Content adaptive prediction unit size decision algorithm for HEVC intra coding

The quadtree based picture partition scheme in HEVC contributes to significant coding efficiency improvement, especially for high resolution videos. But encoding complexity also increases dramatically. This paper brings forward a two-stage prediction unit size decision algorithm to speed up the original intra coding in HEVC. In the pre-stage, texture complexity of down-sampled largest coding unit (LCU) and its four sub-blocks are analyzed according to video content, to filter out unnecessary prediction units for both the LCU and its sub-blocks. Secondly, during intra coding, prediction unit sizes of encoded neighboring blocks are utilized to skip small prediction unit candidates for current block. Experimental results show that proposed algorithm can speed up original HEVC intra coding by a factor of up to 73.7% and by averagely 44.91% for 4k2k video sequences. Meanwhile, the peak signal-to-noise ratio degradation is less than 0.04dB and bit-rate stays almost the same compared with that of original HEVC intra coding.