Conference version:
Efficient matching of pictorial structures. P. Felzenszwald and D. Huttenlocher. Proc. CVPR 2000.
Journal version:
Pictorial structures for object recognition.
International Journal of Computer Vision. Vol. 61, No. 1, 2005.
(need to be on a CMU computer to access the PDF)
Classical techniques for representing the multi-part structures for recognition.
Subscribe to:
Post Comments (Atom)
The tech report describing recursion is also useful:
ReplyDeleteP. Felzenszwalb, D. Huttenlocher
Distance Transforms of Sampled Functions
Cornell Computing and Information Science Technical Report TR2004-1963, September 2004
http://people.cs.uchicago.edu/~pff/papers/dt.pdf
Though the talks have been fairly high-level so far, I would like to see a little more details discussed regarding the distance transform during the presentation.
ReplyDeleteI have a minor issue about section 5 (in the CVPR paper). When it describes the first pass to compute D[x, y, phi, s], it says that "This computation is done "in place", changing values in the array as it goes." Why is that? is it also done in the second pass? It might be that my brain is already done for the day, but I don't see what difference does it make to change values in place or not.
ReplyDeleteWhat is the alternate to changing in place (dynamic programming) and having it efficient?
ReplyDelete