An Efficient Method for Early Detecting All-Zero Blocks in H.264 Video Encoding
Chou-Chen Wang, Yukai Lin · 2014
In H.264 video coding, the effect of the rate-distortion optimization (RDO) on compression efficiency increases while its complexity accounts for an important part of the computation burden. Discrete cosine transform (DCT) is one of the main operations performed as a part of the RDO operation. Therefore, the elimination of DCT operations may lead to a significant reduction of the computational burden in RDO. An early detecting all-zero blocks (AZB) is an efficient DCT elimination technique in currently state-of-art video coding standards. The computations of DCT and quantization can be remarkably reduced if AZB are detected prior to DCT and quantization. This paper presents an efficient method for early AZB by combining both Parseval energy conservation theorem and the successive elimination algorithm (SEA). A mathematical inequality is derived from Parseval theorem, and we apply SEA for motion estimation (ME) to more efficient detecting AZBs. The simulation results show that the proposed method achieves approximately a 12%~22% computational saving compared to the existing methods. Furthermore, when the proposed method is implemented with other fast detection algorithms, it can further improve AZB detection ratio without any video-quality degradation.