This paper is developing a dynamic background generation scheme for color video sequences. We calculate the pixels vibration and statistics for the 32 clusters to develop an effective background generation method in color video sequences. Since the color image has computation consumption problems, we use a color quantization technique to shrink the pixel value from 256 into 32 grayscales called 32 clusters firstly. And we find the main value of the pixel by a cluster statistic stage which static several frames of the video sequence. After a statistical analysis, it will have several clusters. We check and find the maximum cluster (FMC) as the pixel value for further processing. Since the background has two parts: static part and dynamic part, we measure the probability of pixels (BP) and judge if it belongs to static background or dynamic background. If the BP is less than a threshold SP we process, it is the static background. Otherwise it is the dynamic background. Finally, by adding the static background and dynamic background, the completed background is generated. In order to demonstrate our scheme is an effective method, several test sequences are used for simulations. According to the test results, it is evident that our scheme can exactly obtain an integrated color background from the color video sequences. (11 refs)
