spatial - temporal model of visual attention to a video screen sequence

摘要

An improved visual attention model uses a robust adaptive segmentation algorithm to divide a current frame of a video sequence into a plurality of regions based upon both color and luminance, with each region being processed in parallel by a plurality of spatial feature algorithms including color and skin to produce respective spatial importance maps. The current frame and a previous frame are also processed to produce motion vectors for each block of the current frame, the motion vectors being compensated for camera motion, and the compensated motion vectors being converted to produce a temporal importance map. The spatial and temporal importance maps are combined using weighting based upon eye movement studies. IMAGE