Reducing Encoder Complexity of Intra-mode Decision using CU Early Termination Algorithm

Abstract

In this thesis an intra prediction algorithm is proposed that terminates complete full search prediction for the CU and is replaced by CU early termination algorithm which determines the complexity of the CU block and then decision is made to further split or non-split the CU. When the CU texture is complex the CU is split into smaller sub units to find the best size and when the CU texture is flat, the CU is not divided further into sub – units. Down sampling is done first after which complexity is calculated by which a threshold value is set. This threshold value dictates early termination of CU block. This is followed by a TU mode decision to find the optimal prediction mode from the 35 prediction modes. Proposed method will use tree split/tree merge algorithm. Experimental results based on several video test sequences suggest a decrease of about 12%-24% in encoding time is achieved with implementation of the proposed CU early termination algorithm and fast intra mode decision algorithm for intra predication with negligible degradation in peak signal to noise ratio (PSNR). Metrics such as BD-bitrate (Bjøntegaard Delta bitrate), BD-PSNR (Bjøntegaard Delta Peak Signal to Noise Ratio), RD graph (Rate Distortion) and computational complexity are also used.