Piecewise Mapping in HEVC Lossless Intra-Prediction Coding

DPCM-Based Edge Prediction for Lossless Screen Content Coding in HEVC

“…The modes in SEAP are implemented in all prediction unit (PU) sizes available in the quadtree-based coding structure, with a maximum CU size of 64×64 and minimum PU size of 4×4. 13 P x,y = b 7 P x,y = (3 * a + b) >>2 14 P x,y = (7 * b + c)>>3 8 P x,y = (5 * a + 3*b) >>3 15 P x,y = (3 * b + c)>>2 9 P x,y = (a +b) >>1 16 P x,y = (5 * b + 3 * c)>>3 10 P x,y = (3*a + 5*b) >>3 17 P x,y = (b + c)>>1 11 P x,y = (a + 3 * b) >>2 18 Our previous work in [7] has shown that among several DPCM-based methods, SAP-E and SAP+SWP2+DTM attain the best lossless coding performance for a variety of SC and camera-captured sequences compared to current intraprediction and RDPCM, which is the DPCM-based method standardized in HEVC. Therefore, the first set of evaluation results (Section IV.A) compares SEAP, SAP-E, and SAP+SWP2+DTM with RDPCM.…”

“…Although RDCPM is less effective with repeated patterns than IBC, this tool is capable of outperforming IBC for lossless coding of camera-captured material where repeated patterns are not common [4,7]. It is also important to note that RDCPM improves coding efficiency without a considerable increase in encoding times, compared to the encoding times required by IBC [7,9]. This is done by employing simple sample-wise operations to shorten the prediction distance along the same direction as the blockwise intra-prediction.…”

“…It is important to mention that the coding improvements attained by SEAP+ALL can be further improved if the energy of the residual blocks computed by the DPCM-based modes is reduced. We have recently shown in [7] that this can be attained by applying piecewise mapping functions on the residual blocks so that specific residual values are mapped to unique lower values. As part of our ongoing work, we are applying the ideas introduced in [7] to the modes in SEAP to further improve lossless coding performance when all SCC tools are enabled.…”

“…Among these SCC tools, IBC has been shown to provide important coding improvements for sequences with numerous repeated patterns. For those SC sequences not characterized by depicting numerous repeated patterns, RDPCM has also been shown to provide important improvements to lossless intra-coding [6,7].…”

“…In [17,18], the prediction accuracy of RDPCM is improved by exploiting the gradient information of neighboring samples. Recently, we introduce in [7] the usage of piecewise mapping functions to map specific residual values of DPCM-predicted blocks to unique lower values, thus reducing the overall energy of the residual signal.…”