A fast inter coding algorithm for HEVC based on texture and motion quad-tree models

Abstract

In order to reduce the encoding complexity of the emerging high efficiency video coding (HEVC) standard, a fast inter coding algorithm for HEVC based on texture and motion quad-tree models is proposed. First, the coding units are classified into motion or static blocks according to a motion/static detection algorithm. Then, different encoding schemes are adopted for each type of block. In particular, we can make full use of statistical properties and temporal correlations to determine the depth range and prediction mode of static largest coding tree units (LCTUs). Moreover, fast coding unit size and mode decision are made by means of quad-tree models and spatiotemporal correlations for the motion LCTUs. Finally, we merge the above schemes together to gain better performance. Experimental results show that the proposed overall algorithm, compared with the original HEVC encoding scheme, reduces encoding time by 47.5% on average with a Bjonteggard delta bit rate increase of only 1.6% for various test sequences under random access condition. Compared with a state-of-the-art algorithm, the proposed method can save more encoding time while maintaining comparative rate distortion performance.