Video coding using adaptive eigenspace

Abstract

In this paper, an eigenspace coding method introduces to encode a given video sequence using the eigenimages extracted from the previous frames. The eigenspace is updated to adopt into the changing scheme in the video sequence. Two alternative techniques have been developed, where the first approach updates the eigenspace regularly in a repeated manner (R-AKLT) which is independent of the quality of the decoded frames. The second approach uses a PSNRY based threshold (TR-AKLT) to determine if the eigenspace is to be updated with new frames or not. The experimental results clearly indicate that TR-AKLT outperforms the R-AKLT. Furthermore, when compared with H.264 AVC, the TR-AKLT method generates competitive bit rate performance with low PSNRY levels and superior bit rate performance for the tested Akiyo video sequence. © 2012 Institute of Telecommunica.