An HVS-based motion vector composition algorithm for spatial resolution transcoding

In spatial resolution reduction transcoding, the original resolution of the incoming frames is reduced to a lower resolution in the outgoing frames. In the reduced resolution the motion vector of every macroblock needs to be calculated from the corresponding four macroblock of the original frame. Instead of reestimating these motion vectors, a number of techniques recompose these vectors using the incoming information to reduce computational complexity. Many techniques utilize the quantized DCT coefficients of residual blocks for composing the invalid vectors. This paper proposes a new motion vector composition technique by normalizing the low frequency DCT coefficients with the visual quantization matrix (VQM) derived from the study of human visual system (HVS). Simulation results show that the performance of this modified technique is better compared to other similar techniques.