Abstract
This thesis proposes an approach to extract the moving foreground objects for surveillance videos by background-subtraction method. We first represent each pixel with a hybrid feature vector, which includes its GMM likelihood, color and spatial features, and estimate the density for each video frame by a non-parametric method. Next, we apply a clustering process to segment the video frame into clusters with similar hybrid features. Finally, we replace the background likelihood for each cluster with the GMM likelihood in the cluster mode. Hence, the resulting background model becomes a smoothed GMM in terms of spatial and color coherency. In order to combine the temporal information of previous extracting results, we follow the DCRF model which uses the conditional random field to model the information of the temporal and spatial neighborhood and replace the original background model with our proposed modified background model. We also add the chrominance component into the original shadow model in the DCRF model to have a better approximation of shadow. Moreover, in order to reduce the computational load, we also propose a filtering step to skip pixels from the time-consuming clustering process. Our experimental results and comparisons demonstrate that the proposed method indeed achieves better detection results with accurate object contours even in dynamic scenes. Index Terms—Static background, dynamic background, DCRF model, GMM, mean-shift clustering process.