In this work we propose a new method for detecting the moving objects using pixel-level change by coupling the belief propagation(BP) method in the 3D MRF(Markov Random Field) model and Clausius Entropy model. The BP method estimates the motion possibility by considering the spatial and temporal interaction of pixels of the moving objects, and therefore has the high detect correctness. In addition this method is applicable to the environment with the changing background due to the movement of the camera. Clausius Entropy method, recently proposed, is one for detecting the objects by the entropy difference by transforming the image region to entropy region and is stable against the noise. In this work, we successfully detect all the moving objects from the small objects to the disguised objects by applying the 3D MRF model in entropy region by coupling these two methods.
展开▼
