Mode decision acceleration for scalable video coding through coded block pattern

This paper presents a fast mode decision to accelerate the encoding process of scalable video coding (SVC) through coded block pattern (CBP) analysis. CBP contains residual information of each macroblock (MB) and provide cues for mode prediction in encoding process. The proposed algorithm makes use of the existing mode information in base layer (BL), enhancement layer (EL) and CBP to remove invalid modes for the mode prediction of the current MB in EL coding. Also, the EL’s modes in the highest temporal level of a group of picture are determined by high correlation exists in adjacent frames. Experimental results show that the proposed algorithm saves an average of 68% of EL motion estimation time with minor degradation in PSNR and with few bitrate increase of SVC quality and spatial scalabilities when compared to JSVM 9.19.14. Our algorithm outperforms the best-known method by 17% EL motion estimation time saving in average.

► This research is motivated by the strong demand for fast mode decision in SVC.
► Most existing approaches do not consider using metadata like CBP to accelerate SVC.
► Extensive experiments are performed to verify the effectiveness of our algorithm.
► Our algorithm has an average 68% ELME time saving when compared to JSVM 9.19.14.
► Our algorithm outperforms the best-known method published in IEEE CSVT 2010.