Exploring Motion Vector Cost with Partial Distortion Elimination in Sum of Absolute Differences for HEVC Integer Motion Estimation

Abstract

Integer Motion Estimation (IME) is one of the most costly steps in the video coding process, being responsible for 60% of the total encoding time. Two parameters are considered when calculating the redundancies (distortion) between video blocks: Sum of Absolute Differences (SAD) and motion vector cost (MVcost). Partial Distortion Elimination (PDE) techniques may be used to optimize the calculation of the SAD unit itself, in order to avoid the computation of candidates that will certainly not be selected in the IME. This paper proposes an optimization in the distortion calculation by taking advantage of the PDE technique in SAD operation by including MVcost, in order to save even more cycles when compared to standard PDE usage, considering the High Efficiency Video Coding (HEVC) standard. We found out that the inclusion of MVcost in the distortion cost achieves a quality increase of up to 1.28%. We compare three versions: adding MVcost before/after SAD using a multiplexer, and before SAD with a Carry-Save Adder (CSA). The use of PDE including MVcost before SAD calculation using a multiplexer achieves an average reduction of 17.5% in the energy consumption when compared to accumulating MVcost after SAD. When using the CSA, we obtain an additional average reduction of 1.94%.