KR102112019B1 - Apparatus and method for generating intermediate view image - Google Patents

Abstract

An apparatus for generating a mid-view image of a stereoscopic image, the feature point detector detects a contour line from a left-view image and a right-view image, and detects a feature point from the detected contour line, for the feature point of the left-view image and the right-view image An intermediate point image generating apparatus is provided which includes a correlation point detector for detecting a correlation point, and a synthesis unit for generating an intermediate point image based on disparity information of the feature point and the correlation point.

Description

Apparatus and method for generating mid-view images {APPARATUS AND METHOD FOR GENERATING INTERMEDIATE VIEW IMAGE}

The present description relates to an apparatus and method for generating a mid-view image of a stereoscopic image.

As a method of generating the mid-view image, a method of using uniform parallax information for all areas of the image, a method of using parallax information for all pixels of the image, a method of using parallax information for some pixels of the image There is this.

The method of using uniform parallax information for all regions of the image is to designate a rectangular region for a part or all of the left and right images, to find a square region at the middle point by 3D transformation of the square region, An image is generated by morphing an image from a rectangular region to a rectangular region of the middle viewpoint. Since this method assumes that parallax information is uniform in all areas of a rectangle, there is a problem that it can be used only for images without complicated changes.

As a method of utilizing parallax information for all pixels of the image, a parallax map is generated by obtaining parallax information for all pixels of the left and right images, and an intermediate viewpoint image is generated by referring to the parallax map and the left and right images. Specifically, the left and right images are projected and combined into a 3D space with reference to the corresponding parallax map, and then each pixel information in the 3D space is projected back to a designated intermediate position of a plane in which the left and right images exist to generate an intermediate viewpoint image. .

The above method can obtain an intermediate point of view even for a complex image, but has a problem in that the overall quality of the intermediate point of view generated depends on the quality of the parallax map used. When the parallax map is incomplete, each pixel of the left and right images is projected to the wrong location in the 3D space, and then projected to the wrong location in the mid-view plane, causing a hole where the original pixel will be located, and the pixel value at the wrongly projected location. Disturbed. This method has a problem in that it is difficult to obtain a high-quality parallax map because it requires a long time to calculate parallax information for all pixels of the left and right images, and noise containing false parallax information is prone to occur.

The method of obtaining parallax information for a pixel at a part of the image rather than the whole has a problem of minimizing finding a required point and minimizing an error of parallax information obtained at each point.

One embodiment provides an apparatus for generating a mid-view image of a stereoscopic image.

Another embodiment provides a method of generating a mid-view image of a stereoscopic image.

According to an embodiment of the present invention, an apparatus for generating a mid-view image of a stereoscopic image includes a feature point detection unit that detects a contour line from a left-view image and a right-view image, and detects a feature point from the detected contour line, the left-view image and the right view point It includes a correlation point detection unit for detecting a correlation point for the feature point of the image, and a synthesis unit for generating an intermediate view image based on the disparity information of the feature point and the correlation point.

The correlation point detection unit detects an inverse correlation point with respect to the correlation point, removes an abnormal correlation point from the correlation points based on the position of the inverse correlation point and the feature point, selects an effective correlation point, and the synthesis unit includes the The mid-view image may be generated using disparity information between a feature point and the effective correlation point.

The correlation point detection unit may determine that the detected correlation point is abnormal when the positions of the inverse correlation point and the feature point are different, and remove the correlation point determined as abnormal.

The correlation point detector separates a feature point included in a background and a foreground among feature points of the left-view image and the right-view image, and a correlation point for a feature point included in the background and a correlation point for a feature point included in the foreground Can be detected and integrated.

According to another embodiment, a method of generating a mid-view image of a stereoscopic image includes detecting a contour line from a left-view image and a right-view image, detecting a feature point from the contour line, and detecting a correlation point with respect to the feature point And generating an intermediate view image based on disparity information of the feature point and the correlation point.

After the step of detecting the correlation point, detecting an inverse correlation point with respect to the correlation point, and selecting an effective correlation point by removing an abnormal correlation point from the correlation points based on the position of the inverse correlation point and the feature point The generating of the mid-view image may include generating a mid-view image using disparity information between the feature point and the effective correlation point.

In selecting the effective correlation point, if the positions of the inverse correlation point and the detected feature point are different, the detected correlation point is determined to be abnormal, and the correlation point determined to be abnormal can be removed.

The detecting of the correlation points may include separating a feature point included in a background and a foreground among feature points of the left and right view images, and a correlation point for the feature points included in the background and included in the foreground. And detecting and integrating the correlation points for the feature points.

The detecting and integrating the correlation points may include detecting a correlation point for the feature points included in the background and a correlation point for the feature points included in the foreground, respectively for the correlation points included in the background and the foreground Detecting an inverse correlation point of the step, determining the abnormal correlation point when the position of the inverse correlation point and the background and the feature point included in the foreground are different, and removing the abnormal correlation point, and the background from which the abnormal correlation point is removed, and And integrating the correlation points included in the foreground.

According to another embodiment, an apparatus for generating a mid-view image of a stereoscopic image includes a processor and a memory, and the processor executes a program stored in the memory to detect contours in a left-view image and a right-view image, Detecting a feature point from the detected contour, detecting a correlation point for the feature point of the left-view image and the right-view image, and generating an intermediate view image based on disparity information of the feature point and the correlation point Perform the steps.

By obtaining parallax information for a valid feature point without obtaining parallax information from the entire image, a high-quality mid-view image can be generated.

In addition, since parallax information is not obtained from the entire image, an intermediate view image can be generated at a high speed.

1 is a view showing a left view image (a) and a right view image (b).
2 is a view showing a parallax map derived through a method of obtaining parallax information in all regions of an image.
3 is a view showing an intermediate viewpoint image generated through a method of obtaining parallax information in all regions of an image.
4 is a view showing a left view image (a) and a right view image (b).
5 is a view showing parallax information.
FIG. 6 is a diagram illustrating an intermediate view image generated through a method of obtaining parallax information from some pixels of an image.
7 is a block diagram of an apparatus for generating an intermediate viewpoint according to an embodiment.
8 is a diagram illustrating a gray scale image converted by an intermediate viewpoint image generating apparatus according to an embodiment.
9 is a diagram showing feature points detected in a gray scale image.
FIG. 10 is a diagram showing contours detected by performing canny filtering performed by the mid-view image generating apparatus according to an embodiment.
11 is a view showing the feature points detected in the contour.
12 is a view showing an image obtained by combining a Sobel filtered image and a Canny filtered image.
13 is a diagram showing feature points detected in a canny-filtered image.
14 and 15 are diagrams showing a correlation point and a parallax vector detected by an intermediate viewpoint image generating apparatus according to an embodiment.
16 to 19 are views for explaining an abnormal correlation point.
20 and 21 are views showing feature points detected by separating a background or a foreground.
22 is a view showing an intermediate viewpoint image generated through morphing.
23 and 24 are flowcharts illustrating a method for generating an intermediate viewpoint image according to an embodiment.
25 is a flowchart illustrating a step of detecting a correlation point according to an embodiment.
26 is a block diagram illustrating an apparatus for generating an intermediate viewpoint image according to another embodiment.

Hereinafter, exemplary embodiments of the present invention will be described in detail with reference to the accompanying drawings so that those skilled in the art to which the present invention pertains can easily practice. However, the present invention can be implemented in many different forms and is not limited to the embodiments described herein. In addition, in order to clearly describe the present invention in the drawings, parts irrelevant to the description are omitted, and like reference numerals are assigned to similar parts throughout the specification.

Throughout the specification, when a part “includes” a certain component, it means that the component may further include other components, not to exclude other components, unless otherwise stated.

The method of obtaining the mid-view image between the captured left and right images is composed of a method of obtaining disparity information between the left and right images and a method of synthesizing the left and right images using the disparity information.

The method of obtaining parallax information is divided into a method of obtaining parallax information in all regions of the image and a method of obtaining parallax information in some regions of the image.

FIG. 1 is a view showing a left image (a) and a right image (b), and FIG. 2 is a view showing a parallax map derived through a method of obtaining parallax information in all regions of the image, and FIG. 3 is a whole region of the image A diagram showing an intermediate viewpoint image generated through a method of obtaining parallax information in.

A method of obtaining parallax information in all regions of an image is a method of calculating parallax by using all pixels of the left and right images as a feature point, and obtaining a correlation point corresponding to the feature point. In this method, a disparity map having the same size as an image is obtained through stereo matching technology. This method takes a lot of calculation time because the parallax is obtained from all the pixels in the left and right images, and it is difficult to increase the quality of the mid-view image because it is impossible to obtain an accurate parallax value for all pixels.

Referring to FIG. 2, it can be seen that the parallax map contains a lot of noise indicating incorrect parallax information. Due to this noise, a phenomenon in which the contour part of the person is broken in the mid-view image as shown in FIG. 3 occurs.

FIG. 4 is a view showing a left image (a) and a right image (b), FIG. 5 is a view showing parallax information, and FIG. 6 is a view of an intermediate viewpoint generated through a method of obtaining parallax information from some pixels of the image It is a figure to show.

The method of obtaining parallax information in some areas of an image is to find several feature points in one image of a left-view image or a right-view image, find a correlation point corresponding to each feature point in the other image, and then find the distance between the feature point and the correlation point ( Pixel). Referring to FIG. 5, the feature point may be a point for distinguishing edges or the like, for example, a finger tip point, and a correlation point may be a corresponding finger tip point of the other image.

As a method of finding the correlation point, after finding all the feature points in each of the left and right images, a feature point matching technique using the feature point of the other image corresponding to each feature point as a correlation point and as shown in FIG. Likewise, there is a method of finding a correlation point using optical flow technology.

If the length of the horizontal line connecting the correlation points corresponding to the feature points is long, the parallax is large, and if the length of the line is short, the parallax is small.

Referring to FIG. 5, it can be seen that the parallax is large for a person close to the camera and the parallax is small for a background far from the camera. Referring to FIG. 6, it can be seen that a part in which the parallax between the feature point and the correlation point is not correct appears blurry or overlapped in the mid-view image.

7 is a block diagram of an apparatus for generating an intermediate viewpoint according to an embodiment, and FIG. 8 is a diagram illustrating a gray scale image converted by the apparatus for generating an intermediate viewpoint according to an embodiment. FIG. 10 is a diagram showing a feature point detected in a scale image, and FIG. 10 is a diagram showing a contour line detected through Canny filtering performed by the mid-view image generating apparatus according to an embodiment, and FIG. 11 is a line detected in the contour line of FIG. 10. FIG. 12 is a diagram showing a composite of a Sobel filtered image and a Canny filtered image, and FIG. 13 is a diagram showing the feature points detected in the Canny filtered image.

Referring to FIG. 7, the mid-view image generating apparatus 1 according to an embodiment includes a feature point detection unit 30, a correlation point detection unit 50, and a synthesis unit 70.

The feature point detection unit 30 detects a contour line or an important point in a left-view image and a right-view image, and detects a feature point in the detected contour or important point. Not all of the feature points found in the video are important. Feature points detected in an image that includes patterned clothes or wallpaper have low importance because there is little parallax change near the feature points. On the other hand, the feature points found in a person's outline are highly important because the parallax changes are large near the feature points.

The feature point detection unit 30 converts a left-view image and a right-view image into a gray-scale image and performs filtering such as Sobel filtering, Laplacian filtering, and Canny filtering, The filtered images are combined into one to detect contours or important points. The image converted to the gray scale image is shown in FIG. 8, and when the feature point is detected in this state, it can be confirmed that the feature point is relatively evenly distributed as shown in FIG. 9. When Canny filtering is performed on a gray scale image, an image in which only contours appear as shown in FIG. 10 is generated.

The feature point detection unit 30 may detect a feature point from an image detected through filtering. The image in which the feature point is detected in the image is shown in FIG. 11.

The feature point detection unit 30 may detect a feature point in an image obtained by synthesizing a Sobel filtered image including a detailed gradient and a canny filtered image based on a contour as an embodiment. The combined image of the Sobel filtered image and the Canny filtered image is shown in FIG. 12.

If the canny-filtered image is used, the feature points can be detected mainly around the outline, but a quality problem of the outline may occur. If the threshold is set low when performing canny filtering, contours and feature points can be detected abundantly. However, since a double line occurs around the actual contour or noise segments are generated, a problem occurs in that feature points are not detected in the actual contour but are detected in the vicinity of the contour. The characteristic correlation points near the contour are difficult to detect because the parallax change is not clear.

When the threshold value is set high when performing canny filtering, only the contours are detected with certainty, so the feature points can be detected neatly as shown in FIG. 13. However, there is a problem in that a feature point is not detected in a portion where the outline is not detected, such as the square box portion in FIG. 13.

In order to solve the above problems, the feature point detection unit 30 approaches stepwise when obtaining a contour stone, first obtains a feature point of an important contour, and then detects a feature point of a less important contour. As an embodiment, the threshold value is set to the first value and the second value to perform canny filtering with each threshold value, and the feature point detected from the canny filtering by the first value and the feature point detected from the canny filtering by the second value To incorporate. The first value may be larger than the second value, and the first value and the second value are values that may vary depending on the setting.

The feature point detection unit 30 may perform several steps of performing canny filtering and detecting the feature points by variously setting a threshold value as one embodiment, and combine the feature points detected in each step into one. The feature point detection unit 30 maintains the feature points detected in a state where the threshold is set high, and selectively selects feature points detected in a state where the threshold value is set low near the feature points detected in the state where the threshold value is set high. Can be added. Through this, it is possible to detect feature points that are effective, have high importance, and have high quality.

14 and 15 are diagrams illustrating correlation points and parallax vectors detected by the mid-view image generating apparatus according to an embodiment, and FIGS. 16 to 19 are views illustrating abnormal correlation points, and FIGS. 21 is a view showing the feature points detected by separating the background or the foreground, and FIG. 22 is a view showing the mid-view image generated through morphing.

The correlation point detection unit 50 detects a correlation point for the feature points of the left-view image and the right-view image. The correlation point detection unit 50 detects a correlation point for an effective feature point detected through the feature point detection unit 30.

The correlation point can be detected through a method of calculating optical flow in an image. This method calculates to which point the feature point of one image of the left-view image or the right-view image moves to the other image, where the moved point becomes a correlation point.

14 and 15, a change in movement from the feature point 100 to the correlation point 200 may be represented by a parallax vector 300. Correlation points corresponding to the feature points obtained from the left view image may be detected in the right view image, and correlation points corresponding to the feature points obtained from the right view image may be detected from the left view image. By adding the disparity information between the feature point and the correlation point obtained from the left-view image and the right-view image, richer disparity information can be calculated.

In general, although correlation points for most of the feature points can be accurately obtained through a method of calculating the optical flow, a problem in which abnormal correlation points are detected for some feature points may occur.

Referring to the square box of FIG. 16, it can be confirmed that the correlation point for the single feature point was generated at the wrong location. Referring to the square box of FIG. 17, it can be seen that an abnormal correlation point group is detected as a whole with respect to a group of feature points in a partial region. Although the correlation points of the tree of FIG. 16 were relatively accurately detected, it can be confirmed that some correlation points of the tree of FIG. 17 were abnormally detected. This problem of FIG. 17 most often occurs in the background near the contours of objects with very large parallax.

The problem in which the correlation point for the single feature point is generated at the wrong location can be visually confirmed and removed when the distance between the correlation point at the wrong location and the feature point is long, as shown in the square box in FIG. 16, but when the distance is short, it can be visually observed. Difficult to check and remove. If all the correlation points above a certain distance are filtered out, the correlation points corresponding to the feature points placed on the contour of the person close to the camera are all filtered due to the long distance as shown in FIG. 18. If the distance condition is relaxed, abnormal correlations may not be eliminated.

In order to solve the problem that a correlation point for a single feature point is generated in the wrong location, the correlation point detection unit 50 detects an inverse correlation point for the detected correlation point, and determines the position of the feature point detected by the inverse correlation point and the feature point detection unit 30. The effective correlation point is selected by removing the abnormal correlation point from the detected correlation points, and the synthesis unit 70 uses the parallax information of the effective correlation point and the effective feature points detected by the feature point detection unit 30 to generate an intermediate viewpoint image. To create.

Specifically, referring to FIGS. 14 and 15, the correlation point detection unit 50 uses all the detected correlation points 200 as a feature point 100 to detect an inverse correlation point of the opposite image, and then detects the feature point detection unit 30 The position of the first feature point 100 and the position of the inverse correlation point are compared. If the correlation point is correctly detected, since the inverse correlation point is also accurately detected, there is no difference between the position of the feature point and the position of the inverse correlation point.

The correlation point detection unit 50 determines that the detected correlation point is abnormal and removes the correlation point determined as abnormal if the positions of the inverse correlation points and the feature points detected by the feature point detection unit 30 are different. Through this, as shown in FIG. 19, abnormal correlation points are effectively removed.

In order to solve a problem in which a group of feature points in a certain region is detected as a group of abnormal correlation points, the correlation point detection unit 50 is included in the background and the foreground of the feature points of the left and right view images as shown in FIGS. 20 and 21. The feature points are separated, and the correlation points for the feature points included in the background and the correlation points for the feature points included in the foreground are detected and integrated.

Specifically, the correlation point detection unit 50 separates feature points corresponding to a background and a foreground from a left-view image and a right-view image. Foreground refers to main objects such as people. Each correlation point for the background and foreground feature points is detected and integrated.

In order to improve the quality of the correlation point, the correlation point detection unit 50 separates the feature points included in the background and the foreground among the feature points of the left and right view images, and then determines each of the inverse correlation points of the correlation points included in the background and the foreground. To detect. By comparing the position of each inverse correlation point with each feature point included in the background and foreground, it is determined whether the correlation point included in the background and foreground is normal, and if the position of the inverse correlation point and the feature points included in the background and foreground is different, it is used as an abnormal correlation point. The abnormal correlation point is removed by judging. Incorporate the remaining correlations included in the background and foreground where the abnormal correlations were removed.

The synthesizing unit 70 generates an intermediate view image using disparity information between the effective feature point detected through the feature point detection unit 30 and the effective correlation point detected by the correlation point detection unit 50. The synthesizing unit 70 generates a mid-view image by performing morphing or warping using parallax information.

Referring to FIG. 22, the synthesis unit 70 morphs between the triangular patch obtained from the detected feature point and the triangular patch obtained from the correlation point to generate an intermediate view image.

23 is a flowchart illustrating a method for generating an intermediate viewpoint image according to an embodiment.

Referring to FIG. 23, a method for generating an intermediate viewpoint image according to an embodiment includes: detecting a contour in a left-view image and a right-view image (S100), detecting a feature point in the contour (S200), and correlating the feature points It includes the step of detecting the point (S300), and generating a mid-view image based on the disparity information of the feature point and the correlation point (S400).

24 is a flowchart illustrating a method for generating an intermediate viewpoint image according to an embodiment.

Referring to FIG. 24, in the method of generating an intermediate viewpoint image, after the step of detecting a correlation point (S300), an inverse correlation point for the correlation point is detected, and an abnormal correlation point is removed from the correlation points based on the positions of the inverse correlation point and the feature point. By selecting an effective correlation point (S310) may be further included. In the generating of the mid-view image (S400), the mid-view image is generated using the disparity information of the detected feature point and the effective correlation point.

In step S310 of selecting an effective correlation point, if the positions of the inverse correlation point and the detected feature points are different, the detected correlation point is determined to be abnormal, and the correlation point determined to be abnormal is removed.

25 is a flowchart illustrating a step of detecting a correlation point according to an embodiment.

Referring to FIG. 25, the step of detecting a correlation point (S300) includes: separating a feature point included in a background and a foreground among feature points of a left-view image and a right-view image (S320), and for a feature point included in the background It may include the step of detecting and integrating the correlation point for the correlation point and the feature point included in the foreground (S330).

The step of detecting and integrating the correlation point (S330) includes: detecting a correlation point for a feature point included in the background and a correlation point for a feature point included in the foreground (S332), for a correlation point included in the background and the foreground Detecting each inverse correlation point (S333), if the position of the inverse correlation point and the background and feature points included in the foreground is different, determining the abnormal correlation point and removing the abnormal correlation point (S334), and the background in which the abnormal correlation point is removed And integrating the correlation points included in the foreground (S335).

In the generating of the mid-view image (S400), the mid-view image may be generated by performing morphing or warping based on parallax information.

Detecting a contour (S100), Detecting a feature point from the contour (S200), Detecting a correlation point for the feature point (S300), and generating a mid-view image based on disparity information between the feature point and the correlation point Step S400 is the same as the operation contents of the feature point detection unit 30, the correlation point detection unit 50, and the synthesis unit 70 described above, detailed description thereof will be omitted.

26 is a block diagram illustrating an apparatus for generating an intermediate viewpoint image according to an embodiment.

Referring to FIG. 26, the apparatus for generating a mid-view image according to an embodiment may be implemented as a computer system, for example, a computer-readable medium. The computer system 700 includes at least one of a processor 710, a memory 730, a user interface input device 760, a user interface output device 770, and a storage device 780 communicating through the bus 720. It may include. Computer system 700 may also include a network interface 790 coupled to the network. The processor 710 may be a central processing unit (CPU) or a semiconductor device that executes instructions stored in the memory 730 or the storage device 780. The memory 730 and the storage device 780 may include various types of volatile or nonvolatile storage media. For example, the memory may include read only memory (ROM) 731 and random access memory (RAM) 732. Embodiments of the present disclosure may be implemented as a computer-implemented method or as a non-transitory computer-readable medium having computer-executable instructions stored thereon. In one embodiment, when executed by a processor, computer readable instructions may perform a method according to at least one aspect of the present disclosure.

The apparatus for generating a mid-view image according to an embodiment includes a processor 710 and a memory 730, and the processor 710 executes a program stored in the memory 730, thereby contouring the left-view and right-view images. Detecting, detecting a feature point in the detected contour, detecting a correlation point for a feature point of a left-view image and a right-view image, and generating an intermediate view image based on disparity information of the feature point and the correlation point It can be done.

Although the embodiments of the present invention have been described in detail above, the scope of the present invention is not limited thereto, and various modifications and improvements of those skilled in the art using the basic concept of the present invention defined in the following claims are also provided. It belongs to the scope of rights.

Claims (10)

A device for generating a mid-view image of a stereoscopic image,
The feature point detection unit detects a contour line from the left-view image and the right-view image, and detects a feature point from the detected contour line.
Detects a correlation point for the feature point of the left-view image and the right-view image, detects an inverse correlation point for the correlation point, and detects an abnormal correlation point among the correlation points based on the position of the inverse correlation point and the feature point. Correlation point detection unit to remove and select an effective correlation point, and
Synthesis unit for generating mid-view image using disparity information of the feature point and the effective correlation point
Intermediate view image generating device comprising a. delete In claim 1,
The correlation point detection unit,
When the positions of the inverse correlation point and the feature point are different, the detected correlation point is determined to be abnormal, and the correlation point determined to be abnormal is removed. In claim 1,
The correlation point detection unit,
Among the feature points of the left-view image and the right-view image, a feature point included in a background and a foreground is separated, and a correlation point for a feature point included in the background and a correlation point for a feature point included in the foreground are detected and integrated, Mid-point image generation device. As a method of generating a mid-view image of a stereoscopic image,
Detecting a contour line from the left-view image and the right-view image,
Detecting a feature point from the contour,
Detecting a correlation point with respect to the feature point,
Detecting an inverse correlation point with respect to the correlation point,
Selecting an effective correlation point by removing an abnormal correlation point from the correlation points based on the positions of the inverse correlation point and the feature point; and
Generating an intermediate viewpoint image using disparity information between the feature point and the effective correlation point
Method for generating an intermediate viewpoint image comprising a. delete In claim 5,
The step of selecting the effective correlation point
Determining that the detected correlation point is abnormal when the positions of the inverse correlation point and the detected feature points are different, and
And removing the correlation point determined to be abnormal. In claim 5,
The step of detecting the correlation point,
Separating the feature points included in the background and the foreground among the feature points of the left-view image and the right-view image, and
And detecting and integrating a correlation point for the feature points included in the background and a correlation point for the feature points included in the foreground. In claim 8,
Detecting and integrating the correlation point,
Detecting a correlation point for a feature point included in the background and a correlation point for a feature point included in the foreground,
Detecting each inverse correlation point with respect to a correlation point included in the background and the foreground,
If the location of the inverse correlation point and the feature point included in the background and the foreground are different, determining the abnormal correlation point and removing the abnormal correlation point, and
And integrating the correlation point included in the background and the foreground from which the abnormal correlation point is removed. A device for generating a mid-view image of a stereoscopic image,
Includes a processor and memory,
The processor executes a program stored in the memory,
Detecting a contour line from the left-view image and right-view image, and detecting a feature point from the detected contour line,
Detecting a correlation point for the feature point of the left-view image and the right-view image,
Detecting an inverse correlation point with respect to the correlation point,
Selecting an effective correlation point by removing an abnormal correlation point from the correlation points based on the positions of the inverse correlation point and the feature point; and
Generating an intermediate viewpoint image based on disparity information between the feature point and the effective correlation point
Performing, mid-view image generating device.

Images