Adaptive Segmentation Technique For Object-Base MPEG-4 System

This paper presents a segmentation application based on change-detection techniques. The algorithm is able to deal with sequences having different amounts of motion by adjusting the number of frames to skip, to reduce the light drift by DC correction and dealing with noise in a locally adaptive manner. An adaptive truncation processing is used to remove background content as well as to discard the background noise to achieve a binary shape. Finally, a boundary smoothing method is used to improve the segmentation quality. The segmented algorithm can pass various sequences testing with acceptable error-ratio. While segmented algorithm applying on MPEG-4 based surveillance systems, the coding efficiency can be greatly promoted as compared to the conventional MPEG-2 coding. The algorithm uses low computational complexity to make it applicable to real-time surveillance systems using software implementation.