关键词: YUV空间, 码本模型, 信号能量分析, 码字, 前景检测

Abstract: This paper proposes a YUV-color-space codebook-model-based motion detection algorithm which fuses the technique of codebook quantization/clustering with energy analysis of video signal. The proposed algorithm constructs codebook model for each pixel on the YUV-color-space. Then the mean value of each pixel fitting into the codeword and the energy value which fluctuating above and below that mean value is calculated on the stage of codewords training. The image frame is partitioned to foreground and background by means of energy threshold during the phase of foreground detection. Experimental results show that comparing with the  YUV-color-space codebook method, not only does the proposed algorithm gain more speed and accuracy on the process of foreground extracting, but also show more resistence to strong disturbance resulted from shaking leaves, water wave and mutation of luminance in the complicated condition. Furthermore it has better robustness on shadow suppression.

Key words: YUV space, codebook model, analysis of signal energy, codeword, foreground detection