CS194-26: Intro to Computer Vision and Computational Photography, Fall 2021

Project 4A: Image Warping and Mosaicing

INSTRUCTOR: Alexei (Alyosha) Efros, Angjoo Kanazawa

Jingqi Huang, CS194-26



Overview

First, manually pick corresponding points for pictures. Then use least square solution to solve and find the homography H. Use inverse warping similar as the one in proj3, warp the pictures to desired appearance.

reference for building homography.

After building this Ah = b equation, solving using least square solution where h = (A^TA)^-1A^Tb

Rectify


faculty club original
book original
college clothing original
rectify by wall
rectify by book
rectify by logo

Mosaicing

Scene 1: bedroom
stitches with blending
after crop

Scene 2: living room


Scene 3: campus through window

Takeaway

Taking pictures towards the light makes the task way harder, as the different brightness of pictures is hard to deal with. Long press iphone(screen but not the shuttle) can have wonderful AE/AF which makes taking nice proper pictures easier.

As homography is prospective translation, having degree of 8, there is clever way to use only 8 unknowns and construct the homography equation. It is cool to see overlap part of pictures after homography applied is highly identical.

Choose a little more corresponding points makes result better. Choose points on corners!