2004 International Conference on Image Processing (ICIP) THE EFFECT OF GLOBAL MOTION PARAMETER ACCURACIES ON THE EFFICIENCY OF VIDEO CODING

In this paper, we present theoretical analysis on how the global motion parameter accuracies affect the efficiency of motion compensated video coding. The inaccurate global motion compensation is modelled by introducing probabilis-tic rotation. scale and translation parameter errors. Approx-imate expressions that relate the power spectrum of the pre-diction error to motion parameter errors is derived. By dou-bling the accuracy of the motion parameters, up to 6dB the-oretical gains can be obtained in prediction error variance. 1. LNTRODUCTION Global motion estimation (GME) is essential in several ap-plications such as video compression, segmentation, mo-saicing, and image registration. Global motion refers to higher order motion models like aftine or bilinear, which improves the prediction gain over simple translational model. Although many global motion estimation techniques have been proposed, the parameter accuracy vs coding rate anal-ysis have not been performed. Efficiency of motion com-pensated coding with translational motion errors have been studied in [ I, 31. Optimizing the translational accuracy for block-based coders is analyzed in [ 2]. This paper studies the effect of GM parameter inaccura-cies on motion compensated video coding. The 4-RST (ro-tation, scale translation) motion model used in the analysis is described in Section 11. The power spectrum of the pre-diction error is related to GM parameter errors, and approx-imate expressions for prediction error spectrum is derived in Section 111. The effect of motion accuracy level, the effect of video signal model, and finally the effect of approxima-tion on the prediction gain are studied in Section W, which is followed by conclusion in the last section. This work i s supponed by Skyworks Inc. and by UC DIMI