CS 194-26: Image Manipulation and Computational Photography, Fall 2018

Project 6A: Image Warping and Mosaicing

Jordan Flores, CS194-26-agf



Overview

The goal of this project is to combine multiple photos taken from a single point with different angles, mosaicing them together. The pictures must be warped with a transformation matrix to rectify the image, or in other words, correct the angle of an image to match the another image. We do this by first shooting the pictures, recovering the homography (aka the transformation matrix), warping an image to its reference image, then finally blending the two images together

Shooting Pictures

The pictures that we want to mosaic must overlap in order to have corresponding points that will allow us to recover the homography.

Recovering Homography

To recover the homography of two images, we must solve for H in the equation p'H = p. I set up the equation to match the graphics below.

Image Rectification

To align the images with their reference images, I warped the images using the homography as the transformation matrix.

Mosaicing

Now that we have our two images both aligned, we can blend them together to create a larger mosaic! Horray!

Summary

This project taught me the power of the homography, allowing us to recover a transformation matrix to recreate the original angle of a picture. The concept of stitching multiple pictures together that were taken at different angles makes me interested in the extrapolation the idea to larger panoramic stitching.