computerVision A

This repo contains some crucial computer vision tasks including:
• Homography which warps an image taken by phone into an image similar to the one obtained through a scanner
• Photometric Stereo, Specularity Removal and Surface Rendering
• Edge Detection with Non-Maximum Suppression
• Implementing multiple sparse stereo matching methods
• Implement Optical Flow with Multi-resolution Lucas-Kanade

Homework Results Demonstration

HW2 - Homography warpping

Original Image Homography Warpped Image

UCSD Logo and wrapped version:

Hw3: Photometric Stereo, Specularity Removal and Surface Rendering

Table:

Surface Rendering Results:

(1) Lambertian model:

(2) Phong model:

(A) For Sphere

(B) For Pear:

Specularity Removal:

Robust Photometric Stereo:

(A) For Sphere:

(B) For Sphere

Hw3: Edge Detection with Non-Maximum Suppression

Original Image:

Gradient Magnitude:

NMS with 5x5 window:

HW4: multiple sparse stereo matching

(1) Corner Detection

(2) An exmaple of Naive Matching

(3) Epipolar Line plotting:

(4) Image Rectification

(5) Matching Using epipolar geometry

HW5: Optical Flow with Multi-resolution Lucas-Kanade

(1) Moving left and up

(2) Rotate

(3) Zoom In