티스토리 뷰

Corner Detection method 중 하나인 Harris Detector의 한계로

  • Not invariant to image scale
  • Not invariant to Affine(projective) transformation

이라는 게 있다.  viewpoint에 따라 이미지에 Affine transformation이 일어나는 경우가 있어서 중요한 문제인듯 하다.

Affine Transformation

Affine transformation is a linear mapping method that preserves points, straight lines, and planes. Sets of parallel lines remain parallel after an affine transformation. - https://kr.mathworks.com/discovery/affine-transformation.html
아핀 변환은 선형변환 중 하나로 점, 직선, 평면을 보존한다. 평행선이 보존된다.

쉽게 말해서 A Combination of Translation, Rotatoin, Shear, Scale 이다. (여기서 는 Reflect 도 추가한다)

Matlab 홈페이지에 나와있는 그림설명

 Shear matrix is an elementary matrix that represents the addition of a multiple of one row or column to another.
https://www.quora.com/What-is-shear-matrix

직관적인 설명이 없어서 그냥 위의 의미를 선형대수학적으로 이해하는 게 빠를 것 같다.
쉽게 말해 어떤 도형이 있을 때 한 dimension의 길이 비례하는 값을 다른 dimension에 더하는 변형이다. 
x,y축에 독립적으로 translation이 이루어지기 때문에 점 사이의 상대적 거리도 유지하고, 평행선도 보존한다.

shear transformation in x axis 

차라리 matrix를 보는 게 편할 것 같다. 

위키피디아 이미지


아 애초에 글을 쓰기 시작한 이유는 

  • Not invariant to Affine(projective) transformation

가 문제인 이유가 단순히 Affine transformation에 scale transformation이 포함되어 있어서인지 궁금해서였다.

Rotation, translation은 Autocorrelation matrix M의 eigenvalue에 영향을 안 줄테니 상관없고, (사실 자세히는 모르겠다 예시를 본 것도 아니고..)
Shear transformation에서 Scale change가 있을테니 문제가 될 것이다. 점 사이의 상대적인 거리는 보존해도 절대적인 Scale이 변할테니

'Research (연구 관련)' 카테고리의 다른 글

2D 이미지와 퓨리에 변환  (1) 2019.04.23
리눅스 make subdirectories  (0) 2019.04.22
리눅스 scp, find  (0) 2019.04.19
Bilateral Filter / Guidance Filter  (0) 2019.04.18
Procrustes Analysis  (0) 2019.04.13
공지사항
최근에 올라온 글
최근에 달린 댓글
Total
Today
Yesterday
링크
«   2024/04   »
1 2 3 4 5 6
7 8 9 10 11 12 13
14 15 16 17 18 19 20
21 22 23 24 25 26 27
28 29 30
글 보관함