Color Particle Filter Tracking using Frame Segmentation based on JND Color and Spatial Connectivity 4-D Histogram

Object tracking is one of the most important components in numerous applications of computer vision. Color can provide an efficient visual feature for tracking non-rigid objects in real-time. The color is chosen as tracking feature to make the process scale and rotation invariant. The color of an object can vary over time due to variations in the illumination conditions, the visual angle and the camera parameters. This paper presents the integration of color distributions into particle filtering. The color feature is extracted using our novel 4D color histogram of the image, which is determined using JND color similarity threshold and connectivity of the neighboring pixels. Particle filter tracks several hypotheses simultaneously and weighs them according to their similarity to the target model. The popular Bhattacharyya coefficient is used as similarity measure between two color distributions. The tracking results are compared on the basis of precision over the data set of video sequences from the website http://visualtracking.net of CVPR13 bench marking paper. The proposed tracker yields better precision values as compared to previous reported results.