Bayesian Foreground and Shadow Detection in Uncertain Frame Rate Surveillance Videos

摘要: In in this paper, we propose a new model regarding foreground and shadow detection video sequences. The works without detailed priori object-shape information, it is also appropriate for low unstable frame rate sources. Contribution presented three key issues: 1) novel adaptive model, show the improvements versus previous approaches scenes with difficult lighting coloring effects; 2) give description based on spatial statistics of neighboring pixel values, which enhances background or shadow-colored object parts; 3) how microstructure analysis can be used proposed framework as additional feature components improving results. Finally, Markov random field to enhance accuracy separation. We validate our method outdoor indoor sequences including real surveillance videos well-known benchmark test sets.