Human detection in far-infrared images based on histograms of maximal oriented energy map

This paper proposes a new scheme for human detection in infrared images. Firstly, we present a new segmentation method based on the image histogram cluster analysis using K-means clustering method. Then, we propose a human feature extraction method based on histograms of maximal oriented energy map using Log-Gabor wavelets as the filters for orientation selecting and use a radial basis function (RBF) based support vector machines (SVM) classifier to test the performance of our feature extraction method. The detection system is based on single frame image and doesn´t need motion information of objects. The experiment results show that the algorithm is efficient.