image correspondence - using feature detection a la SIFT or SURF structure from motion - using bundle adjustment and/or RANSAC the photo graph - I need to think hard about an efficient implementation rendering - a good graphics library and platform should be chosen interface - ease of navigation will be paramount in implementation.
This is microsoft's image mosaic software, photosynth:
www.photosynth.com
This is a TED Talk describing it:
http://www.ted.com/index.php/talks/blaise_aguera_y_arcas_demos_photosynth.html
And this is an introductory tutorial to image stitching:
http://research.microsoft.com/apps/pubs/default.aspx?id=75695
by a co-creator of photosynth, Richard Szeliski.
For even more fun, this is the working bibliography for the project:
http://photograph.sourceforge.net/ref.html
And perhaps most fun, here is a ?broken? link to download the thesis PDF:
http://photograph.sourceforge.net/thesis.pdf
My three increments for the project are:
1. Choose and implement the image correspondence algorithms. 2. Design and implement the actual graph structure for the photos. 3. Choose and set up rendering library on an appropriate platform to render the photos.And here is my sorta schedule.
I've finished the first part of my first increment; I have a working implementation of the SIFT feature detector.
Please see the sift page for details and code.
If you'd like to download my sift detector, its in the PhotoGraph git repo on the main sourceforge project page.
Further install instructions can be found there :).
My name is Nick Insalata and I attend Portland State University.
I was taking the morning session of OSS 2009.
I can be reached by email: insalatn at pdx dot edu.