抽象的

Moving Object Extraction Based on Background Reconstruction

Prajakta A Patil, Prachi A Deshpande

Object extraction from video is one of the most important areas of video processing in which objects from video sequences are extracted and used for many applications such as surveillance systems. This thesis describes the theoretical bases, development and testing of moving object detection framework. Many systems uses motion and color information to detect the moving object from the video sequences, but these methods cannot produce robust segmentation results. We propose an algorithm for automatically detecting and segmenting a moving object from single concept videos, i.e. videos which have only one object category of interest but may have multiple object instances with pose, scale, etc. variations. Given such a video, we first reconstruct the background image from the input frames. From that background image we can construct motion cues based on motion information and colour cues based on color information present in frames. Then combine that both motion and color cues in MRF framework to get objet which is separated from the background.