Depth-based human body enhancement in the infrared video

Abstract

Based on the depth information acquired by the popular RGBD camera such as Kinect, the human body image areas in the infrared video can be selectively enhanced. In this paper, we firstly utilized the Optimal Contrast-Tone Mapping (OCTM) method instead of Histogram Equalization (HE) method to make a good contrast balance for the infrared video image acquired in a low illumination condition. Secondly, we used multiple iterations of the Level Set algorithm to improve the human body silhouette which initially recognized by the RGBD camera in each infrared frame. Finally, in order to improve the image quality of the human body area in each infrared frame, a fast bilateral filter had been employed to eliminate the spot noise while maintaining good edge features. Experimental results show that the proposed method can effectively enhance the human subjects in the infrared video images.