Propagation based saliency detection for infrared pedestrian images
Yu Huang Zheng, Fugen Zhou, Lu Li, Xiangzhi Bai · 2017
Saliency detection is popular in image processing, but it is still a challenging problem for infrared pedestrian images. In this paper, an effective saliency detection method for infrared pedestrian images is proposed. Taking into consideration the characteristics of pedestrians including luminance and shape, the MSER-based local stableness (MLS) is firstly introduced. Then vertical edge-weighted contrast (VEC) is calculated. Finally, an intra-scale and inter-scale neighborhood based saliency propagation method is constructed to optimize and integrate the two features. Extensive experiments demonstrate the effectiveness of the proposed saliency method for infrared pedestrian images.