Lecture 3 Template Matching Edge Detection
2 Processes for Assignment 1 Understand Image Format Pre Processing - Gaussian, Mean Filter to clean up the image Thresholding - Binary Image Segmentation - Blob Coloring 8-neighbor Thinning Template Matching - Similar to convolution Template - “Filter” - Sum Square Difference skeleton กุ้ง
3 Template Matching You may have a template Put a template, do subtraction low numbe = good match Number = 0
4 Not only documents Chips - need numbers bank checks in industrial - fixed font is enough Template Matching - Train your machine for fixed font OCR- Optical Character Recognition
5 Try matching a segment with every template The lowest score is the best match Preserve Aspect Ratio width/height Template Matching Rule of Thumb 8 12 Scale down Template Matching Templates
6 When people say edges, they means object contour Edge Detection - finding edge contours Contour Edge (something to do with shape) Texture Edge
7 Edge Detector Edge - Points in image with a lot of change in intensity Scan Line Edge Points x y Intensity Bluring (Smoothing) Step Edge
8 Simple Edge Detector edge else no edge Edge if I / X > 30 First Difference Image
9 First Difference We should blur the image first Use Gaussian Filter to reduce noise
10 First Difference
11 Second Derivative 0 = Edge First Derivative Blurring Second Derivative edge
12 Second Derivative = Edge Detector Using 2nd Derivative
13 Edge Detector Using 2nd Derivative Any zero Crossing Edge Zero-Crossing Change + to -, - to + In 2D
14 Edge Detector by Photoshop (1) 1. Open peppers.jpg 2. Change to Grayscale
15 Edge Detector by Photoshop(2) 3. Custom Filter
16 Edge Detector by Photoshop(2) 3. Custom Filter