Background subtraction is a common first step in the field of video processing and it is used to reduce the effective image size in subsequent processing steps by segmenting the mostly static background from the moving or changing foreground. In this paper previous approaches towards background modeling are extended to handle videos accompanied by information gained from a novel 2D/3D camera. This camera contains a color and a PMD chip which operates on the Time-of-Flight operating principle. The background is estimated using the widely spread Gaussian mixture model in color as well as in depth and amplitude modulation. A new matching function is presented that allows for better treatment of shadows and noise and reduces block artifacts. Problems and limitations to overcome the problem of fusing high resolution color information with low resolution depth data are addressed and the approach is tested with different parameters on several scenes and the results are compared to common and widely accepted methods.
分享
引用
