Search

US-12626375-B2 - Homography matrix generation apparatus, control method, and computer-readable medium

US12626375B2US 12626375 B2US12626375 B2US 12626375B2US-12626375-B2

Abstract

A homography matrix generation apparatus ( 2000 ) detects two or more key-point pairs from a first image ( 10 ) and a second image ( 20 ). The homography matrix generation apparatus ( 2000 ) detects, for each of the key-point pairs, a derived point pair, which is a pair of a point separated by a first distance in a first direction from a point on the first image ( 10 ) included in the key-point pair and a point separated by a second distance in a second direction from a point on the second image ( 20 ) included in the key-point pair. The homography matrix generation apparatus ( 2000 ) uses each of the detected key-point pairs and the derived point pairs to generate a homography matrix ( 40 ) representing a homography from a point on the first image ( 10 ) to a point on the second image ( 20 ).

Inventors

  • Gaku Nakano

Assignees

  • NEC CORPORATION

Dates

Publication Date
20260512
Application Date
20201110

Claims (12)

  1. 1 . A homography matrix generation apparatus comprising: at least one memory storing instructions; and at least one processor configured to execute the instructions to: detect, from a first image and a second image, two or more key-point pairs each of which is a pair of key-points corresponding to each other; detect, for each of the key-point pairs, a derived point pair that is a pair of a first point and a second point; the first point being separated by a first distance in a first direction from a point on the first image included in the key-point pair, and the second point being separated by a second distance in a second direction from a point on the second image included in the key-point pair; and generate a homography matrix representing a homography from a point on the first image to a point on the second image using each of the detected key-point pairs and the derived point pairs, wherein the first direction and the first distance are determined based on a feature computed for the point on the first image included in the key-point pair, wherein the second direction and the second distance are determined based on a feature computed for the point on the second image included in the key-point pair, and wherein the at least one processor is further configured to extract two points from the key-point pairs and the derived point pairs to compute a signed area, and to determine whether to generate the homography matrix based on a sign of the computed signed area.
  2. 2 . The homography matrix generation apparatus according to claim 1 , wherein the first direction and the first distance are respectively determined based on a principal axis direction and a scale length of a scale-invariant feature computed for the point on the first image, and wherein the second direction and the second distance are respectively determined based on a principal axis direction and a scale length of a scale-invariant feature computed for the point on the second image, respectively.
  3. 3 . The homography matrix generation apparatus according to claim 1 , wherein the first direction and the first distance are respectively determined based on a direction and a length of a specific axis of an affine invariant feature computed for the point on the first image, and wherein the second direction and the second distance are respectively determined based on a direction and a length of a specific axis of an affine invariant feature computed for the point on the second image.
  4. 4 . The homography matrix generation apparatus according to claim 1 , wherein the at least one processor is further configured to repeatedly generate the homography matrix while changing the key-point pairs to be used for detection of the derived point pairs, and to output the homography matrix with highest accuracy of a plurality of the generated homography matrices.
  5. 5 . A control method to be executed by a computer, the control method comprising: detecting, from a first image and a second image, two or more key-point pairs each of which is a pair of key-points corresponding to each other; detecting, for each of the key-point pairs, a derived point pair that is a pair of a first point and a second point, the first point being separated by a first distance in a first direction from a point on the first image included in the key-point pair, and the second point being separated by a second distance in a second direction from a point on the second image included in the key-point pair; and generating a homography matrix representing a homography from a point on the first image to a point on the second image using each of the detected key-point pairs and the derived point pairs, wherein the first direction and the first distance are determined based on a feature computed for the point on the first image included in the key-point pair, wherein the second direction and the second distance are determined based on a feature computed for the point on the second image included in the key-point pair, and wherein the control method further comprises: extracting two points from the key-point pairs and the derived point pairs to compute a signed area; and determining whether to generate the homography matrix based on a sign of the computed signed area.
  6. 6 . The control method according to claim 5 , wherein the first direction and the first distance are respectively determined based on a principal axis direction and a scale length of a scale-invariant feature computed for the point on the first image, and wherein the second direction and the second distance are respectively determined based on a principal axis direction and a scale length of a scale-invariant feature computed for the point on the second image.
  7. 7 . The control method according to claim 5 , wherein the first direction and the first distance are respectively determined based on a direction and a length of a specific axis of an affine invariant feature computed for the point on the first image, and wherein the second direction and the second distance are respectively determined based on a direction and a length of a specific axis of an affine invariant feature computed for the point on the second image.
  8. 8 . The control method according to claim 5 , comprising: repeatedly generating the homography matrix while changing the key-point pairs to be used for detection of the derived point pairs; and outputting the homography matrix with highest accuracy of a plurality of the generated homography matrices.
  9. 9 . A non-transitory computer-readable medium storing a program causing a computer to execute: detecting, from a first image and a second image, two or more key-point pairs each of which is a pair of key-points corresponding to each other; detecting, for each of the key-point pairs, a derived point pair that is a pair of a first point and a second point, the first point being separated by a first distance in a first direction from a point on the first image included in the key-point pair, and the second point being separated by a second distance in a second direction from a point on the second image included in the key-point pair; and generating a homography matrix representing a homography from a point on the first image to a point on the second image using each of the detected key-point pairs and the derived point pairs, wherein the first direction and the first distance are determined based on a feature computed for the point on the first image included in the key-point pair, wherein the second direction and the second distance are determined based on a feature computed for the point on the second image included in the key-point pair, and wherein the program causes the computer to execute: extracting two points from the key-point pairs and the derived point pairs to compute a signed area; and determining whether to generate the homography matrix based on a sign of the computed signed area.
  10. 10 . The computer-readable medium according to claim 9 , wherein the first direction and the first distance are respectively determined based on a principal axis direction and a scale length of a scale-invariant feature computed for the point on the first image, and wherein the second direction and the second distance are respectively determined based on a principal axis direction and a scale length of a scale-invariant feature computed for the point on the second image.
  11. 11 . The computer-readable medium according to claim 9 , wherein the first direction and the first distance are respectively determined based on a direction and a length of a specific axis of an affine invariant feature computed for the point on the first image, and wherein the second direction and the second distance are respectively determined based on a direction and a length of a specific axis of an affine invariant feature computed for the point on the second image.
  12. 12 . The computer-readable medium according to claim 9 , wherein the program causes the computer to execute: repeatedly generating the homography matrix while changing the key-point pairs to be used for detection of the derived point pairs; and outputting the homography matrix with highest accuracy of a plurality of the generated homography matrices.

Description

This application is a National Stage Entry of PCT/JP2020/041962 filed on Nov. 10, 2020, the contents of all of which are incorporated herein by reference, in their entirety. TECHNICAL FIELD The present disclosure relates to generation of a homography matrix. BACKGROUND ART Techniques for generating a homography matrix representing a homography from a point on one image to a point on another image have been developed. For example, Patent Literature 1 discloses a technique for generating a homography matrix for acquiring, from an image obtained by photographing the subject at an oblique angle, a bird's-eye-view image in which a subject is viewed from vertically above. In Patent Literature 1, four or more pairs of key-points corresponding to each other are detected from two images to generate a homography matrix from the set of the detected key-point pairs using a robust estimation algorithm, such as random sample consensus (RANSAC). CITATION LIST Patent Literature Patent Literature 1: Japanese Unexamined Patent Application Publication No. 2015-233212 Non Patent Literature Non Patent Literature 1: D. Barath and Z. Kukelova, “Homography from two orientation- and scale-covariant features”, Computer Research Repository, arXiv:1906. 11927, Jun. 27, 2019. SUMMARY OF INVENTION Technical Problem The inventor has studied a new technique for generating a homography matrix. A purpose of the present disclosure is to provide a new technique for generating a homography matrix. Solution to Problem A homography matrix generation apparatus according to the present disclosure includes: a first detecting unit configured to detect, from a first image and a second image, two or more key-point pairs each of which is a pair of key-points corresponding to each other; a second detecting unit configured to detect, for each of the key-point pairs, a derived point pair that is a pair of a point separated by a first distance in a first direction from a point on the first image included in the key-point pair and a point separated by a second distance in a second direction from a point on the second image included in the key-point pair; and a generating unit configured to generate a homography matrix representing a homography from a point on the first image to a point on the second image using each of the detected key-point pairs and the derived point pairs. The first direction and the first distance are determined based on a feature computed for the point on the first image included in the key-point pair. The second direction and the second distance are determined based on a feature computed for the point on the second image included in the key-point pair. A control method according to the present disclosure is to be executed by a computer. The control method includes: a first detection step of detecting, from a first image and a second image, two or more key-point pairs each of which is a pair of key-points corresponding to each other; a second detection step of detecting, for each of the key-point pairs, a derived point pair that is a pair of a point separated by a first distance in a first direction from a point on the first image included in the key-point pair and a point separated by a second distance in a second direction from a point on the second image included in the key-point pair; and a generation step of generating a homography matrix representing a homography from a point on the first image to a point on the second image using each of the detected key-point pairs and the derived point pairs. The first direction and the first distance are determined based on a feature computed for the point on the first image included in the key-point pair. The second direction and the second distance are determined based on a feature computed for the point on the second image included in the key-point pair. A computer-readable medium according to the present disclosure stores a program that causes a computer to execute the control method according to the present disclosure. Advantageous Effects of Invention According to the present disclosure, a new technique for generating a homography matrix is provided. BRIEF DESCRIPTION OF DRAWINGS FIG. 1 is a diagram showing an example of an overview of the operation of a homography matrix generation apparatus according to a first example embodiment. FIG. 2 is a diagram showing examples of a key-point pair and a derived point pair. FIG. 3 is a block diagram showing an example of a functional configuration of the homography matrix generation apparatus according to the first example embodiment. FIG. 4 is a block diagram showing an example of a hardware configuration of a computer that implements the homography matrix generation apparatus. FIG. 5 is a flowchart showing an example of a flow of processing performed by the homography matrix generation apparatus according to the first example embodiment. FIG. 6 is a flowchart showing an example of a flow of processing performed by a homography matrix generation