Computational Complexity Reduction of HEVC by Early Termination of Transform Unit Partitioning Using Raster Scan Approach

High efficiency video coding (HEVC) is the newest video coding standard which has been developed by the Joint Collaborative Team on Video Coding (JCT-VC). The performance of the HEVC is superior to its predecessor standard, H.264. The compression efficiency of the HEVC is higher by 50% at one half of the bit-rate without compromising on the video quality. Each Coding Unit (CU) in the HEVC is accompanied by a Transform Unit (TU). The Quad-Tree partition of the CU and the Residual Quad Tree (RQT) partitioning of TU accounts for the significant computation complexity at the intra prediction mode of HEVC. We have extended the idea of partitioning each TU to the same depth of its corresponding CU. In the proposed paper, the Raster-Scan approach for pruning of the Quad-Tree structure of the TU is implemented. The proposed approach could reduce the computational complexity of HEVC at the intra prediction mode. The CU is partitioned into a Quad-Tree structure based on the Rate-Distortion (RD) cost computation. The CU partition structure is scanned by using the Raster Scan technique. The purpose of the scanning is to get an estimate of the depth of the partitioning of the current CU. During scanning, the depth of the current CU partition is extracted. The extracted partition depth level is considered as the best depth level for the subsequent TU. Once the best TU depth is identified, further partitioning of the TU within the given CU is terminated. Thus, the pruning of the RQT of the TU reduces the computational burden of the HEVC at intra prediction mode. The reduction in computation complexity is measured in terms of the saving in the overall encoding time. We have achieved an average reduction of 37% in the total encoding time in comparison to the standard HM reference model. The peak signal to noise ratio (PSNR) is slightly increased by an average of 0.0052 with the average bit-rate reduction of 0.9 bits/s. The detailed analysis of the effect of the quantization parameter (QP) on the total encoding time is presented for four different QP values. The encoding time is drastically reduced with a negligible loss in the image quality.