What entropy tells about man-made structures

This paper explores the challenge of optimally categorizing regions for man-made environment. We propose using the histogram of oriented gradients (HOG) features for characterizing image regions, and propose an algorithm based on the entropy of HOG to select relevant regions. We also propose a region sensitive feature selection algorithm for image registration. The algorithms are applied to several mobile applications, including image search and photo registration for panorama on mobile phones. We demonstrate the effectiveness of our approach by experimental results on a large dataset.