Use MATLAB to implement a simplified edge detector (pb-lite) consisting of the following (first two steps in Arbelaez et al., 2011 http://www.cs.berkeley.edu/~arbelaez/publications/amfm_pami2011.pdf):

1) Low-level feature extraction: (1) brightness, (2) color, and (3) textons

2) Multiscale cue combination with non-maximum suppression

Then pre-define:

1) a filter bank of multiple scales and orientations.

2) half-disc masks of multiple scales and orientations

Then for every image: 1) Create a texton map by filtering and clustering the responses with kmeans. (Textures available here: http://www.robots.ox.ac.uk/~vgg/research/texclass/filters.html). 2) Compute per pixel texture gradient (tg) and brightness gradient (bg) by comparing local half-disc distributions. 3) Output a per-pixel boundary score based on the magnitude of these gradients combined with a baseline edge detector.