Computational Photography, Fall 2014

Assignment 4:Panorama


1.1 Description:

The objective was to construct panoramas from four images. It is done by finding the homography between the images. The points for finding homography can be manually assigned or detected using interest point detectors like SIFT/SURF. For the automatic version, RANSAC is used for finding the best fit for homography. Finally the images are blended using alpha blending.

1.2 Homography

Tho homography is found after normailzing and subtracting the mean for stability purposes as suggested by the slides.
It should be noted that imtransform requires transpose of homography matirx.


1.3 RANSAC

The RANSAC implemented has two stopping criteria. One is the maximum number of iterations and other is ratio of correct points by total matched points.

1.4 Some comments

Utmost care must be taken when you are making a big canvas for all the four images to put in. Also, care must be taken so that the images are not too large. If they are, you may run out of memory while processing.

Manual panorama

Clicking is soo boring!!


Also what you click is not perfect, hence many artifacts are introduced. Hence automatic system should be used




Individual images
Individual images
Individual images
Individual images
Panorama



Individual images
Individual images
Individual images
Individual images
Panorama



Individual images
Individual images
Individual images
Individual images
Panorama



Individual images
Individual images
Individual images
Individual images
Panorama

1.5 Matlab Files

Download the zip file


Code:zipped files