Method of and an apparatus for 3-dimensional structure estimation
AssigneeNEC Corporation
InventorShizuo Sakamoto
Life of this patent
Estimating the term…
Abstract
A method of 3-dimensional structure estimation of the invention, making use of a plurality of stereo-pictures, repeats, for each pixel of a first picture, a step of extracting corresponding small regions (4) corresponding to the concerning pixel according to a depth estimation (68); a step of calculating a neighboring correspondence value for each of the corresponding small regions representing correspondence among neighboring corresponding small regions of picures taken by neighboring cameras; a step of obtaining a sum (61) of the neighboring correspondence values; and selecting a value of the depth estimation (68) which gives a singular value of the sum (61) representing correspondence among the corresponding small regions.
Claims13
A method of 3-dimensional structure estimation for estimating a 3-dimensional structure of an object from image data of a plurality of pictures of the object each taken from each viewing position arranged on a straight line by a camera with an optical axis parallel to a direction perpendicular to the straight line;
- said method having a step of performing, for each pixel of image data of a first of the plurality of pictures, steps of:
- extracting corresponding small regions, each having a size of at least one pixel, each from the image data of each of the plurality of pictures, a position of each of said corresponding small regions in corresponding each of the plurality of pictures being defined by a focal distance of the camera, a distance between a viewing position wherefrom said corresponding each of the plurality of pictures is taken and a viewing position wherefrom the first of a plurality of pictures is taken, a position of a concerning pixel of image data of the first of the plurality of pictures, and a variable representing a depth of a point of the object corresponding to said concerning pixel;
- calculating a neighboring correspondence value for each of said corresponding small regions, said neighboring correspondence value representing correspondence among said corresponding small regions of neighboring certain of the plurality of pictures;
- obtaining a sum of said neighboring correspondence value of all of said corresponding small regions;
- selecting an estimation value in a predetermined range of said variable which gives a minimum value of said sum of said neighboring correspondence value for each pixel of image data of the first picture by controlling a depth counting unit and a reference pixel position counting unit;
- and
- outputting said estimation value as an estimation of said depth of said point corresponding to said concerning pixel.
A method of estimating a 3-dimensional structure of an object from image data of a plurality of pictures of the object each taken from each viewing position arranged on a straight line by a camera with an optical axis parallel to a direction perpendicular to the straight line;
- said method comprising:
- extracting corresponding small regions each from the image data of each of the plurality of pictures, a position of each of said corresponding small regions in corresponding each of the plurality of pictures being defined by a focal distance of the camera, a distance between a viewing position wherefrom said corresponding each of the plurality of pictures is taken and a viewing position wherefrom a first of the plurality of pictures is taken, a position of a reference pixel of image data of the first of the plurality of pictures, and a variable representing a depth of a point of the object corresponding to said reference pixel;
- calculating, for each of the plurality of pictures, an average pixel value of said corresponding small regions of neighboring certain of the plurality of pictures;
- calculating a differential of pixel values of each of said corresponding small regions to said average pixel value corresponding to said each of said corresponding small regions;
- outputting a sum of said differential of all of said corresponding small regions as a differential sum;
- varying a value of said variable in a range;
- revising said position of said reference pixel;
- and
- obtaining 3-dimensional structure information of the object by finding a value of said variable which makes said differential sum minimum for each pixel of image data of the first picture by controlling a depth counting unit and a reference pixel position counting unit.
A method of estimating a 3-dimensional structure of an object from image data of a plurality of pictures of the object each taken from each viewing position arranged on a straight line by a camera with an optical axis parallel to a direction perpendicular to the straight line;
- said method comprising:
- extracting corresponding pixels each from the image data of each of the plurality of pictures, a position of each of said corresponding pixels in corresponding each of the plurality of pictures being defined by a focal distance of the camera, a distance between a viewing position wherefrom said corresponding each of the plurality of pictures is taken and a viewing position wherefrom a first of the plurality of pictures is taken, a position of a reference pixel of image data of the first of the plurality of pictures, and a variable representing a depth of a point of the object corresponding to said reference pixel;
- calculating, for each of the plurality of pictures, an average pixel value of said corresponding pixels of neighboring certain of the plurality of pictures;
- calculating a differential of a pixel value of each of said corresponding pixels to said average pixel value corresponding to said each of said corresponding pixels;
- outputting a sum of said differential of all of said corresponding pixels as a differential sum;
- varying a value of said variable in a range;
- revising said position of said reference pixel;
- and
- obtaining 3-dimensional structure information of the object by finding a value of said variable which makes said differential sum minimum for each pixel of image data of the first picture by controlling a depth counting unit and a reference pixel position counting unit.
A method of estimating a 3-dimensional structure of an object from image data of a plurality of pictures of the object each taken from each viewing position arranged on a straight line by a camera with an optical axis parallel to a direction perpendicular to the straight line;
- said method comprising:
- extracting corresponding small regions each from the image data of each of the plurality of pictures, a position of each of said corresponding small regions in corresponding each of the plurality of pictures being defined by a focal distance of the camera, a distance between a viewing position wherefrom said corresponding each of the plurality of pictures is taken and a viewing position wherefrom a first of the plurality of pictures is taken, a position of a reference pixel of image data of the first of the plurality of pictures, and a variable representing a depth of a point of the object corresponding to said reference pixel;
- calculating, for each of the plurality of pictures, a variance of pixel values in said corresponding small regions of neighboring certain of the plurality of pictures;
- outputting a sum of said variance of all of said corresponding small regions as a variance sum;
- varying a value of said variable in a range;
- revising said position of said reference pixel;
- and
- obtaining 3-dimensional structure information of the object by finding a value of said variable which makes said variance sum minimum for each pixel of image data of the first picture by controlling a depth counting unit and a reference pixel position counting unit.
A method of estimating a 3-dimensional structure of an object from image data of a plurality of pictures of the object each taken from each viewing position arranged on a straight line by a camera with an optical axis parallel to a direction perpendicular to the straight line;
- said method comprising:
- extracting corresponding pixels each from the image data of each of the plurality of pictures, a position of each of said corresponding pixels in corresponding each of the plurality of pictures being defined by a focal distance of the camera, a distance between a viewing position wherefrom said corresponding each of the plurality of pictures is taken and a viewing position wherefrom a first of the plurality of pictures is taken, a position of a reference pixel of image data of the first of the plurality of pictures, and a variable representing a depth of a point of the object corresponding to said reference pixel;
- calculating, for each of the plurality of pictures, a variance of values of said corresponding pixels of neighboring certain of the plurality of pictures;
- outputting a sum of said variance of all of said corresponding pixels as a variance sum;
- varying a value of said variable in a range;
- revising said position of said reference pixel;
- and
- obtaining 3-dimensional structure information of the object by finding a value of said variable which makes said variance sum minimum for each pixel of image data of the first picture by controlling a depth counting unit and a reference pixel position counting unit.
Dependencies are inferred from the claim text.