Automatic Detection for Tracking Moving Objects in H.264 Video Sequences Using Multi-Features and Bi-Modal Gaussian Approximation

Automatic moving object detection is essential for various computer vision applications like video surveillance systems. Many previous moving object detection methods work for usually low-res video sequences under certain constraints. Either they perform detection process based on background learning and/or pixel-level motion analysis or they focus on detecting particular objects such as faces. In this paper, we introduce a robust moving object detection method for H.264 hi-resolution video sequences, which does not rely on background learning nor focuses on detecting particular objects. The proposed method also exploits multi-features, motion and color in the process. First, it uses motion vector information given from a reference H.264 decoder, allowing for initial moving object estimation. Second, it extracts a moving-edge map, and combines it with the motion vector mask to produce an overestimate of the moving object blobs. Third, in order to extract a solid body of moving object blob with an accurate boundary, we employ a pyramid color segmentation connecting components that might be under different motions. Lastly, it fuses moving blobs with color segmentation results and performs a region-growing technique based on the blobs' Gaussian distribution of its RGB information to further refine moving blobs. Results are shown to demonstrate the accuracy of our method.