Pedestrian Detection Based on HOG and LBP

In this paper, we present a feature extraction approach for pedestrian detection by extracting the sparse representation of histograms of oriented gradients (HOG) feature and local binary pattern (LBP) feature using K-SVD. Moreover, we use PCA to reduce the dimension of HOG and LBP. We combine the low dimension principal features with the sparse representations of HOG feature directly for fast pedestrian detection from images. In addition, we compare the performance of sparse representations and PCA based features. Experimental results on INRIA databases show that the proposed approach provides a better detection result and spends less time.