Keypoint unwarping for machine vision applications
First Claim
Patent Images
1. An image processing system, comprising:
- one or more memories, which, in operation, store image data; and
image processing circuitry, which, in operation;
identifies a plurality of keypoints within received image data representative of a first image;
generates, using the received image data representative of the first image, descriptor data for at least some keypoints within the received image data;
transforms a subset of keypoints of the identified plurality of keypoints within the received image data representative of the first image, wherein the subset of keypoints does not include all of the keypoints of the identified plurality of keypoints, the transforming the subset of keypoints including;
generating, for the identified plurality of keypoints within the received image data, first keypoint data; and
transforming, using an image deformation model, first keypoint data of only the subset of keypoints of the identified plurality of keypoints, producing second keypoint data corresponding to the subset of keypoints of the identified plurality of keypoints;
generates output image data based on the second keypoint data and the descriptor data; and
determines whether one or more features of the first image match one or more features of at least one comparison image based on the output image data.
1 Assignment
0 Petitions
Accused Products
Abstract
Apparatus and methods to unwarp at least portions of distorted, electronically-captured images are described. Keypoints, instead of an entire image, may be unwarped and used in various machine-vision algorithms, such as object recognition, image matching, and 3D reconstruction algorithms. When using unwarped keypoints, the machine-vision algorithms may perform reliably irrespective of distortions that may be introduced by one or more image capture systems.
-
Citations
36 Claims
-
1. An image processing system, comprising:
-
one or more memories, which, in operation, store image data; and image processing circuitry, which, in operation; identifies a plurality of keypoints within received image data representative of a first image; generates, using the received image data representative of the first image, descriptor data for at least some keypoints within the received image data; transforms a subset of keypoints of the identified plurality of keypoints within the received image data representative of the first image, wherein the subset of keypoints does not include all of the keypoints of the identified plurality of keypoints, the transforming the subset of keypoints including; generating, for the identified plurality of keypoints within the received image data, first keypoint data; and transforming, using an image deformation model, first keypoint data of only the subset of keypoints of the identified plurality of keypoints, producing second keypoint data corresponding to the subset of keypoints of the identified plurality of keypoints; generates output image data based on the second keypoint data and the descriptor data; and determines whether one or more features of the first image match one or more features of at least one comparison image based on the output image data. - View Dependent Claims (2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15)
-
-
16. An image processing method, comprising:
-
receiving, by image processing circuitry, image data representative of a first image; identifying, by the image processing circuitry, a plurality of keypoints within the received image data; generating, by the image processing circuitry and using the received image data representative of the first image, descriptor data for at least some of the plurality of keypoints within the received image data; transforming a subset of keypoints of the identified plurality of keypoints within the received image data representative of the first image, wherein the subset of keypoints does not include all of the keypoints of the identified plurality of keypoints, the transforming the subset of keypoints including; generating, by the image processing circuitry, first keypoint data corresponding to the plurality of identified keypoints; and transforming, by the image processing circuitry and using an image deformation model, first keypoint data corresponding to only the subset of keypoints of the identified plurality of keypoints, producing second keypoint data corresponding to the subset of keypoints of the identified plurality of keypoints; generating, by the image processing circuitry, output image data based on the second keypoint data corresponding to the subset of keypoints of the identified plurality of keypoints and the descriptor data; and determining, by the image processing circuitry, whether one or more features of the first image match one or more features of at least one comparison image based on the output image data. - View Dependent Claims (17, 18, 19, 20, 21, 22)
-
-
23. A non-transitory computer-readable medium whose contents configure image processing circuitry to perform a method, the method comprising:
-
identifying a plurality of keypoints within image data representing a first image; generating, using the image data representing the first image, descriptor data for at least some of the identified plurality of keypoints within the image data representing the first image; transforming a subset of keypoints of the identified plurality of keypoints within image data representative of the first image, wherein the subset of keypoints does not include all of the keypoints of the identified plurality of keypoints, the transforming the subset of keypoints including; generating first keypoint data corresponding to the plurality of identified keypoints; and transforming, using an image deformation model, first keypoint data of only the subset of keypoints of the identified plurality of keypoints, producing second keypoint data corresponding to the subset of keypoints of the identified plurality of keypoints; generating output image data based on the second keypoint data corresponding to the subset of keypoints of the identified plurality of keypoints and the descriptor data; and determining whether one or more features of the first image match one or more features of at least one comparison image based on the output image data. - View Dependent Claims (24, 25, 26, 27)
-
-
28. An image processing system, comprising:
-
one or more memories, which, in operation, store image data; and digital image processing circuitry, which, in operation; identifies a plurality of keypoints within received image data representative of a first image; generates, using the image data representative of the first image, descriptor data for at least some of the identified plurality of keypoints; transforms a subset of first keypoint data of the identified plurality of keypoints within the image data representative of the first image, wherein the subset does not include first keypoint data of all of the keypoints of the identified plurality of keypoints, the transforming the subset including; generating, for the identified plurality of keypoints within the received image data, the first keypoint data; and transforming, using an image deformation model, only the subset of the first keypoint data to produce second keypoint data, the subset of the first keypoint data corresponding to a subset of the identified plurality of keypoints; generates output image data based on the second keypoint data and the descriptor data; and determines whether one or more features of the first image match one or more features of at least one comparison image based on the output image data. - View Dependent Claims (29, 30, 31, 32)
-
-
33. An image processing system, comprising:
-
one or more memories, which, in operation, store image data; and digital image processing circuitry, which, in operation; identifies a plurality of keypoints within received image data representative of a first image; generates, for the identified plurality of keypoints within the received image data, first keypoint data; transforms keypoint data of at least a portion of the identified plurality of keypoints within the image data representative of the first image, the transforming the keypoint data of the at least a portion of the identified plurality of keypoints including;
transforming, using an image deformation model, the first keypoint data of the at least a portion of the identified plurality of keypoints to produce second keypoint data;generates, using the image data representative of the first image, descriptor data for at least some of the identified plurality of keypoints, wherein the generating of the descriptor data is independent of the transforming of the first keypoint data of the at least a portion of the identified plurality of keypoints; generates output image data based on the second keypoint data and the descriptor data; and determines whether one or more features of the first image match one or more features of at least one comparison image based on the output image data. - View Dependent Claims (34, 35, 36)
-
Specification