Accelerating the multiple reference frames compensation in the H.264 video coder

Khalil Hachicha 1 David Faura Olivier Romain 1 Patrick Garda 1
1 SYEL - Systèmes Electroniques
LIP6 - Laboratoire d'Informatique de Paris 6
Abstract : Using the multiple reference frames compensation in the H264 coder improves the coding efficiency for sequences which contain uncovered backgrounds, repetitive motions and highly textured areas. Unfortunately this technique requires excessive memory and computation resources. In this article, we proposed and implemented a technique based on Markov Random Fields Algorithm relying on robust moving pixel segmentation. By the introduction of this technique, we were able to decrease the number of reference frames from five to three while keeping similar video coding performances. The coding time decreased by 35% and the sequence quality was preserved. After the validation of our idea, we evaluated the processing time of the Markov algorithm on architectures intended for embedded multimedia applications. Both DSP and FPGA implementations were explored. We were able to process 50 frames(128 × 128)/s on the EP1S10 FPGA paltform and 35 frames(128 × 128)/s on the ADSP BF533.
Document type :
Journal articles
Complete list of metadatas
Contributor : Lip6 Publications <>
Submitted on : Monday, September 14, 2015 - 2:45:06 PM
Last modification on : Thursday, March 21, 2019 - 2:54:27 PM

Links full text



Khalil Hachicha, David Faura, Olivier Romain, Patrick Garda. Accelerating the multiple reference frames compensation in the H.264 video coder. Journal of Real-Time Image Processing, Springer Verlag, 2009, 4 (1), pp.55-65. ⟨10.1007/s11554-008-0101-1⟩. ⟨hal-01198822⟩



Record views