Project 4A: IMAGE WARPING and MOSAICING

CS194-26 Intro to Computer Vision and Computational Photography | Jingyi Zhou

Overview

This is the first part of the panorama stitching project where I still have to manually define correspondences.

Shoot and digitize pictures

I took three sets of two pictures for panorama:

Recover homographies

After manually defining correspondences, I can compute the homography matrix according to the following:

where i = 1 is our scale factor and a, b, c, d, e, f, g, h are computed using least squares.

Warp the images

In order to see that my warping works correctly, I rectified two images that I took from an Andy Warhol exhibit:

I cropped the first result manually just to verify that the edges have been correctly rectified.

Blend images into a mosaic

I blended using a simple weighted average:

What I've learned

I now feel one step closer to the daily technology (panorama mode in my iPhone's camera) that I used to find soooooo mysterious. Looking forward to the second part of the project!