Week 8-1. Keypoints and Features
1. Motivation
Image Matching Problem: 같은 공간
- Panorama Stitching
Step 1: Extract keypoints (interesting points)
Step 2: Match keypoint featuers
Step 3: Align images
Applications(keypoints and features): Image alignment, 3D reconstruction, Motion tracking, Robot navigation, Indexing and database retrieval, Object recognition
2. What is a Keypoint?
• Salience 유의성
Each keypoint is distinctive
• Repeatability
Can be found despite geometric and photometric transformations 변환에도 반복적으로 등장
• Compactness and efficiency
Many fewer keypoints than image pixels
• Locality
Occupies small area of the image; robust to clutter and occlusion 어수선함과 막힘에 강건한
- small window로 봤을 때 특징이 있음
corner, blob(위치 파악에도 좋음)
3. Blob Detection
ripple: 잔물결, wave
From Edges to Blobs
Scale Selection
