Moving object detection based on temporal and spatial resolution mosaic

Algorithms that offer a simultaneous accurate detection of moving objects and low computational complexity are yet a challenge for video surveillance applications. Multiresolution analysis and the wavelet transform can form bases for the design of such algorithms. They can be functioned to enable the analysis and the processing in dimensions to match the nature of the input data and to allow the analysis of the images in a degree of visual details corresponding to the relevance of the information to be extracted. The algorithm proposed in this paper uses the multiresolution analysis to represent the images in a mosaic of different resolutions based on the view conditions of the scene. The 3D wavelet packet is then used to analyse the image sequence spatially and temporally. The algorithm is tested and evaluated using many scenes for vehicles traffic and pedestrian zones. In a comparison with previous work, the results visually and statistically were much improved.