EE/CSE 576 Project 2 -- Panoramaic Image Stitching

Andrew Chekerylla
University of Washington
Department of Electrical Engineering

Submitted: April 28, 2005

This project aimed to extend the functionality of the previous project by adding image stitching to the developed program. Specifically, methods for warping, aligning, and accumulating images were implemented. Material taken from the lectures and readings was used to develop the approaches and algorithms used for this project. Skeleton code and provided images were used to facilitate development. This project was developed in Microsoft Windows using the Microsoft Visual Studio 2003 .NET development environment.

This was a very interesting but time consuming project. Not all of the intended functionality was implemented successfully. It is estimated that this project was completed by the developer during 45 hours of effort.

Given the ability to embark on this project again, high priority would be given to developing a clear focus for code development. Much of the time during this project was spent trying to get different pieces of the puzzle to work independently but then there was much time wasted when it was time to integrate all of the components. A block diagram would be followed closely and less focus would be given to specific functionality and greater focus would be placed on completing full functionality before the deadline.

Heres a link to the HUB lawn lo res hi res

And a link to the pictures I took behind the Allegro cafe lo res link hi res link

If the links do not load the IVR files correctly, please see the ims provided below.

• Hub lo res high res
• Allegro alley lo res high res

This was a very time consuming task. More prototyping would have been performed (in MATLAB, for example) so that visual cues could be used to verify proper algorithm execution (using easy interface and graphical capabilites of MATLAB). I used it at the end but should have used it more as well.

I attempted to adjust feature matching to more effective (because I had great trouble getting it to work well enough to adjust the images properly in this project). My thoughts were that I could somehow correlate edge detection in the three color channels and then generate an edge detection (and matching) approach that was more effective than I previously had.

Unfortunately, regardless of countless (50++) hours of effort on this project, I was unable to complete the assignment to my own satisfaction. This all hopefully taught me many important lessons about projects such as this and I will do my best to apply these lessons to the rest of the assignments in this class.

(c) Andrew Chekerylla
April 28, 2005
deft1462 (at) ee.washington.edu