Highly efficient near lossless video compression using selective intra prediction for HEVC lossless mode

Lossless coding mode of high efficiency video coding (HEVC) relies on efficient intra prediction methods for achieving higher compression. Sample based angular intra prediction (SAP) is a modification proposed for the conventional block-based intra prediction (BP) in the HEVC standard. In this paper a selective intra prediction (SIP) algorithm is proposed for lossless intra mode coding within HEVC, which uses an adaptive switching strategy between the two intra prediction methods, BP and modified SAP (MSAP), on a per pixel basis. MSAP is a modification proposed in this work to improve the prediction accuracy of SAP. The proposed SIP algorithm is a near lossless algorithm that selects the best prediction for each pixel to reduce the residual energy thereby increasing the coding efficiency. To circumvent the huge overhead required for the transmission of the selection from encoder to decoder, SIP adopts piggybacking the selection on the least-significant-bit (LSB) of the transmitted residual for each pixel. The proposed approach removes overhead at the expense of a ±1 reconstruction error for each pixel which is negligible for 8-bit or 10-bit input videos. Experimental results show significant coding gains for the proposed SIP over the HEVC anchor and similar state-of-the-art lossless coding algorithms.