WO2014069196A1 - 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 PDF

Info

Publication number
WO2014069196A1
WO2014069196A1 PCT/JP2013/077418 JP2013077418W WO2014069196A1 WO 2014069196 A1 WO2014069196 A1 WO 2014069196A1 JP 2013077418 W JP2013077418 W JP 2013077418W WO 2014069196 A1 WO2014069196 A1 WO 2014069196A1
Authority
WO
WIPO (PCT)
Prior art keywords
block
image
image processing
stereo image
area
Prior art date
Application number
PCT/JP2013/077418
Other languages
English (en)
French (fr)
Japanese (ja)
Inventor
亮 太田
裕史 大塚
寛人 三苫
Original Assignee
日立オートモティブシステムズ株式会社
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by 日立オートモティブシステムズ株式会社 filed Critical 日立オートモティブシステムズ株式会社
Publication of WO2014069196A1 publication Critical patent/WO2014069196A1/ja

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; 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 OR CALCULATING; 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 OR CALCULATING; COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/30—Subject of image; Context of image processing
    • G06T2207/30248—Vehicle exterior or interior
    • G06T2207/30252—Vehicle exterior; Vicinity of vehicle
    • G—PHYSICS
    • G08—SIGNALLING
    • G08G—TRAFFIC CONTROL SYSTEMS
    • G08G1/00—Traffic control systems for road vehicles
    • G08G1/16—Anti-collision systems
    • G08G1/166—Anti-collision systems for active traffic, e.g. moving vehicles, pedestrians, bikes

Definitions

  • the present invention relates to a stereo image processing apparatus and stereo image processing method for measuring the distance to an object according to the parallax of a pair of images (stereo images).
  • Patent Document 1 As described in Japanese Patent Application Laid-Open No. 2001-92969 (Patent Document 1), as a method for calculating the parallax of an object in a stereo image, one image is divided into blocks of a predetermined size, and the image of each block is divided. There is known “block matching” which searches for a corresponding area (block) from the other image. Then, the deviation between the position of the block in one image and the position of the block in the other image is calculated as the parallax of the object of distance measurement.
  • the size of the block into which the image is divided is constant, the following problems may occur. That is, when the area occupied by the object on the image is small, such as when the object for which the parallax is to be calculated is at a distance, an object having different parallax and a background may be mixed in one block. On the other hand, when the area occupied by the object is large, such as when the object for parallax calculation is in the vicinity, the object exists over a plurality of blocks, and for example, a block including no object edge is formed. Sometimes. Even if block matching is performed in such a state, it is difficult to appropriately calculate the parallax of the object, and there is a possibility that the distance to the object can not be accurately measured.
  • Patent Document 2 JP 2012-2683 A (Patent Document 2), a plurality of reduced images with different magnifications are generated from the stereo image, and each reduced image has the same size.
  • a technique has been proposed which performs block matching using blocks and adopts the one with the best evaluation.
  • an object of this invention is to provide the stereo image processing apparatus and the stereo image processing method which reduced the processing load which calculates the parallax of the object in a stereo image.
  • the stereo image processing apparatus includes a pair of cameras for capturing a stereo image, a unit that calculates the parallax of the object by block matching from the stereo image, and a unit that calculates the distance to the object from the parallax of the object. Then, the means for calculating the parallax of the object changes the size of the block used in the block matching according to the progress of the block matching.
  • FIG. 5 is a flowchart illustrating an example of image processing.
  • 5 is a flowchart illustrating an example of image processing.
  • FIG. 6 is an explanatory diagram of a capture range of image data. It is explanatory drawing of an image processing block. It is explanatory drawing of the mask area
  • FIG. 1 shows an example of a stereo image processing apparatus mounted on a vehicle.
  • the stereo image processing apparatus 100 includes a left camera 110, a right camera 120, a left image memory 130, a right image memory 140, a left image block setting unit 150, a right image block setting unit 160, and a left image block mask unit 170, right image block mask unit 180, right image block holding unit 190, right image mask block holding unit 200, matching processing unit 210, normalization processing unit 220, disparity calculation unit 230, distance calculation unit And 240 are configured.
  • the left camera 110 and the right camera 120 are attached to the upper inner surface of the windshield at a base length L [m] apart in the vehicle width direction (horizontal direction) in order to capture the situation in front of the vehicle.
  • the left camera 110 and the right camera 120 are disposed in the wiping area of the wiper in consideration of the fact that the windshield is dirty when it rains.
  • the left camera 110 and the right camera 120 are, for example, image sensors such as charge coupled devices (CCDs) and complementary metal oxide semiconductors (CMOSs), and have predetermined focal lengths f [m] and pixel pitches [m / pixel]. Have.
  • the alignment direction of the left camera 110 and the right camera 120 is not limited to the vehicle width direction, and may be the vehicle height direction (vertical direction). Further, three or more cameras may be attached to the upper inner surface of the windshield, and a pair of cameras may be selected and used among them depending on, for example, vehicle traveling conditions and the like. Furthermore, using an infrared camera as the left camera 110 and the right camera 120 may improve nighttime visibility or use a color camera for capturing a color image to utilize color information in subsequent processing. it can.
  • the left image memory 130 and the right image memory 140 include, for example, volatile memory such as RAM (Random Access Memory) that can be read and written at high speed, and the left image captured by the left camera 110 and the right camera 120 has different viewpoints. And store the right image respectively.
  • the left image memory 130 and the right image memory 140 are provided with a function of taking the left image and the right image captured by the left camera 110 and the right camera 120 into the volatile memory.
  • the left image block setting unit 150 and the right image block setting unit 160 set the block of the maximum size for calculating the parallax between the left image and the right image by block matching, and set the image data included in the area divided by the block. Extract.
  • block matching means, for example, a score representing the similarity of appearance between blocks while sequentially searching the left image for each block (range of rectangular shape) at a certain point of the right image It is a method of calculating a numerical value) and obtaining a block of the left image having a good score. Note that the left image block setting unit 150 and the right image block setting unit 160 appropriately change the positions at which blocks are set according to the progress of block matching.
  • the left image block mask unit 170 and the right image block mask unit 180 mask partial areas of the left and right image data extracted by the left image block setting unit 150 and the right image block setting unit 160, and Make it a valid image area. That is, the left image block mask unit 170 and the right image block mask unit 180 limit the area to be subjected to block matching among the image data of the area divided by the blocks.
  • the left image block mask unit 170 and the right image block mask unit 180 sequentially reduce the area to be masked, that is, sequentially enlarge the effective image area according to the progress of the block matching.
  • the right image block holding unit 190 holds the data of the right image extracted by the right image block setting unit 160, for example, in volatile memory. In this way, the image data of the area divided by the blocks can be used without accessing the right image memory 140, so the processing speed can be improved.
  • the right image mask block holding unit 200 holds, for example, volatile image data of the right image masked by the right image block mask unit 180. In this way, as in the case of the right image block holding unit 190, the masked right image data can be used without accessing the right image memory 140, so that the processing speed can be improved.
  • the matching processing unit 210 performs block matching between the data of the right image masked by the right image block mask unit 180 and the data of the left image masked by the left image block mask unit 170. The details of block matching will be described later.
  • the normalization processing unit 220 executes a process of normalizing the score in block matching so that the block sizes are not necessarily the same in the block matching performed by the matching processing unit 210, so that they can be evaluated under the same condition. .
  • FIGS. 2 and 3 illustrate an example of image processing that is repeatedly performed for each imaging frame of the left camera 110 and the right camera 120 when the stereo image processing apparatus 100 is activated.
  • step 1 the left image memory 130 captures the left image captured by the left camera 110 into the volatile memory of the left image memory 130.
  • the left image data taken into the volatile memory of the left image memory 130 is not the entire left image, as shown in FIG. Multiple lines according to the vertical size).
  • the left image memory 130 captures image data of a plurality of lines from above at the first capture, and captures subsequent left image data of a plurality of lines in the subsequent capture (the same applies to the right image). Note that the entire left image may be captured in the volatile memory of the left image memory 130.
  • the right image memory 140 captures the right image captured by the right camera 120 into the volatile memory of the right image memory 140.
  • right image data taken into volatile memory of right image memory 140 is, as shown in FIG. 4, similar to left image data, not the entire right image but, for example, a maximum of four lines performing block matching. Multiple lines according to the block height of the size. Note that the entire right image may be captured in the volatile memory of the right image memory 140.
  • step 3 the right image block setting unit 160 divides the right image data captured in the volatile memory of the right image memory 140 into an area partitioned by the block of the largest size at the block matching start position. Right image data is cut out and sent to the right image block holding unit 190.
  • the right image data of the area divided by the block of the largest size is referred to as “right image processing block” (the same applies to the left image).
  • step 4 the right image block holding unit 190 holds the right image processing block in the volatile memory.
  • step 5 the left image block mask unit 170 and the right image block mask unit 180 set the mask amount of the maximum size as the initial value of the mask amount for masking the left image processing block and the right image processing block. Therefore, at the position where block matching starts, the effective image area of the left image processing block and the right image processing block becomes minimum.
  • the area for masking the left image processing block and the right image processing block is only the left and right ends of the image processing block. Therefore, in the image processing block after masking, the central part in the horizontal direction is the effective image area. In this way, even when the mask amount is changed, the features of the entire image can be prevented from significantly fluctuating. For example, when a vehicle appears in the left half of the image processing block and a road surface appears in the right half, setting the mask area to the left or right of the image processing block causes only the road surface to appear in the image processing block depending on the mask status. Or the situation where only the vehicle appears. Therefore, there is a possibility that the appearing image may be significantly different between the image processing block before masking and the image processing block after masking. Therefore, by setting the left and right center portions of the image processing block as an effective image area, the vehicle and the road surface appear to be half even if the size of the effective image area is changed.
  • step 6 the right image block mask unit 180 masks the right image processing block and sends it to the right image mask block holding unit 200. As shown in FIG. 6, in the right image processing block masked, the central portion in the left-right direction is an effective image area.
  • step 7 the right image mask block holding unit 200 holds the masked right image processing block in the volatile memory.
  • step 8 the left image block setting unit 150 sets the block position of the left image to be subjected to block matching with the right image processing block. At the block matching start position, as shown in FIG. 5, the block position is the same as the position of the right image processing block.
  • step 9 the left image block mask unit 170 resets the variable n for counting the shift amount of the block position to 0 in order to change the mask amount according to the progress of the block matching.
  • step 10 the left image block mask unit 170 increments the variable n, that is, adds 1 to the variable n.
  • the left image block setting unit 150 cuts out the left image processing block of the area divided by the block of the largest size at the block position of the left image from the left image data taken into the volatile memory of the left image memory 130 .
  • the left image block mask unit 170 masks the left image processing block. As shown in FIG. 6, in the left image processing block masked, the central part in the left-right direction is an effective image area.
  • the matching processing unit 210 performs block matching between the masked right image processing block and the left image processing block. That is, for each pixel included in the effective image area of the right image processing block and the effective image area of the left image processing block, the matching processing unit 210 uses SAD (Sum of Absolute Difference), for example, The sum (score) of the luminance differences is calculated, and the smaller the value is, the higher the similarity is.
  • SAD Sum of Squared Difference
  • NCC Normalized Cross Correlation
  • ISC Increment Sign Correlation
  • SRF Statistical Reach Feature
  • the left image block setting unit 150 shifts the block position of the left image to the right by a predetermined amount (for example, one pixel) as shown in FIG.
  • the amount of shifting the block position of the left image is not limited to one pixel, and may be two or more and less than one pixel.
  • the shift amount is 2 pixels or more, speeding up of block matching can be expected, and when the shift amount is less than 1 pixel, the accuracy of block matching can be improved.
  • the shift amount is not an integer, for example, a weighted shift amount of pseudo decimal can be realized by weighted averaging of the luminances of adjacent pixels, but it is realized by using other known methods. May be
  • the left image block setting unit 150 determines whether or not the shift amount of the block position of the left image is larger than the threshold A.
  • the threshold value A can be a value according to the distance to the front end of the vehicle. That is, when the threshold A takes a large value, the parallax is searched for in a region where the distance to the object located in front of the vehicle is closer than the front end of the vehicle. Therefore, by setting the threshold value A to a value according to the distance to the front end of the vehicle, the parallax search range can be narrowed and the processing load can be reduced.
  • the left image block setting unit 150 determines that the shift amount of the block position is larger than the threshold A (Yes), the left image block setting unit 150 determines that the shift amount of the block position is the threshold A or less. If it does, the process is transited to step 22 (No).
  • the normalization processing unit 220 and the disparity calculation unit 230 cooperate to evaluate the result of block matching. That is, since the size of the block subjected to the block matching is not necessarily the same, the normalization processing unit 220 performs a normalization process that allows the blocks to be evaluated under the same condition. Specifically, the normalization processing unit 220 divides the score, which is the result of block matching, by the number of pixels included in the effective image area, thereby evaluating it regardless of the size of the effective image area as shown in FIG. To be able to
  • the disparity calculating unit 230 selects an effective image area corresponding to the score having the minimum value among all the normalized scores in one line.
  • an effective image area having a score that is not the smallest may be selected in some cases.
  • step S17 the disparity calculating unit 230 calculates the disparity ⁇ between the effective image area in the right image and the effective image area in the left image.
  • the distance calculation unit 240 calculates the distance Z to the object based on the parallax ⁇ .
  • the right image block setting unit 160 shifts the block position of the right image to the right, for example, by the width of the block of the largest size, as shown in FIG.
  • the right image block setting unit 160 determines whether or not the shift amount of the block position of the right image is larger than the threshold B.
  • the threshold B is for determining whether or not block matching has been performed to the right end of the right image, and can be, for example, a value according to the resolution of the right image and the maximum size of the block.
  • the right image block setting unit 160 transitions the process to step 21 (Yes), while determining that the shift amount of the block position is the threshold B or less If it does, the process is transited to step 3 (No).
  • step 21 the right image block setting unit 160 determines whether block matching has been performed up to the final line. If the right image block setting unit 160 determines that block matching has been performed up to the final line (Yes), the process proceeds to step 1 if it is determined that block matching has not been performed up to the final line (Yes). Let it be (No). The process of step 21 may be performed by another setting unit or processing unit, not limited to the right image block setting unit 160.
  • the left image block setting unit 150 determines whether or not the variable n has become larger than the threshold C.
  • the threshold C is for determining the timing of changing the size of the effective image area to be subjected to block matching, and is set to an appropriate value, for example, through experiments. If the left image block setting unit 150 determines that the variable n is larger than the threshold C, the process proceeds to step 23 (Yes), while if it is determined that the variable n is less than or equal to the threshold C, the process is performed. Transition to (No).
  • step 23 as shown in FIG. 11, the left image block mask unit 170 and the right image block mask unit 180 reduce the mask amount for masking the left image processing block and the right image processing block, that is, the left image processing block and Enlarge the effective image area of the right image processing block. Then, after the effective image area of the left image processing block and the right image processing block is enlarged, the processing is shifted to step 6.
  • the effective image area of the image processing block increases discretely (stepwise) according to the shift amount of the left image processing block, that is, effective every time the shift amount increases by i pixels.
  • the size of one side of the image area is increased by j pixels.
  • i and j may be fixed values in accordance with the processing speed, parallax calculation accuracy, etc. determined by the stereo image processing apparatus 100, or may be variable values in accordance with the parallax calculation result.
  • the change timing of the mask amount and the change amount thereof are determined such that the shift amount of the left image processing block and the length of one side of the effective image area are substantially proportional.
  • the change timing and the change amount of the mask amount may have a constant relationship in advance, or may be variable according to the situation.
  • left image data and right image data for a plurality of lines are taken into the left image memory 130 and the right image memory 140, respectively, as shown in FIG.
  • the right image processing block of the area divided by the block of the largest size for block matching is cut out and held in the volatile memory .
  • the right image processing block held in the volatile memory is masked at both ends so that the minimum effective image area is set at the center in the horizontal direction, and this is held in the volatile memory.
  • the left image processing block of the area divided by the block of the largest size to be subjected to block matching is cut out from the left image data taken into the left image memory 130.
  • the left image processing block is masked at both ends so that the minimum effective image area is set at the center in the horizontal direction. Then, the masked left image processing block and the right image processing block are block matched, and a score representing the similarity of the left image processing block to the right image processing block is calculated.
  • the block position of the left image is shifted to the right by a predetermined amount, and an area divided by the shifted blocks from the left image data fetched into the left image memory 130.
  • Left image processing block is cut out.
  • the processing load can be reduced as compared with the method of performing block matching by the block of the smallest size.
  • the image data of the same line as the previous block matching is used, it is possible to use the image data that has already been fetched into the left image memory 130 and the right image memory 140 as it is.
  • the left image processing block is masked at both ends so that the minimum effective image area is set at the center in the horizontal direction.
  • the masked left image processing block and the right image processing block are block matched, and a score representing the similarity of the left image processing block to the right image processing block is calculated.
  • the right image processing block held in the volatile memory can be used.
  • Such processing is sequentially performed, and as a result of shifting the block position of the left image to the right by a predetermined amount, as shown in FIG. 11, the left image processing is performed each time the shift amount of the block of the left image reaches the threshold C.
  • the amount of mask for masking the block and the right image processing block is successively reduced.
  • the left image processing block and the right image processing block are masked, and the left image processing block obtained by enlarging the effective image area and the right image processing block are block matched, and the similarity of the left image processing block to the right image processing block is calculated.
  • the representing score is calculated.
  • the right image processing block held in the volatile memory may be masked, and the time for reading the image data from the right image memory 140 may be omitted. it can.
  • the mask amount for masking the left image processing block and the right image processing block finally becomes 0, and as shown in FIG. 13, the effective image area of the left image processing block and the right image processing block is a block of the largest size. It becomes size.
  • the shift amount of the block of the left image reaches the threshold A
  • the block matching result of the left image processing block and the right image processing block is evaluated, and the left image processing block corresponding to the right image processing block is obtained.
  • the distance to the object is calculated from the parallax between the image processing block and the left image processing block.
  • the block position of the right image is shifted to the right by the width of the block of the largest size, and the left image corresponding to the right image processing block A search for processing blocks is performed.
  • the left image memory 130 and the right image memory 140 receive the next plurality of lines as shown in FIG.
  • the left image data and the right image data of a minute are respectively taken.
  • the processing described above is sequentially performed on the image data captured in the left image memory 130 and the right image memory 140 until the final line is reached.
  • the parallax of a distant object is calculated by gradually increasing the size of the block used in block matching according to the progress of block matching, a small block is used, and the parallax of the nearby object is used.
  • a large block will be used. Therefore, distance measurement accuracy of an object can be ensured without generating a plurality of reduced images from a stereo image as in the prior art, and the processing load for calculating the parallax of the object in the stereo image can be reduced.
  • the size of the block to be subjected to block matching may be fixed, and block matching may be performed in the area divided by the block. By doing this, the load of block matching and block size change can be reduced, and speeding up can be achieved.
  • the block size is changed so that adjacent blocks or blocks further apart can be covered, considering that the correct value of disparity may be moved to adjacent blocks or blocks further apart You may Furthermore, in this case, the block size may be changed to a predetermined size, or the block size may be fixed in order to omit the normalization process and the block size changing process.
  • the effective image area of the image processing block is not the maximum, as shown in FIGS. 15 and 16, a parallax unsearched area is generated.
  • the disparity of the unsearched area may be searched using a block of a size according to the unsearched area.
  • a block may extend beyond the searched area or the image area, the maximum size of the block centered on the unsearched area may be defined, and the block matching may be performed using the block.
  • the position of the next search area of the searched area is such that the block of the smallest size is adjacent to the right of the searched area. They may be adjacent to each other and partially overlap the searched area when the block size is enlarged. In this case, there is an advantage that the parallax is calculated with the size of the minimum block regardless of the size of the block corresponding to the parallax.
  • part of the block is an image It may be permitted to be out of the region.
  • the part which is out of the image area has the same predetermined luminance value, for example, 0, for the left and right images so that the area does not affect the block matching.
  • the method excludes the area of the block area outside the image area from the effective area, sets the score ratio before and after the change point of the block size as the normalization coefficient, extracts a certain number of pixels from the block, and blocks When matching is performed, pixel extraction may not be performed from outside the image region.
  • the normalization coefficient may not be a value proportional to a simple block size, but may be a value proportional to the area of the block excluding the outside of the image area.
  • the position of the block at the left end of the image is such that the block of the largest size is in contact with the left end, but even if there is an unsearched area at the left end of the image, the parallax of this part is not searched You may ignore it. In this case, since the unsearched area is relatively small and at the image edge, it is not necessary to consider exceptional processing of the area having a small influence.
  • the left image block mask unit 170 and the right image block mask unit 180 mask the area of the peripheral portion of the left image processing block and the right image processing block as shown in FIG. Let the area be the effective image area.
  • the left image block mask unit 170 and the right image block mask unit 180 appropriately change the mask amount while keeping the center and the aspect ratio of the effective image area constant, as shown in FIG. Expand one by one. Then, the distance to the object of the stereo image is measured through the same procedure as in the first embodiment.
  • the operation and effects are the same as those of the first embodiment except for the area for masking the image processing block, and therefore the description thereof is omitted. If necessary, refer to the first embodiment.
  • the basic configuration of the stereo image processing apparatus 100 is the same as that of the first embodiment, and thus the description thereof is omitted. In the following description, only the parts different from the first embodiment will be described from the viewpoint of eliminating the redundant description.
  • the parallax calculation processing in a certain block When the parallax calculation processing in a certain block is completed, the movement to an unprocessed block in the image is performed, and the parallax calculation processing is similarly performed.
  • the block for which the disparity calculation process has been performed is a block of a size matching the block matching.
  • the search is performed twice for that area and is wasted. For this reason, it is desirable to set the search order so that duplication does not occur as much as possible.
  • the position of the initial search block is that of the smallest effective image area masked.
  • the upper left of the image processing block is in contact with the upper left of the image rather than the upper left is in contact with the upper left of the image. In this way, in the process of performing block matching while gradually reducing the mask amount, it is possible to prevent a situation in which the effective image area is not originally included in the block.
  • the image processing block is made to contact the parallax search end area. By doing this, it is possible to prevent the situation in which the block includes the disparity search end area in the process of performing the block matching while gradually reducing the mask amount.
  • the part that is out of the image area has the same predetermined luminance, for example, 0, for both left and right images, so that the presence of this area does not affect the block matching. Also, in the normalization process, as in the first embodiment, a process corresponding to the presence of an area outside the image area is performed.
  • the disparity of the unsearched areas and the block sizes to be corresponded are also likely to be equal to these.
  • the length or width of the unsearched area or both of the unsearched area is made as long as possible as long as the surrounding block size without exceeding the unsearched area, and only the shift amount of the search range is searched with the size of the block You may Further, while the block size of the unsearched area is fixed, the search may be performed in the entire range of the shift amount.
  • the block of the right image is fixed and the block of the left image is shifted to perform block matching.
  • the block of the left image is fixed and the block of the right image is shifted. Matching may be performed.
  • the direction in which the block of the right image is shifted is opposite because of the direction in which the parallax occurs.
  • the effective image area is maximized without masking the image processing block, and a search is performed in the direction of the shift amount 0 from the shift position of the block where the parallax is maximum.
  • the amount may be increased by one step.
  • the normalization processing unit 220 may not perform normalization processing, and the matching processing unit 210 may perform block matching in a method having the same effect as normalization.
  • the number of pixels to be compared with the luminance difference may be fixed regardless of the size of the image area. Therefore, a fixed number of pixels are extracted from the image area, and block matching is performed on this set of pixels.
  • the block size may be reduced in advance to set the pixel extraction area.
  • the image range in which block matching is performed is not limited to the configuration in which the image processing block is defined by the mask, and may be defined by changing the size of the image processing block fetched from the left image memory 130 and the right image memory 140 . In this case, the process of masking the image processing block is unnecessary.
  • stereo image processing apparatus 110 left camera 120 right camera 150 left image block setting unit 160 right image block setting unit 170 left image block mask unit 180 right image block mask unit 210 matching processing unit 220 normalization processing unit 230 disparity calculation unit 240 distance Calculation unit

Landscapes

  • Engineering & Computer Science (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • Image Analysis (AREA)
  • Measurement Of Optical Distance (AREA)
  • Testing, Inspecting, Measuring Of Stereoscopic Televisions And Televisions (AREA)
  • Length Measuring Devices By Optical Means (AREA)
  • Image Processing (AREA)
PCT/JP2013/077418 2012-10-29 2013-10-09 ステレオ画像処理装置及びステレオ画像処理方法 WO2014069196A1 (ja)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
JP2012-237592 2012-10-29
JP2012237592A JP5918684B2 (ja) 2012-10-29 2012-10-29 ステレオ画像処理装置及びステレオ画像処理方法

Publications (1)

Publication Number Publication Date
WO2014069196A1 true WO2014069196A1 (ja) 2014-05-08

Family

ID=50627110

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/JP2013/077418 WO2014069196A1 (ja) 2012-10-29 2013-10-09 ステレオ画像処理装置及びステレオ画像処理方法

Country Status (2)

Country Link
JP (1) JP5918684B2 (enrdf_load_stackoverflow)
WO (1) WO2014069196A1 (enrdf_load_stackoverflow)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
EP4611366A1 (en) * 2024-02-29 2025-09-03 Canon Kabushiki Kaisha Image processing apparatus, imaging apparatus, image processing method, and storage medium

Families Citing this family (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
KR101690645B1 (ko) * 2015-09-21 2016-12-29 경북대학교 산학협력단 다단계 시차영상 분할이 적용된 시차탐색범위 추정 방법 및 이를 이용한 스테레오 영상 정합장치
WO2018037479A1 (ja) * 2016-08-23 2018-03-01 株式会社日立製作所 画像処理装置、ステレオカメラ装置及び画像処理方法
KR102297147B1 (ko) * 2019-08-12 2021-09-03 고려대학교 산학협력단 어류의 시기능 평가를 위한 3차원 영상 분석 방법
JP7554105B2 (ja) * 2020-12-02 2024-09-19 ヤマハ発動機株式会社 距離認識システムおよびその制御方法、船舶

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JP2001082927A (ja) * 1999-09-16 2001-03-30 Nippon Telegr & Teleph Corp <Ntt> 三次元画像処理方法、装置、および三次元画像処理プログラムを記録した記録媒体
JP2002247604A (ja) * 2001-02-22 2002-08-30 Olympus Optical Co Ltd 視差量測定装置
JP2004234423A (ja) * 2003-01-31 2004-08-19 Seiko Epson Corp ステレオ画像処理方法およびステレオ画像処理装置、並びにステレオ画像処理プログラム
JP2007218922A (ja) * 2007-03-27 2007-08-30 Topcon Corp 画像測定装置
WO2012137511A1 (ja) * 2011-04-07 2012-10-11 パナソニック株式会社 画像処理装置および画像処理方法

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JP2001082927A (ja) * 1999-09-16 2001-03-30 Nippon Telegr & Teleph Corp <Ntt> 三次元画像処理方法、装置、および三次元画像処理プログラムを記録した記録媒体
JP2002247604A (ja) * 2001-02-22 2002-08-30 Olympus Optical Co Ltd 視差量測定装置
JP2004234423A (ja) * 2003-01-31 2004-08-19 Seiko Epson Corp ステレオ画像処理方法およびステレオ画像処理装置、並びにステレオ画像処理プログラム
JP2007218922A (ja) * 2007-03-27 2007-08-30 Topcon Corp 画像測定装置
WO2012137511A1 (ja) * 2011-04-07 2012-10-11 パナソニック株式会社 画像処理装置および画像処理方法

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
EP4611366A1 (en) * 2024-02-29 2025-09-03 Canon Kabushiki Kaisha Image processing apparatus, imaging apparatus, image processing method, and storage medium

Also Published As

Publication number Publication date
JP5918684B2 (ja) 2016-05-18
JP2014089498A (ja) 2014-05-15

Similar Documents

Publication Publication Date Title
JP6167525B2 (ja) 距離計測装置及び車両
JP6577565B2 (ja) 深度画像処理用の方法及び装置
WO2014069196A1 (ja) ステレオ画像処理装置及びステレオ画像処理方法
JP6158779B2 (ja) 画像処理装置
CN108629763B (zh) 一种视差图的评判方法、装置及终端
JP5752618B2 (ja) ステレオ視差算出装置
CN103080973B (zh) 立体图像处理装置及立体图像处理方法
JP4521235B2 (ja) 撮影画像の変化抽出装置及び方法
JP2008065634A (ja) 物体検出装置および物体検出方法
JP2010224924A (ja) 画像処理装置
JP5776771B2 (ja) 対応点探索装置、および距離測定装置
CN111105452B (zh) 基于双目视觉的高低分辨率融合立体匹配方法
CN103383776A (zh) 一种基于分段匹配和贝叶斯估计的递进式立体匹配算法
TW202022802A (zh) 辨識影像中前景物件的方法與電子裝置
US10043106B2 (en) Corresponding point searching method and distance detection device
JP5091897B2 (ja) 停止線検出装置
CN103968772A (zh) 一种基于单目视觉检测的活塞环检测方法
WO2021167910A1 (en) A method for generating a dataset, a method for generating a neural network, and a method for constructing a model of a scene
CN108460348A (zh) 基于三维模型的道路目标检测方法
CN105138979A (zh) 基于立体视觉的运动人体头部检测方法
JP4831084B2 (ja) 対応点探索装置
JP6060612B2 (ja) 移動面状況認識装置、移動体、及びプログラム
CN113486825A (zh) 非接触式指纹采集装置及其方法、系统、介质
CN113706622A (zh) 基于双目立体视觉的路面拟合方法、系统和智能终端
KR101229376B1 (ko) 렌즈 왜곡 보정에 사용된 보간 패턴 검출을 통한 촬영 카메라 식별 방법 및 장치

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 13850457

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

122 Ep: pct application non-entry in european phase

Ref document number: 13850457

Country of ref document: EP

Kind code of ref document: A1