Presentation is loading. Please wait.

Presentation is loading. Please wait.

Se-Hoon Park 26 th August 2014 Backgrounds for feature extraction.

Similar presentations


Presentation on theme: "Se-Hoon Park 26 th August 2014 Backgrounds for feature extraction."— Presentation transcript:

1 Se-Hoon Park 26 th August 2014 Backgrounds for feature extraction

2 Table of contents  Edge detection -> Canny edge detector A combined corner and edge detector.A combined corner and edge detector. C Harris, M Stephens - Alvey vision conference, 1988 (citation : 9984)C Harris  Corner detection -> Harris corner detector A computational approach to edge detection J CannyJ Canny - Pattern Analysis and Machine Intelligence, IEEE, 1986 (citation:19512)

3  Edge detection -> Canny edge detector  Corner detection -> Harris corner detector

4 Canny edge ● ○ ○ ○ ○ Sobel filtering I : intensity of image

5 Canny edge ● ● ○ ○ ○

6 Canny edge ● ● ● ○ ○ Sobel filtering

7 Canny edge ● ● ● ● ○ Direction of gradient x y Non – maximum suppression Suppression 수행 전 Suppression 수행 후 Magnitude of gradient Direction of gradient

8 Canny edge ● ● ● ● ● Double thresholding noiseWeak edgeStrong edge Double threshold 수행 전 Double threshold 수행 후 Edge tracking by hysteresis

9  Edge detection -> Canny edge detector  Corner detection -> Harris corner detector

10 Canny edge ● ● ● ● ● Harris corner ● ○ ○ ○ ○ What is corner? “flat” no change in all direction“edge” no change along the edge direction“corner” significant change in all directions

11 Canny edge ● ● ● ● ● Harris corner ● ● ○ ○ ○

12 Canny edge ● ● ● ● ● Harris corner ● ● ● ○ ○ Harris corner detection

13 Canny edge ● ● ● ● ● Harris corner ● ● ● ● ○ properties Rotation invariance Partial intensity invariance

14 Canny edge ● ● ● ● ● Harris corner ● ● ● ● ● Problem Non-invariant image scale ● Lowe, D.G., "Distinctive image features from scale-invariant keypoints", IJCV 2004.1

15 Canny edge ● ● ● ● ● Harris corner ● ● ● ● ● Scale space Scale invariance 영상의 특징 점이 scale 축을 따라서 반복 검출될 경우, scale invariant 한 특징 점이 된다. SIFT 의 경우 스케일 축을 따라서도 Laplacian 이 극대, 극소가 되는 점들이 scale invariant 특징 점.


Download ppt "Se-Hoon Park 26 th August 2014 Backgrounds for feature extraction."

Similar presentations


Ads by Google