A superpixel-based codebook model for real time foreground detection
Shicai Fan, Xiangfeng Zeng · 2017
A superpixel-based codebook model was presented to improve the efficiency of foreground detection in this paper. By clustering the neighboring pixels into superpixels according to the similarities of brightness and texture, the proposed codebook could reduce the redundancy of the codewords and the postprocessing complexity. Simulation results on some benchmark videos indicate that, the superpixel-based codebook model could significantly reduce the memory consumption and improve the processing speed without sacrificing the foreground detection precision. Applied on the embedded video processing systems based on the DM6437 processor, our algorithm could satisfy the real time requirement in foreground detection, which is straightforward to real world application.