Advanced Digital Image Processing - Project Work

This is a project about digital image processing, with particular regard to image segmentation.

The goals are:

• Implement an efficient approach to estimate the oriented gradient of histograms.
• Utilize the local minimum of the latter as a seed for a morphological watershed segmentation

This repository implements the work described in:

Contour Detection and Hierarchical Image Segmentation
P. Arbelaez, M. Maire, C. Fowlkes and J. Malik. IEEE TPAMI, Vol. 33, No. 5, pp. 898-916, May 2011 http://www.eecs.berkeley.edu/Research/Projects/CS/vision/grouping/papers/amfm_pami2010.pdf

Results

We successfully managed to achieve both tasks. See "Final Report" to get an intro to the topics of this research and an overview of our work. The docs folder contains our presentation at a student conference inside Delft University of Technology.

Gradient of Histogram

Influence of scale in gradients' magnitudes

Big scales images give rise to fine contours in contrast with the coarse contours acquired in the smaller scale image.

Influence of θ in gradients’ magnitudes

Using one single orientation (θ=0°,90°,45°,125°) is not enough to acquire contour quality, but the combination of all give rise to an efficient contour detector.

Watershed Segmentation