Computation Photography Project 5

By: Timothy Farkas

Project Overview

Created panoramas from several images with the same center of projection. This was done by manually defining points of correspondence between the images in part 1, computing a homography that warped the second image onto the first, then blending the two images together. In part 2 the points of correspondence were determine by finding the harris corners in the images, performing non maximum suppression, then feature space outlier rejection, and the finally ransac to comput the homography that can warp the second image onto the first.

Warped Images

original	warped
original	warped.

Steps in feature extraction

1. Harris corners

2. Non maximal suppression

3. Feature space outlier rejection

Panoramas

Left image.

Middle image.

Right image.

Manual correspondence panorama.

RANSAC

Left image.

Middle image.

Right image.

Manual correspondence panorama.

RANSAC

Left image.

Middle image.

Right image.

Manual correspondence panorama.

RANSAC