoh Co Ltd | 画像処理デバイス、カラー画像形成装置、画像処理方法、画像処理プログラム及び記録媒体 |
Family Cites Families (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US5684890A (en) * | 1994-02-28 | 1997-11-04 | Nec Corporation | Three-dimensional reference image segmenting method and apparatus |
US20020012459A1 (en) * | 2000-06-22 | 2002-01-31 | Chips Brain Co. Ltd. | Method and apparatus for detecting stereo disparity in sequential parallel processing mode |
US8275195B2 (en) * | 2005-11-30 | 2012-09-25 | Telecom Italia S.P.A. | Method for determining scattered disparity fields in stereo vision |
JP2008140360A (ja) * | 2006-11-02 | 2008-06-19 | Konica Minolta Holdings Inc | 画像照合装置 |
RU2522480C2 (ru) * | 2008-01-31 | 2014-07-20 | Нэшнл Инститьют Фо Байолоджикэл Сайенсес | Растения с измененным ростом и/или развитием и способ их создания |
US8233730B1 (en) * | 2008-04-22 | 2012-07-31 | Marvell International Ltd. | Filter bank based phase correlation architecture for motion estimation |
JP5387905B2 (ja) * | 2009-12-25 | 2014-01-15 | ソニー株式会社 | 画像処理装置および方法、並びにプログラム |
-
2011
- 2011-12-19 EP EP11851244.1A patent/EP2657901B1/en not_active Not-in-force
- 2011-12-19 US US13/818,473 patent/US9055284B2/en not_active Expired - Fee Related
- 2011-12-19 JP JP2012549631A patent/JP5891415B2/ja not_active Expired - Fee Related
- 2011-12-19 WO PCT/JP2011/007084 patent/WO2012086173A1/ja active Application Filing
- 2011-12-19 CN CN201180040250.2A patent/CN103080973B/zh not_active Expired - Fee Related
Patent Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JP2008123141A (ja) | 2006-11-09 | 2008-05-29 | Yamatake Corp | 対応点探索方法および3次元位置計測方法 |
WO2009096520A1 (ja) * | 2008-02-01 | 2009-08-06 | Konica Minolta Holdings, Inc. | 対応点探索装置および対応点探索方法 |
WO2010113389A1 (ja) * | 2009-03-31 | 2010-10-07 | パナソニック株式会社 | ステレオ画像処理装置およびステレオ画像処理方法 |
JP2010283622A (ja) | 2009-06-04 | 2010-12-16 | Ricoh Co Ltd | 画像処理デバイス、カラー画像形成装置、画像処理方法、画像処理プログラム及び記録媒体 |
Cited By (10)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
WO2014049919A1 (ja) * | 2012-09-27 | 2014-04-03 | パナソニック株式会社 | ステレオ画像処理装置およびステレオ画像処理方法 |
CN104685535A (zh) * | 2012-09-27 | 2015-06-03 | 松下知识产权经营株式会社 | 立体图像处理装置和立体图像处理方法 |
JPWO2014049919A1 (ja) * | 2012-09-27 | 2016-08-22 | パナソニックIpマネジメント株式会社 | ステレオ画像処理装置およびステレオ画像処理方法 |
CN104685535B (zh) * | 2012-09-27 | 2017-11-28 | 松下知识产权经营株式会社 | 立体图像处理装置和立体图像处理方法 |
US10009594B2 (en) | 2012-09-27 | 2018-06-26 | Panasonic Intellectual Property Management Co., Ltd. | Stereo image processing device and stereo image processing method |
CN104903915A (zh) * | 2013-01-14 | 2015-09-09 | 罗伯特·博世有限公司 | 用于监视车辆的周围环境的方法和设备以及用于实施紧急制动的方法 |
EP2803944A3 (en) * | 2013-05-14 | 2015-04-29 | Ricoh Company, Ltd. | Image Processing Apparatus, Distance Measurement Apparatus, Vehicle-Device Control System, Vehicle, and Image Processing Program |
WO2021256020A1 (ja) * | 2020-06-19 | 2021-12-23 | 日立Astemo株式会社 | ステレオ画像処理装置及び距離測定方法 |
JP2022001830A (ja) * | 2020-06-19 | 2022-01-06 | 日立Astemo株式会社 | ステレオ画像処理装置及び距離測定方法 |
JP7394709B2 (ja) | 2020-06-19 | 2023-12-08 | 日立Astemo株式会社 | ステレオ画像処理装置及び距離測定方法 |
Also Published As
Publication number | Publication date |
---|---|
JPWO2012086173A1 (ja) | 2014-05-22 |
EP2657901A1 (en) | 2013-10-30 |
JP5891415B2 (ja) | 2016-03-23 |
EP2657901B1 (en) | 2018-07-04 |
US9055284B2 (en) | 2015-06-09 |
CN103080973B (zh) | 2016-06-15 |
CN103080973A (zh) | 2013-05-01 |
US20130147922A1 (en) | 2013-06-13 |
EP2657901A4 (en) | 2014-02-26 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
JP5891415B2 (ja) | ステレオ画像処理装置 | |
JP5567179B2 (ja) | ステレオ画像処理装置およびステレオ画像処理方法 | |
JP5874039B2 (ja) | ステレオ画像処理装置及びステレオ画像処理方法 | |
JP6258209B2 (ja) | ステレオ画像処理装置およびステレオ画像処理方法 | |
EP2597421B1 (en) | Stereo distance measurement apparatus and stereo distance measurement method | |
US9275463B2 (en) | Stereo image processing device and stereo image processing method | |
JP5810314B2 (ja) | ステレオ画像処理装置およびステレオ画像処理方法 | |
US20090304266A1 (en) | Corresponding point searching method and three-dimensional position measuring method | |
CA3206206A1 (en) | Device and method for correspondence analysis in images | |
JP6494402B2 (ja) | 画像処理装置、撮像装置、画像処理方法、プログラム | |
US11475233B2 (en) | Image processing device and image processing method | |
Chen et al. | Recovering depth from a single image using spectral energy of the defocused step edge gradient | |
CN107451992A (zh) | 一种sar图像变化检测的方法与装置 |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
WWE | Wipo information: entry into national phase |
Ref document number: 201180040250.2 Country of ref document: CN |
|
121 | Ep: the epo has been informed by wipo that ep was designated in this application |
Ref document number: 11851244 Country of ref document: EP Kind code of ref document: A1 |
|
WWE | Wipo information: entry into national phase |
Ref document number: 2012549631 Country of ref document: JP |
|
WWE | Wipo information: entry into national phase |
Ref document number: 13818473 Country of ref document: US Ref document number: 2011851244 Country of ref document: EP |
|
NENP | Non-entry into the national phase |
Ref country code: DE |