Adaptive weighted non-parametric background model for efficient video coding

The latest HEVC video coding standard [1] has improved the coding performance by applying a number of innovative tools compared to its predecessor H.264/AVC [2][3] including a wider range of variable block size motion estimation (ME), motion compensation (MC), prediction, and transformation units. The use of multiple reference frames (MRFs) with variable block sizes typically provides better coding performance than the single reference frame approach [1]-[4] for video with repetitive motion, uncovered background, non-integer pixel displacement, lighting change, etc. However, MRFs-based schemes require index codes to identify a particular reference frame and the computational time increases almost linearly with each additional reference frames due to ME and MC. The decision on appropriate number of reference frames is dependent on the video content and the computational time constraint which may not always allow large number of reference frames[5][6].