Propose a motion segmentation algorithm based on Markov random model

it can has good segmentation effect in complex scenarios on moving objects only using motion vectors and block coding mode information in compressed stream. Preprocessing motion vector using motion vector quantization method

and Markov modeling is carried out on motion vector

the minimum energy function is used for optimization to obtain effect of moving object segmentation. Experiment shows that this method can segment moving object more accurately from complex scene compared with existing methods.