Pedestrian detection in infrared images using local thresholding

This paper presents a new and efficient approach for automatic pedestrian detection in infrared images. The approach consists of three steps; initially, background subtraction model designed based on the input image intensity properties of target and background region to suppress the noise in image. Secondly, order statistics filter is applied on the background suppressed image to enhance the target and suppress the noise leftover, if any, from the image. Finally, Gaussian function based local thresholding is used to detect the pedestrian in image. The experimental results of the proposed method are compared with the existing methods. From the results observed, the proposed method outperforms in both subjective and objective evaluation.