Abstract: This paper proposes a background modeling and subtraction algorithm to detect moving objects from color video sequence captured by surveillance system. Background values are quantized into codebooks with respect to color distortion and brightness distortion. Pixel values of new frames are compared with the codebooks to identify foreground pixels. The algorithm is efficient in computation and takes up less memory. It can also handle scenes containing moving backgrounds and illumination variations.

Key words: background subtraction, background codebook, surveillance video, object detection

摘要： 针对监控系统获得的彩色视频序列，根据连续采样值的颜色相似度及其亮度范围，将背景像素值量化后用码本表示，利用减背景的思想对新输入的像素值与其对应位置的码本作比较判断，提取出前景运动目标像素。该算法计算复杂度小、占内存少，能够在存在前景运动的过程中提取背景，并能处理光照变化。

关键词: 减背景, 背景码本, 监控视频, 目标检测

CLC Number: 

• TP301.6