HOGG: Gabor and HOG-Based Human Detection

Aimed at the multi-scale feature of human body in visual surveillance, a new multi-scale and multi-orientation feature extraction algorithm for human detection is put forwarded in this paper. First, extract the multi-orientation multi-scale amplitude spectrum of human body using Gabor filter. Second, fuse the multi-orientation Gabor features which has the same scale information into a single feature in order to reduce the dimension of the features. Third, Histograms of Oriented Gradients of each fused image is extracted and combined as human picture features. Experimental results on PETS2006 and CAVIAR database show that the proposed algorithm has a better performance.