WO2011033673A1 - 画像処理装置 - Google Patents
画像処理装置 Download PDFInfo
- Publication number
- WO2011033673A1 WO2011033673A1 PCT/JP2009/066444 JP2009066444W WO2011033673A1 WO 2011033673 A1 WO2011033673 A1 WO 2011033673A1 JP 2009066444 W JP2009066444 W JP 2009066444W WO 2011033673 A1 WO2011033673 A1 WO 2011033673A1
- Authority
- WO
- WIPO (PCT)
- Prior art keywords
- pixel
- depth
- value
- disparity
- map
- Prior art date
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T15/00—3D [Three Dimensional] image rendering
- G06T15/10—Geometric effects
- G06T15/20—Perspective computation
- G06T15/205—Image-based rendering
-
- 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
- H04N13/261—Image signal generators with monoscopic-to-stereoscopic image conversion
Definitions
- the present invention relates to an image processing apparatus that generates an image for 3D display from a 2D image.
- Patent Document 1 In the 3D image generation method disclosed in Japanese Patent Application No. 2008-504887 (Patent Document 1), first, a moving object is detected and tracked from a 2D image including the moving object. Then, a standard template is set at the detection position, and correction is performed using a line having the size and contrast of the detected object, thereby separating the image into a foreground portion and a background portion including the detected object. A depth model is given to each of the foreground and background, thereby generating a three-dimensional image.
- Patent Document 1 since segmentation is performed using line information such as edges in an image in a two-dimensional image, it matches the actual three-dimensional information. There is a problem that only the periphery of the object.
- the method of giving a depth model after a segment has a problem that the image quality may be remarkably deteriorated when a three-dimensional image is formed because the given depth model does not necessarily match the three-dimensional information.
- the present invention provides an image processing apparatus capable of generating a higher quality 3D image from a 2D image.
- An image processing apparatus includes a detection unit that detects an object in an input image, and a depth corresponding to the type of the detected object among at least one depth template that describes a pixel depth value. Selecting the template and placing the depth template selected according to the position of the detected object in the input image on the depth map, to describe the depth map describing a depth value for each pixel in the input image.
- Depth map generation unit to generate, at least one target pixel in the depth map, the weight of the peripheral pixel, the pixel value of the corresponding target pixel in the input image corresponding to the target pixel, and the peripheral pixel The depth value of the target pixel is calculated from the relationship with the pixel value of the corresponding peripheral pixel in the input image.
- a correction unit that corrects the depth value of the target pixel from the weighted sum of the weights of the peripheral pixels and the depth value, and a plurality of parallax images based on the depth map corrected by the correction unit and the input image.
- an image generation unit for generating.
- a higher quality 3D image can be generated from a 2D image.
- FIG. 1 shows a configuration example of an image processing apparatus according to a first embodiment.
- 2 shows a configuration example of an image processing apparatus according to a second embodiment.
- 9 shows a configuration example of an image processing apparatus according to a third embodiment.
- 10 shows a configuration example of an image processing apparatus according to Embodiment 4.
- An example of a depth template is shown. The state which displayed the depth template of FIG. 5 in 3D is shown. An example in which a depth template is arranged on a map is shown. The calculation method of a parallax vector is shown. The left disparity vector and the right disparity vector obtained by dividing the disparity vector are shown.
- FIG. 1 shows a configuration example of an image processing apparatus according to the present embodiment.
- the image input unit 10 inputs a two-dimensional image to be processed.
- the input two-dimensional image includes pixel values of a plurality of pixels.
- the image input unit 10 can input an input image from any device or medium.
- the image input unit 10 may input image data from a recording medium such as an HDD, or may input image data from an external device connected via a network.
- the object detection unit 100 analyzes the input image and detects an object included in the input image and its position.
- the depth template storage unit 120 stores a depth template 20 in which the depth value (depth value) of each pixel of the corresponding object is described for each type of object.
- the depth map generation unit 200 reads the depth template 20 corresponding to the object detected by the object detection unit 100 from the storage unit 120, and arranges the depth template 20 on the initial map according to the position of the detected object. A depth map describing a depth value corresponding to each pixel of the input image is generated.
- the depth map correction unit (correction unit) 300 sequentially selects each pixel on the depth map as a target pixel, and corrects the depth value of the target pixel by performing weighted smoothing on the target pixel and its surrounding pixels.
- the weight of the target pixel and each peripheral pixel is determined by calculating the pixel value of the pixel (corresponding target pixel) of the input image corresponding to the target pixel and the pixel value of the pixel (corresponding peripheral pixel) in the input image corresponding to each peripheral pixel. Calculate based on the difference.
- the three-dimensional image generation unit 400 generates a plurality of parallax images (right-eye image and left-eye image) from the input image and the corrected depth map.
- the plurality of generated parallax images are used for displaying a stereoscopic image. For example, by displaying a plurality of parallax images alternately in a time-division manner, the observer can recognize a stereoscopic image.
- the observer wears special glasses such as liquid crystal glasses, and the left and right liquid crystal shutters are switched according to the display of each parallax image, so that the left eye image and the right eye image are alternately incident on the left and right eyes. Thereby, the observer can recognize the stereoscopic image.
- the object detection unit 100, the depth template 20, the depth map generation unit 200, the depth map correction unit 330, and the three-dimensional image generation unit 400 will be described in detail.
- the following items (1) to (4) are assumed.
- (1) Set the upper left corner of the input image as the origin, and set the x-axis in the horizontal direction and the y-axis in the vertical direction.
- the method for setting coordinates is not limited to this.
- the pixel value of the coordinates (x, y) of the input image is represented as P (x, y).
- the pixel value only needs to represent the brightness or color component of the image, and corresponds to, for example, luminance, brightness, a specific color channel, and the like.
- the pixel value at the map coordinates (X, Y) is represented as Z (X, Y). At this time, the pixel value represents depth information, and the greater the value, the greater the depth (depth).
- the coordinates of the input image and the coordinates of the map should be in a one-to-one correspondence. Unless otherwise specified, it is assumed that the size of the input image is equal to the size of the map, and the coordinates (x, y) of the input image and the coordinates (X, Y) of the map correspond to each other.
- the pixel value of the input image is described as “pixel value” and its value range is [0,255] (0 to 255). Further, the pixel value of the depth map is described as “depth value”, and its value range is set to [0,255] (0 to 255).
- the object detection unit 100 will be described.
- the object detection unit 100 detects the whole or part of the target object and its position from the input image.
- it may be the whole person, a part of the person (face, hand, foot), vehicle, plant, etc., and each may be treated as a different type of object depending on the orientation of the person's face.
- the position on the input image of the i-th object detected at the coordinates (x, y) on the input image is represented as Ai (x, y).
- a generally known method can be used as a method for detecting an object. Further, various methods may be used in combination depending on the object to be detected. For example, when the detection target object is a person, a method using a face detection method for detecting a face that is a part of the person can be considered. For example, Reference 1 (Yoshi Mita, Toshimitsu Kaneko, Osamu Hori, “Joint Haar-like feature based on co-occurrence suitable for face detection” IEICE Transactions D-II Vol.J89-D-II No.8 pp .1791-1801, 2006) can be used.
- a face is detected from the Haar-like feature of the image, and in detail, the position and size of a rectangle arranged so as to surround the face can be obtained. That is, the position and size of the face can be known.
- the orientation of the face can be detected by changing the dictionary used for detection.
- the depth template 20 is prepared for each type of object.
- the depth template represents the approximate shape of the actual three-dimensional shape of the object. Specifically, when the object is viewed from the direction in which the object is desired to be detected, the depth template represents the depth as a pixel value as a two-dimensional image. For example, when the detection target object is the upper body of a person, the depth template is as shown in FIG. In FIG. 5, the smaller the depth value (closer to black), the smaller the depth, and the larger the depth value (closer to white), the greater the depth.
- Fig. 6 shows 3D display of Fig. 5.
- the depth template has such three-dimensional information.
- One or more such depth templates are prepared for each type of object to be detected, and stored in the storage unit 120.
- the depth map generation unit 200 arranges a depth template corresponding to the detected object at a position Bi (X, Y) on the map corresponding to the position Ai (x, y) of the object detected by the object detection unit 100. . This generates a depth map.
- Fig. 7 shows an example in which a person is detected from a certain input image and a depth template is arranged on the map (initial map).
- a human face is detected by the object detection unit 100 from the input image in FIG. 7 (a), and the upper left corner of the face is the coordinate A1 (x, y).
- the corresponding depth template is arranged so that the upper left corner is positioned at position B1 (X, Y) on the depth map corresponding to the upper left corner of the person's face.
- the depth templates may overlap on the depth map due to the detection of a plurality of objects.
- various methods for giving depth values to the coordinates at which a plurality of depth templates overlap For example, the following methods (1) to (6) can be considered.
- (1) Use the average. That is, the average of the depth values at the coordinates of a plurality of depth templates that overlap the coordinates is used.
- (2) Use the minimum value. That is, the minimum value of the depth value at the coordinates of the depth template overlapped with the coordinates is used.
- Use a weighted average A weighted average of depth values at the coordinates of a plurality of depth templates that overlap the coordinates is used. For example, the weight is increased as the template has a smaller depth.
- (4) Use the median.
- the median depth value at the coordinates of the depth template that overlaps the coordinates is used.
- initial values are set for each coordinate of the initial map.
- the reference value for example, the reference value 255 having the largest depth (the largest depth) may be set.
- the reference value is updated (overwritten) by the depth value of the template.
- the depth value may be updated according to the methods (1) to (5).
- the depth map correction unit 300 corrects the depth map by performing weighted smoothing on the pixel of interest D (X, Y) and its surrounding pixels on the depth map.
- Peripheral pixels are pixels that exist within a range close to the target pixel. For example, a pixel existing within a certain fixed distance range from the target pixel is shown.
- the weight used in the correction is set according to the relationship between the pixel value of the corresponding target pixel C (x, y) in the input image corresponding to the target pixel D (X, Y) and its peripheral pixels (corresponding peripheral pixels).
- the weight is basically set according to the difference between the pixel value of the corresponding target pixel C (x, y) and the pixel value of the corresponding peripheral pixel. For example, the weight may be set to be larger as the difference is smaller, and the weight may be smaller as the difference is larger.
- a bilateral filter can be used to correct such a depth map.
- Depth map correction using a bilateral filter can be expressed by Equation 1 where the corrected depth value is Z ′ (X, Y).
- k is the window size of the filter.
- Equation 1 For example, if a Gaussian distribution is used for W 1 and W 2 in Equation 1, the result is as follows.
- ⁇ 1 and ⁇ 2 are standard deviations of the Gaussian distribution.
- W 1 evaluates the spatial distance between the corresponding target pixel and the corresponding peripheral pixel.
- Z ′ (X, Y) in Equation 1 a weighted average of depth values of each peripheral pixel and the target pixel is calculated.
- An ⁇ filter can also be used for depth map correction.
- Depth map correction using the ⁇ filter can be expressed by Equation 2. The value of the ⁇ filter is selected based on the input image and filtered on the depth map.
- W 1 and W 2 may be set as follows.
- k is the filter window size
- ⁇ is the threshold of the ⁇ filter.
- the weighted average of the depth values of each peripheral pixel and the target pixel is calculated. Specifically, the average of the depth value selected for each peripheral pixel and the depth value of the target pixel is calculated.
- W 1 is always set to 1, but a Gaussian distribution or the like can be used for W 1 .
- a median filter can also be used for depth map correction.
- a median value is searched for among the corresponding target pixel and the corresponding peripheral pixel, and when the pixel value matches the median value, filtering is performed so as to select a pixel on the depth map corresponding to the pixel having the pixel value.
- the median is determined by the filter window range.
- the weights W 1 and W 2 in Equation 1 may be set as in Equation 3 below.
- W 2 is 1 when the pixel value P (x + m, y + n) of the corresponding peripheral pixel C (x + m, y + n) matches the median value, and 0 when it does not match. This is to take the average when there are a plurality of medians.
- the weight of W2 is set not to be 0, 1 but to a pixel having a smaller difference from the median value (corresponding target pixel or corresponding peripheral pixel) so that the weight becomes larger. It may be set.
- the three-dimensional image generation unit 400 converts the corrected depth map into a disparity map by the disparity (parallax) conversion unit 410, and generates a parallax image from the disparity map and the input image by the parallax image generation unit 420. To do.
- the disparity conversion unit 410 obtains a disparity vector (disparity value) of each pixel from the depth value z of each pixel in the depth map, and generates a disparity map describing the disparity vector of each pixel.
- the parallax vector represents how much the input image is moved to generate the parallax image. In this way, the disparity conversion unit 410 converts the depth map into a disparity map.
- the disparity vector d is a similarity between a triangle formed by connecting a right eye, a left eye, and an object, and a disparity (right disparity and left disparity) on the screen and the object.
- z, d, b, z s , z 0 and L z are defined.
- Arbitrary values are set for b, z s , z 0 and L z .
- the depth value z is in the range of 0-255, with 0 representing the foreground and 255 representing the farthest (in FIG. 8, the lower horizontal dotted line is 0, and the upper horizontal dotted line is 255. Corresponding). However, this value is only a virtual one and is different from the actual distance.
- the distance from the screen to the object can be expressed as follows.
- Stereoscopic parameters b, z s , z 0 , and L z can be arbitrarily determined based on the stereoscopic vision desired to be provided. For example, z s (distance to the screen) is determined according to the actual screen position, and z 0 (jump distance) is increased when it is desired to increase the pop-up from the screen. The depth of the depth of the real space can be determined by L z.
- the disparity vector d can be calculated from the depth value z by using the similarity of the two triangles described above and the following mathematical expression (depth disparity vector conversion model).
- the disparity conversion unit 410 obtains a disparity vector from the depth value z for each pixel in the depth map, and generates a disparity map in which the disparity vector for each pixel is described.
- the parallax image generation unit 420 generates the number of parallax images to be generated from the input image and the disparity map.
- the left parallax image and the right parallax image have a parallax vector d pixel of ⁇ 1/2 and 1/2. Can be generated from the following left and right disparity vectors.
- the left parallax image pixel value P (x, y) of the input image can be generated by moving according to d L.
- Right parallax image pixel value P (x, y) of the input image can be generated by moving according to d R. Since there is a possibility that a hole will be formed simply by moving, the image may be filled in the hole area by interpolating from surrounding parallax vectors.
- the case of two parallaxes has been described as an example, but the same processing may be performed for multiparallax cases.
- a depth map is generated by arranging a depth template corresponding to an object detected from an input image, and the depth value of the target pixel in the depth map is set to the corresponding target pixel and the corresponding peripheral pixel in the input image.
- the correction is performed based on the weights for the surrounding pixels determined based on the distance between the pixel values.
- FIG. 2 shows a configuration example of the image processing apparatus according to the present embodiment.
- the image input unit 10 inputs a two-dimensional image to be processed.
- the object detection unit 101 detects an object included in the input image and its type, position, size, and orientation.
- the storage unit 120 stores the depth template 20 having the depth value of each pixel of the corresponding object for each object type.
- the depth template correction unit 500 reads a depth template corresponding to the type of object detected by the object detection unit 101 from the storage unit 120, and corrects the depth template according to the size and orientation of the object.
- the depth map generation unit 200 generates a depth map by placing the depth template corrected by the depth template correction unit 500 on the map based on the position of the object detected by the object detection unit 100.
- the depth map correction unit 300 selects each pixel on the depth map as a target pixel, corrects the depth value of the target pixel by performing weighted smoothing on the target pixel and its surrounding pixels, and thereby corrects the depth map. To do.
- the correction method the same method as in the first embodiment can be used.
- the storage unit 130 stores another depth map 30 which is a depth map corresponding to the input image, which is given by some means.
- the depth map synthesis unit 600 reads the other depth map 30 from the storage unit 30 and synthesizes the other depth map 30 with the depth map corrected by the depth map correction unit 300.
- the three-dimensional image generation unit 400 generates a parallax image from the depth map synthesized by the depth map synthesis unit 600 and the input image.
- the object detection unit 101 the depth template correction unit 500, the other depth map information 30, and the depth map composition unit 600 will be described in more detail.
- the object detection unit 101 will be described.
- the object detection unit 101 detects the position, size, and orientation of all or part of the target object from the input image. It also detects the type of object. Except for detecting the size and orientation, the operation is the same as the operation of the object detection unit 100 of the first embodiment.
- the depth template correction unit 500 will be described.
- the depth template correction unit 500 corrects the depth template read from the storage unit 120 according to the size and orientation of the detected object. If all templates are prepared according to the size and orientation of the object, the amount will be large. Therefore, the depth template prepared in advance is modified based on the object detection information.
- the size of the depth template may be enlarged or reduced.
- the scaling can be performed by a generally known method.
- the orientation of the depth template may be changed. The orientation may be changed using a generally known morphing method.
- the other depth map 30 will be described.
- the other depth map 30 is a depth map related to an input image given by other means.
- the other depth map 30 may be, for example, another depth map that describes the background composition depth from the overall composition.
- the depth map synthesis unit 600 will be described.
- the depth map synthesis unit 600 synthesizes the depth map corrected by the depth map correction unit 300 and the other depth map 30. There may be any number of other depth maps to be combined.
- each depth map may be synthesized.
- a synthesis method for each pixel for example, the following method can be considered. (1) Use the average depth value of each pixel. (2) The maximum value among the depth values of each pixel is used. (3) Use the minimum value of the depth values of each pixel. (4) Use a weighted average of the depth values of each pixel. For example, the weight is increased as the depth is decreased. (5) The median depth value of each pixel is used.
- a depth map obtained by arranging a depth template is combined with another depth map, a depth map that achieves high contrast can be obtained.
- a parallax image capable of visually recognizing a quality stereoscopic image can be generated.
- FIG. 3 shows a configuration example of the image processing apparatus according to the present embodiment.
- the image input unit 10 inputs a two-dimensional image to be processed.
- the object detection unit 100 detects an object, its type, and position from the input image.
- the storage unit 140 stores a disparity template 40 describing a disparity value (parallax value) of each pixel of a corresponding object for each type of object.
- the disparity map generation unit 700 reads the disparity template 40 corresponding to the type of the object detected by the object detection unit 100 from the storage unit 40, and displays the disparity template 40 on the map according to the position of the detected object.
- a disparity map is generated by placing the
- the disparity map correction unit (correction unit) 800 selects each pixel on the disparity map as a target pixel, and corrects the disparity value of the target pixel by performing weighted smoothing on the target pixel and its surrounding pixels. This corrects the disparity map.
- the weight of the target pixel and each peripheral pixel is determined by the pixel value of the pixel (corresponding target pixel) of the input image corresponding to the target pixel and the pixel value of the pixel (corresponding peripheral pixel) in the input image corresponding to each peripheral pixel. It calculates according to the difference.
- the three-dimensional image generation unit 400 generates a parallax image from the input image and the corrected disparity map.
- disparity template 40 the disparity map generation unit 700, the disparity map correction unit 800, and the 3D image generation unit 400 will be described in more detail.
- the disparity map has the upper left corner of the map as the origin, the X axis in the horizontal direction, and the Y axis in the vertical direction.
- the method of setting the coordinates is not limited to this.
- the pixel value (disparity value) at the coordinates (X, Y) of the disparity map is represented as d (X, Y).
- the coordinates of the input image and the disparity map have a one-to-one correspondence.
- the size of the input image is equal to the size of the map, and the coordinates (x, y) of the input image correspond to the coordinates (X, Y) of the map.
- the disparity template 40 is prepared for each type of object, and has a disparity (parallax) value of the corresponding type of object.
- the disparity template 40 can be obtained by converting the depth template 20 by the same processing as the disparity conversion unit 410 in FIG.
- the disparity map generation unit 700 will be described.
- the disparity map generation unit 700 is similar to the depth map generation unit 200 of FIG. 1, and the position Bi (X, y) on the disparity map corresponding to the position Ai (x, y) of the object detected by the object detection unit 101 A disparity map is generated by arranging a disparity template corresponding to the type of the detected object in Y).
- the disparity map correction unit 800 will be described. Similar to the depth map correction unit 300 in FIG. 1, the disparity map correction unit 800 corrects the disparity map by performing weighted smoothing on the pixel of interest E (X, Y) on the disparity map and its surrounding pixels. .
- the weight used at this time is set according to the distribution of the pixel values of the corresponding target pixel C (x, y) in the input image corresponding to the target pixel E and the corresponding peripheral pixel (the pixel in the input image corresponding to the peripheral pixel).
- the weight is basically set according to the difference between the pixel values of the corresponding target pixel and the corresponding peripheral pixel. For example, the smaller the difference, the larger the weight, and the larger the difference, the smaller the weight may be set.
- a bilateral filter can be used as in the first embodiment.
- the correction of the disparity map can be expressed by Equation 6 where the corrected disparity value is d ′ (X, Y). It becomes.
- k is the window size of the filter.
- an ⁇ filter a weighted ⁇ filter, a median filter, or a weighted median filter can be used. Refer to the description of the first embodiment for details of the correction method of each filter.
- the three-dimensional image generation unit 400 will be described.
- the 3D image generation unit 400 generates a parallax image from the disparity map obtained by the disparity map correction unit 800 and the input image in the same manner as in the first embodiment.
- a disparity map is generated by disposing disparity templates corresponding to objects detected from an input image, and the disparity value of the target pixel in the disparity map is set as the corresponding target pixel in the input image. And correction based on the weights for the peripheral pixels determined based on the distance between the pixel values of the corresponding peripheral pixels.
- a disparity map adapted to the actual three-dimensional information of the input image can be obtained with high contrast (for example, without blurring the edges), and thus a parallax image capable of visually recognizing a high-quality stereoscopic image can be obtained. Can be generated.
- FIG. 4 shows a configuration example of the image processing apparatus according to the present embodiment.
- the image input unit 10 inputs a two-dimensional image to be processed.
- the object detection unit 101 detects an object and its type, position, size, and orientation from the input image.
- the storage unit 140 stores a disparity template 40 for each object type.
- the disparity template correction unit 900 reads the disparity template 40 corresponding to the type of the object detected by the object detection unit 101 from the storage unit 140, and reads the disparity template 40 from the detected object size, direction, etc. Correct as necessary.
- the disparity map generation unit 700 arranges the disparity template corrected by the disparity template correction unit 900 on the map according to at least the former of the position and type of the object detected by the object detection unit 101. Generate a disparity map.
- the disparity map correction unit 800 corrects the disparity map by performing weighted smoothing on the target pixel on the disparity map and its surrounding pixels. Details of the processing are the same as in the third embodiment.
- the storage unit 150 stores another disparity map 50 that is a disparity map corresponding to an input image given by some means.
- the disparity map combining unit 910 combines the disparity map corrected by the disparity map correcting unit 800 and the other disparity map 50.
- the three-dimensional image generation unit 400 generates a parallax image from the input image and the corrected disparity map.
- the disparity template correction unit 900 the disparity map generation unit 700, the other disparity map 50, the disparity map synthesis unit 910, and the 3D image generation unit 400 will be described in more detail.
- the disparity template correction unit 900 will be described.
- the disparity template correction unit 900 corrects the disparity template according to the size and orientation of the detected object, similarly to the depth template correction unit 500 of FIG. A method similar to the depth template correction unit 500 can be used for the correction.
- the disparity map generation unit 700 will be described.
- the disparity map generation unit 700 is similar to the depth map generation unit 200 of FIG. 2 in that a position Bi (X on the disparity map corresponding to the position Ai (x, y) of the object detected by the object detection unit 101 is used.
- Y the disparity template corrected by the disparity template correcting unit 900 is arranged.
- the other disparity map 50 is a disparity map related to an input image given by other means.
- a background composition disparity can be used from the entire composition or the like.
- a disparity map used for an image before t frames can be used.
- the disparity map combining unit 910 will be described.
- the disparity map combining unit 910 combines the disparity map corrected by the disparity map correcting unit 800 and the other disparity map 50.
- a synthesis method a method similar to the processing of the depth map synthesis unit 600 in FIG. 2 can be used.
- the three-dimensional image generation unit 400 will be described.
- the three-dimensional image generation unit 400 generates a parallax image from the disparity map generated by the disparity map synthesis unit 910 and the input image.
- a disparity map obtained by disposing a disparity template is combined with another disparity map, a disparity map that achieves high contrast can be obtained. As a result, it is possible to generate a parallax image capable of visually recognizing a high-quality stereoscopic image.
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Geometry (AREA)
- Computing Systems (AREA)
- Computer Graphics (AREA)
- General Physics & Mathematics (AREA)
- Signal Processing (AREA)
- Multimedia (AREA)
- Testing, Inspecting, Measuring Of Stereoscopic Televisions And Televisions (AREA)
- Image Processing (AREA)
- Image Analysis (AREA)
- Processing Or Creating Images (AREA)
Abstract
Description
(1)入力画像の左上隅を原点とし、横方向にx軸、縦方向にy軸を設定する。ただし、座標の設定方法はこれに限られるものではない。また入力画像の座標(x,y)の画素値をP(x,y)と表す。ここで画素値とは画像の明るさまたは色成分を表すものであればよく、例えば輝度や明度、特定の色チャンネルなどが該当する。
(2)デプスマップはマップの左上隅を原点とし、横方向にX軸、縦方向にY軸を設定する。ただし座標の設定方法はこれに限られるものではない。またマップの座標(X,Y)での画素値をZ(X,Y)と表す。このとき画素値は奥行き情報を表し、値が大きいほど奥行き(デプス)が大きいということになる。
(3)入力画像の座標とマップの座標は1対1に対応するようにしておく。特に記述しない限り、入力画像のサイズとマップのサイズは等しいものとし、入力画像の座標(x,y)とマップの座標(X,Y)は互いに対応するものとする。
(4)特に記述しない限り、入力画像の画素値を「画素値」と記述しその値域を[0,255](0以上255以下)とする。さらに、デプスマップの画素値を「奥行き値」と記述し、その値域を[0,255](0以上255以下)とする。
デプステンプレート20は、オブジェクトの種類ごとに用意される。デプステンプレートはオブジェクトの実際の3次元形状の概略形状を表現したものである。具体的にはオブジェクトを検出したい方向から眺めたときに、その奥行きを画素値として2次元画像で表現したものがデプステンプレートである。例えば検出対象オブジェクトが人物の上半身である場合、デプステンプレートは図 5に示すようなものとなる。図5では奥行き値が小さい(黒に近い)ほど、奥行きが小さく、奥行き値が大きい(白に近い)ほど、奥行きが大きい。
デプスマップ生成部200は、オブジェクト検出部100で検出されたオブジェクトの位置Ai(x,y)に対応するマップ上の位置Bi(X,Y)に、検出したオブジェクトに応じたデプステンプレートを配置する。これによりデプスマップを生成する。
(1)平均を用いる。すなわち、当該座標に重なった複数のデプステンプレートの当該座標における奥行き値の平均を用いる。
(2)最小値を用いる。すなわち、当該座標に重なったデプステンプレートの当該座標における奥行き値の最小値を用いる。
(3)重み付平均を用いる。当該座標に重なった複数のデプステンプレートの当該座標における奥行き値の重み付け平均を用いる。重みは、例えば奥行きが小さい値のテンプレートほど大きくする。
(4)中央値を用いる。すなわち、当該座標に重なったデプステンプレートの当該座標における奥行き値の中央値を用いる。
(5)オブジェクトの種類に順位をつけて、最も順位の高いオブジェクトに対応するテンプレートの奥行き値を用いる。同じ順位のオブジェクトが複数存在するときは、これらのオブジェクトに対して(1)~(4)の方法を適用する。
デプスマップ補正部300は、デプスマップ上の注目画素D(X,Y)とその周辺画素で重み付き平滑化を行うことによりデプスマップを補正する。周辺画素は注目画素に対して距離が近い範囲内に存在する画素である。例えば、注目画素と特定の一定距離範囲の内に存在する画素を示す。
視差画像生成部420は、入力画像とディスパリティマップから、生成したい枚数分の視差画像を生成する。
オブジェクト検出部101は、入力画像中から対象となるオブジェクトの全体または一部の位置、大きさ、向きを検出する。またオブジェクトの種類を検出する。大きさと向きを検出するという点以外は、実施例1のオブジェクト検出部100の動作と同様である。
デプステンプレート修正部500は、検出されたオブジェクトの大きさおよび向きに応じて、記憶部120から読み出したデプステンプレートを修正する。オブジェクトの大きさ、向きに応じてすべてのテンプレートを用意すると、その量は大きなものとなってしまう。そこで、オブジェクトの検出情報から、事前に用意されているデプステンプレートに修正を加える。
他デプスマップ30とは、他の手段などで与えられる入力画像に関するデプスマップのことである。この他デプスマップ30は、例えば、全体の構図などから背景の構図デプスなどを記述した他デプスマップが考えられる。また、動画像などの処理の場合、tフレーム前の画像に用いたデプスマップなどを他デプスマップとして用いることが考えられる。
デプスマップ合成部600は、デプスマップ補正部300で補正されたデプスマップと他デプスマップ30とを合成する。合成する他デプスマップは何種類あっても良い。
(1)各画素の奥行き値の平均値を用いる。
(2)各画素の奥行き値のうちの最大値を用いる。
(3)各画素の奥行き値のうちの最小値を用いる。
(4)各画素の奥行き値の重み付け平均を用いる。重みは、例えば奥行きが小さいほど大きくする。
(5)各画素の奥行き値の中央値を用いる。
画像入力部10は、処理の対象となる2次元画像を入力する。
オブジェクト検出部100は、入力画像からオブジェクトと、その種類、位置を検出する。
ディスパリティマップ生成部700は、図1のデプスマップ生成部200と同様に、オブジェクト検出部101で検出されたオブジェクトの位置Ai(x,y)と対応するディスパリティマップ上の位置Bi(X,Y)に、検出されたオブジェクトの種類に応じたディスパリティテンプレートを配置することでディスパリティマップを生成する。
ディスパリティマップ補正部800は、図1のデプスマップ補正部300と同様に、ディスパリティマップ上の注目画素E(X,Y)とその周辺画素で重み付き平滑化を行いディスパリティマップを補正する。このとき用いる重みは、注目画素Eに対応する入力画像中の対応注目画素C(x,y)と対応周辺画素(周辺画素に対応する入力画像中の画素)の画素値の分布に従って設定する。重みは基本的に対応注目画素と対応周辺画素の画素値の差分に応じて設定する。例えば差分が小さいほど重みを大きくし、差分が大きいほど重みを小さくするように設定すればよい。
3次元用画像生成部400は、視差画像生成部420において、ディスパリティマップ補正部800で得られたディスパリティマップと、入力画像とから、実施例1と同様にして、視差画像を生成する。
画像入力部10は、処理の対象となる2次元画像を入力する。
ディスパリティテンプレート修正部900は、図2のデプステンプレート修正部500と同様に、検出されたオブジェクトの大きさおよび向きに応じて、ディスパリティテンプレートを修正する。修正する方法はデプステンプレート修正部500と同様の方法を用いることができる。
ディスパリティマップ生成部700は、図2のデプスマップ生成部200と同様に、オブジェクト検出部101で検出されたオブジェクトの位置Ai(x,y)と対応するディスパリティマップ上での位置Bi(X,Y)に、ディスパリティテンプレート修正部900により修正されたディスパリティテンプレートを配置する。
他ディスパリティマップ50は、他の手段で与えられる入力画像に関するディスパリティマップである。他ディスパリティマップ50としては、例えば、全体の構図などから背景の構図ディスパリティなどを用いることができる。また、動画像などの処理の場合、tフレーム前の画像に用いたディスパリティマップなどを用いることができる。
ディスパリティマップ合成部910は、ディスパリティマップ補正部800で補正されたディスパリティマップと、他ディスパリティマップ50とを合成する。合成する他ディスパリティマップは何種類あっても良い。また、合成の方法は図2のデプスマップ合成部600の処理と同様の方法を用いることができる。
3次元用画像生成部400は、視差画像生成部420において、ディスパリティマップ合成部910で生成されたディスパリティマップと、入力画像とから視差画像を生成する。
Claims (7)
- 入力画像内のオブジェクトを検出する検出部と、
画素の奥行き値を記述する少なくとも1つのデプステンプレートのうち、検出された前記オブジェクトの種類に対応するデプステンプレートを選択し、前記入力画像内での前記検出されたオブジェクトの位置に従って選択されたデプステンプレートをデプスマップ上に配置することにより、前記入力画像における画素毎の奥行き値を記述する前記デプスマップを生成するデプスマップ生成部と、
前記デプスマップ内の少なくとも1つの注目画素と、周辺画素の重みを、前記注目画素に対応する前記入力画像内の対応注目画素の画素値と、前記周辺画素に対応する前記入力画像内の対応周辺画素の画素値との関係から算出し、注目画素の前記奥行き値と前記周辺画素の前記奥行き値との前記重みの重み付き和から前記注目画素の奥行き値を補正する補正部と、
前記補正部により補正されたデプスマップと前記入力画像とに基づき複数の視差画像を生成する画像生成部と、
を備えた画像処理装置。 - 前記検出部は、前記オブジェクトの大きさおよび向きの少なくとも一方を検出し、
前記デプスマップ生成部は、前記選択されたデプステンプレートを前記オブジェクトの大きさおよび向きの少なくとも一方に基づいて修正し、修正されたデプステンプレートを前記デプスマップ上に配置する
ことを特徴とする請求項1に記載の画像処理装置。 - 前記補正部は、
前記対応注目画素の画素値との差分が小さい前記対応周辺画素ほど、前記対応周辺画素に対応する周辺画素の重みが大きくなるように、前記周辺画素の重みを算出する
ことを特徴とする請求項2に記載の画像処理装置。 - 前記補正部は、
前記対応注目画素の画素値と前記対応周辺画素の画素値との差が、閾値より大きい対応周辺画素に対応する周辺画素の前記重みに0を割り当てると共に前記注目画素に対する前記重みを大きくする
ことを特徴とする請求項1に記載の画像処理装置。 - 前記補正部は、
前記対応注目画素と前記対応周辺画素の画素値のうち中央値を求め、前記対応周辺画素及び前記対応注目画素のうち、前記中央値に一致しない画素値を有する画素に対しては、前記重みに0を割り当て、前記中央値に一致する画素値を有する画素の前記奥行き値の平均を計算することにより、前記注目画素の奥行き値を補正する
ことを特徴とする請求項1に記載の画像処理装置。 - 前記補正されたデプスマップと、特定の与えられた他のデプスマップとを合成するデプスマップ合成部をさらに備え、
前記画像生成部は、合成されたデプスマップを用いて前記複数の視差画像を生成する
ことを特徴とする請求項1に記載の画像処理装置。 - 入力画像内のオブジェクトを検出する検出部と、
画素のディスパリティ値を記述する少なくとも1つのディスパリティテンプレートのうち、検出されたオブジェクトの種類に対応するディスパリティテンプレートを選択し、前記入力画像内での前記検出されたオブジェクトの位置に従って選択されたディスパリティテンプレートをディスパリティマップ上に配置することにより、前記入力画像における画素毎のディスパリティ値を記述する前記ディスパリティマップを生成するディスパリティマップ生成部と、
前記ディスパリティマップ内の少なくとも1つの注目画素と、周辺画素の重みを、前記注目画素に対応する前記入力画像内の対応注目画素の画素値と、前記周辺画素に対応する前記入力画像内の対応周辺画素の画素値との関係から算出し、前記注目画素の前記ディスパリティ値と前記周辺画素の前記ディスパリティ値との前記重みの重み付き和から前記注目画素のディスパリティ値を補正する補正部と、
前記補正部により補正されたディスパリティマップと、前記入力画像とに基づき複数の視差画像を生成する画像生成部と、
を備えた画像処理装置。
Priority Applications (4)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
JP2011503687A JP4966431B2 (ja) | 2009-09-18 | 2009-09-18 | 画像処理装置 |
PCT/JP2009/066444 WO2011033673A1 (ja) | 2009-09-18 | 2009-09-18 | 画像処理装置 |
CN2009801593209A CN102428501A (zh) | 2009-09-18 | 2009-09-18 | 图像处理装置 |
US13/234,448 US9053575B2 (en) | 2009-09-18 | 2011-09-16 | Image processing apparatus for generating an image for three-dimensional display |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
PCT/JP2009/066444 WO2011033673A1 (ja) | 2009-09-18 | 2009-09-18 | 画像処理装置 |
Related Child Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
US13/234,448 Continuation US9053575B2 (en) | 2009-09-18 | 2011-09-16 | Image processing apparatus for generating an image for three-dimensional display |
Publications (1)
Publication Number | Publication Date |
---|---|
WO2011033673A1 true WO2011033673A1 (ja) | 2011-03-24 |
Family
ID=43758299
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
PCT/JP2009/066444 WO2011033673A1 (ja) | 2009-09-18 | 2009-09-18 | 画像処理装置 |
Country Status (4)
Country | Link |
---|---|
US (1) | US9053575B2 (ja) |
JP (1) | JP4966431B2 (ja) |
CN (1) | CN102428501A (ja) |
WO (1) | WO2011033673A1 (ja) |
Cited By (10)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
WO2012169217A1 (ja) * | 2011-06-10 | 2012-12-13 | シャープ株式会社 | 映像生成装置、映像表示装置、テレビ受像装置、映像生成方法及びコンピュータプログラム |
JP2013058956A (ja) * | 2011-09-09 | 2013-03-28 | Sony Corp | 情報処理装置、情報処理方法、プログラム及び情報処理システム |
WO2013044418A1 (en) * | 2011-09-30 | 2013-04-04 | Intel Corporation | Human head detection in depth images |
EP2536160A3 (en) * | 2011-06-14 | 2013-09-11 | Samsung Electronics Co., Ltd. | Display system with image conversion mechanism and method of operation thereof |
WO2013145554A1 (ja) * | 2012-03-29 | 2013-10-03 | パナソニック株式会社 | 画像処理装置及び画像処理方法 |
WO2014030630A1 (ja) * | 2012-08-20 | 2014-02-27 | 株式会社日本自動車部品総合研究所 | 視差マップを生成する装置及びその方法 |
KR20140088200A (ko) * | 2011-11-02 | 2014-07-09 | 구글 인코포레이티드 | 예제 유사 이미지와 연관된 예제 근사 깊이-맵을 사용하여 입력 이미지에 대한 깊이-맵 생성 |
JPWO2012153447A1 (ja) * | 2011-05-11 | 2014-07-28 | パナソニック株式会社 | 画像処理装置、映像処理方法、プログラム、集積回路 |
JP2016115332A (ja) * | 2014-12-15 | 2016-06-23 | キヤノン株式会社 | 画像処理装置およびその制御方法ならびにプログラム |
KR101870764B1 (ko) * | 2011-06-14 | 2018-06-25 | 삼성전자주식회사 | 영상 변환 메커니즘을 활용한 디스플레이 장치 및 그 동작 방법 |
Families Citing this family (23)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JP5627498B2 (ja) * | 2010-07-08 | 2014-11-19 | 株式会社東芝 | 立体画像生成装置及び方法 |
JP5699609B2 (ja) * | 2011-01-06 | 2015-04-15 | ソニー株式会社 | 画像処理装置および画像処理方法 |
JP2012186781A (ja) | 2011-02-18 | 2012-09-27 | Sony Corp | 画像処理装置および画像処理方法 |
US20130094753A1 (en) * | 2011-10-18 | 2013-04-18 | Shane D. Voss | Filtering image data |
EP2786580B1 (en) * | 2011-11-30 | 2015-12-16 | Fraunhofer-Gesellschaft zur Förderung der angewandten Forschung e.V. | Spatio-temporal disparity-map smoothing by joint multilateral filtering |
KR101356544B1 (ko) * | 2012-03-29 | 2014-02-19 | 한국과학기술원 | 3차원 입체영상 생성 방법 및 장치 |
KR101889952B1 (ko) * | 2012-06-18 | 2018-08-20 | 엘지디스플레이 주식회사 | 신뢰도 기반의 스테레오 매칭 방법 및 장치 |
JP5974376B2 (ja) * | 2012-12-28 | 2016-08-23 | 株式会社日立製作所 | 医用画像処理装置および画像処理方法 |
JP2014220704A (ja) * | 2013-05-09 | 2014-11-20 | ソニー株式会社 | 画像処理装置および画像処理方法 |
US9552633B2 (en) * | 2014-03-07 | 2017-01-24 | Qualcomm Incorporated | Depth aware enhancement for stereo video |
US9998724B2 (en) * | 2014-12-15 | 2018-06-12 | Canon Kabushiki Kaisha | Image processing apparatus and method for processing three-dimensional data that describes a space that includes physical objects |
US10451714B2 (en) | 2016-12-06 | 2019-10-22 | Sony Corporation | Optical micromesh for computerized devices |
US10536684B2 (en) | 2016-12-07 | 2020-01-14 | Sony Corporation | Color noise reduction in 3D depth map |
US10181089B2 (en) | 2016-12-19 | 2019-01-15 | Sony Corporation | Using pattern recognition to reduce noise in a 3D map |
US10178370B2 (en) | 2016-12-19 | 2019-01-08 | Sony Corporation | Using multiple cameras to stitch a consolidated 3D depth map |
US10495735B2 (en) | 2017-02-14 | 2019-12-03 | Sony Corporation | Using micro mirrors to improve the field of view of a 3D depth map |
US10795022B2 (en) | 2017-03-02 | 2020-10-06 | Sony Corporation | 3D depth map |
US10979687B2 (en) * | 2017-04-03 | 2021-04-13 | Sony Corporation | Using super imposition to render a 3D depth map |
US10484667B2 (en) | 2017-10-31 | 2019-11-19 | Sony Corporation | Generating 3D depth map using parallax |
US10549186B2 (en) | 2018-06-26 | 2020-02-04 | Sony Interactive Entertainment Inc. | Multipoint SLAM capture |
CN110288543B (zh) * | 2019-06-21 | 2021-11-30 | 北京迈格威科技有限公司 | 一种深度图像保边处理方法和装置 |
WO2021035746A1 (zh) * | 2019-08-30 | 2021-03-04 | 深圳市大疆创新科技有限公司 | 图像处理方法、装置和可移动平台 |
CN113077504B (zh) * | 2021-04-12 | 2021-11-12 | 中国电子科技集团公司第二十八研究所 | 基于多粒度特征匹配大场景深度图生成方法 |
Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JP2003047027A (ja) * | 2001-07-31 | 2003-02-14 | Canon Inc | 立体画像形成システム、立体画像形成方法、プログラムおよび記憶媒体 |
WO2004071102A1 (ja) * | 2003-01-20 | 2004-08-19 | Sanyo Electric Co,. Ltd. | 立体視用映像提供方法及び立体映像表示装置 |
Family Cites Families (18)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US5339386A (en) * | 1991-08-08 | 1994-08-16 | Bolt Beranek And Newman Inc. | Volumetric effects pixel processing |
US6947587B1 (en) * | 1998-04-21 | 2005-09-20 | Hitachi, Ltd. | Defect inspection method and apparatus |
US6868191B2 (en) * | 2000-06-28 | 2005-03-15 | Telefonaktiebolaget Lm Ericsson (Publ) | System and method for median fusion of depth maps |
US6633297B2 (en) * | 2000-08-18 | 2003-10-14 | Hewlett-Packard Development Company, L.P. | System and method for producing an antialiased image using a merge buffer |
US7113634B2 (en) | 2001-07-31 | 2006-09-26 | Canon Kabushiki Kaisha | Stereoscopic image forming apparatus, stereoscopic image forming method, stereoscopic image forming system and stereoscopic image forming program |
US7221809B2 (en) * | 2001-12-17 | 2007-05-22 | Genex Technologies, Inc. | Face recognition system and method |
CN1641702A (zh) * | 2004-01-13 | 2005-07-20 | 邓兴峰 | 由平面图象设计立体图象的方法 |
WO2006106465A2 (en) | 2005-04-07 | 2006-10-12 | Nxp B.V. | Method and device for three-dimentional reconstruction and rendering |
ES2676055T5 (es) * | 2006-03-31 | 2022-08-03 | Koninklijke Philips Nv | Receptor de imagen eficaz para múltiples vistas |
CN101479765B (zh) * | 2006-06-23 | 2012-05-23 | 图象公司 | 对2d电影进行转换用于立体3d显示的方法和系统 |
US8416479B2 (en) * | 2006-10-26 | 2013-04-09 | Seereal Technologies S.A. | Compact holographic display device |
KR100803220B1 (ko) * | 2006-11-20 | 2008-02-14 | 삼성전자주식회사 | 다중 파이프라인의 3차원 그래픽스 렌더링 방법 및 장치 |
US20090002368A1 (en) * | 2007-06-26 | 2009-01-01 | Nokia Corporation | Method, apparatus and a computer program product for utilizing a graphical processing unit to provide depth information for autostereoscopic display |
CN101398933B (zh) * | 2008-10-17 | 2010-09-15 | 北京大学 | 一种从图像中恢复三维几何信息的方法 |
CN101404091B (zh) * | 2008-11-07 | 2011-08-31 | 重庆邮电大学 | 基于两步形状建模的三维人脸重建方法和系统 |
KR101506926B1 (ko) * | 2008-12-04 | 2015-03-30 | 삼성전자주식회사 | 깊이 추정 장치 및 방법, 및 3d 영상 변환 장치 및 방법 |
BRPI1007451A2 (pt) * | 2009-01-30 | 2016-09-13 | Thomson Licensing | codificação de mapas de profundidade |
US9582889B2 (en) * | 2009-07-30 | 2017-02-28 | Apple Inc. | Depth mapping based on pattern matching and stereoscopic information |
-
2009
- 2009-09-18 CN CN2009801593209A patent/CN102428501A/zh active Pending
- 2009-09-18 JP JP2011503687A patent/JP4966431B2/ja not_active Expired - Fee Related
- 2009-09-18 WO PCT/JP2009/066444 patent/WO2011033673A1/ja active Application Filing
-
2011
- 2011-09-16 US US13/234,448 patent/US9053575B2/en not_active Expired - Fee Related
Patent Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JP2003047027A (ja) * | 2001-07-31 | 2003-02-14 | Canon Inc | 立体画像形成システム、立体画像形成方法、プログラムおよび記憶媒体 |
WO2004071102A1 (ja) * | 2003-01-20 | 2004-08-19 | Sanyo Electric Co,. Ltd. | 立体視用映像提供方法及び立体映像表示装置 |
Cited By (19)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JPWO2012153447A1 (ja) * | 2011-05-11 | 2014-07-28 | パナソニック株式会社 | 画像処理装置、映像処理方法、プログラム、集積回路 |
JP2013004989A (ja) * | 2011-06-10 | 2013-01-07 | Sharp Corp | 映像生成装置、映像表示装置、テレビ受像装置、映像生成方法及びコンピュータプログラム |
WO2012169217A1 (ja) * | 2011-06-10 | 2012-12-13 | シャープ株式会社 | 映像生成装置、映像表示装置、テレビ受像装置、映像生成方法及びコンピュータプログラム |
EP2536160A3 (en) * | 2011-06-14 | 2013-09-11 | Samsung Electronics Co., Ltd. | Display system with image conversion mechanism and method of operation thereof |
KR101870764B1 (ko) * | 2011-06-14 | 2018-06-25 | 삼성전자주식회사 | 영상 변환 메커니즘을 활용한 디스플레이 장치 및 그 동작 방법 |
JP2013058956A (ja) * | 2011-09-09 | 2013-03-28 | Sony Corp | 情報処理装置、情報処理方法、プログラム及び情報処理システム |
WO2013044418A1 (en) * | 2011-09-30 | 2013-04-04 | Intel Corporation | Human head detection in depth images |
US9996731B2 (en) | 2011-09-30 | 2018-06-12 | Intel Corporation | Human head detection in depth images |
US9111131B2 (en) | 2011-09-30 | 2015-08-18 | Intelcorporation | Human head detection in depth images |
KR20140088200A (ko) * | 2011-11-02 | 2014-07-09 | 구글 인코포레이티드 | 예제 유사 이미지와 연관된 예제 근사 깊이-맵을 사용하여 입력 이미지에 대한 깊이-맵 생성 |
JP2014532941A (ja) * | 2011-11-02 | 2014-12-08 | グーグル・インコーポレーテッド | 例示的な類似画像と関連した例示的な近似奥行きマップを用いた、入力画像に対する奥行きマップの生成 |
KR102148538B1 (ko) * | 2011-11-02 | 2020-08-26 | 구글 엘엘씨 | 예제 유사 이미지와 연관된 예제 근사 깊이-맵을 사용하여 입력 이미지에 대한 깊이-맵 생성 |
JPWO2013145554A1 (ja) * | 2012-03-29 | 2015-12-10 | パナソニックIpマネジメント株式会社 | 画像処理装置及び画像処理方法 |
US9495806B2 (en) | 2012-03-29 | 2016-11-15 | Panasonic Intellectual Property Management Co., Ltd. | Image processing apparatus and image processing method |
WO2013145554A1 (ja) * | 2012-03-29 | 2013-10-03 | パナソニック株式会社 | 画像処理装置及び画像処理方法 |
JP2014038546A (ja) * | 2012-08-20 | 2014-02-27 | Nippon Soken Inc | 視差マップ生成装置および視差マップ生成装置用のプログラム |
US9361668B2 (en) | 2012-08-20 | 2016-06-07 | Denso Corporation | Method and apparatus for generating disparity map |
WO2014030630A1 (ja) * | 2012-08-20 | 2014-02-27 | 株式会社日本自動車部品総合研究所 | 視差マップを生成する装置及びその方法 |
JP2016115332A (ja) * | 2014-12-15 | 2016-06-23 | キヤノン株式会社 | 画像処理装置およびその制御方法ならびにプログラム |
Also Published As
Publication number | Publication date |
---|---|
US20120069009A1 (en) | 2012-03-22 |
JPWO2011033673A1 (ja) | 2013-02-07 |
US9053575B2 (en) | 2015-06-09 |
JP4966431B2 (ja) | 2012-07-04 |
CN102428501A (zh) | 2012-04-25 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
JP4966431B2 (ja) | 画像処理装置 | |
KR102003015B1 (ko) | 광류를 사용한 중간 뷰의 생성 | |
US9117295B2 (en) | Refinement of depth maps by fusion of multiple estimates | |
US8655055B2 (en) | Method, system and computer program product for converting a 2D image into a 3D image | |
JP6094863B2 (ja) | 画像処理装置、画像処理方法、プログラム、集積回路 | |
JP5673032B2 (ja) | 画像処理装置、表示装置、画像処理方法及びプログラム | |
JP2013005259A (ja) | 画像処理装置、および画像処理方法、並びにプログラム | |
WO2012157540A1 (ja) | 画像処理装置、画像処理方法、およびプログラム | |
US8094148B2 (en) | Texture processing apparatus, method and program | |
JP6585938B2 (ja) | 立体像奥行き変換装置およびそのプログラム | |
WO2012117706A1 (ja) | 映像処理装置、映像処理方法、プログラム | |
Bleyer et al. | Temporally consistent disparity maps from uncalibrated stereo videos | |
CN107016730A (zh) | 一种虚拟现实与真实场景融合的装置 | |
JPWO2011030399A1 (ja) | 画像処理方法及び装置 | |
CN106981100A (zh) | 一种虚拟现实与真实场景融合的装置 | |
KR20110112143A (ko) | Ldi 기법 깊이맵을 참조한 2d 동영상의 3d 동영상 전환방법 | |
US20120121163A1 (en) | 3d display apparatus and method for extracting depth of 3d image thereof | |
WO2015186284A1 (ja) | 画像処理装置、画像処理方法及びプログラム | |
KR101121979B1 (ko) | 입체 영상 변환 방법 및 입체 영상 변환 장치 | |
KR101632514B1 (ko) | 깊이 영상 업샘플링 방법 및 장치 | |
CN103945206A (zh) | 一种基于相似帧比较的立体图像合成系统 | |
JP2014072809A (ja) | 画像生成装置、画像生成方法、画像生成装置用プログラム | |
Islam et al. | Warping-based stereoscopic 3d video retargeting with depth remapping | |
Seitner et al. | Trifocal system for high-quality inter-camera mapping and virtual view synthesis | |
KR101192121B1 (ko) | 양안시차 및 깊이 정보를 이용한 애너그리프 영상 생성 방법 및 장치 |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
WWE | Wipo information: entry into national phase |
Ref document number: 200980159320.9 Country of ref document: CN |
|
WWE | Wipo information: entry into national phase |
Ref document number: 2011503687 Country of ref document: JP |
|
121 | Ep: the epo has been informed by wipo that ep was designated in this application |
Ref document number: 09849532 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: 09849532 Country of ref document: EP Kind code of ref document: A1 |