Fig 1.
Process diagram of the proposed mode selection technique.
Fig 2.
Illustration of motion features and saliency feature generated at different blocks of 12th frame on Tennis video; (b~d) are the phase shifted plots for no motion (0.4), simple motion (0.7) and complex motion (0.8); (e-f) corresponds to the respective values generated by ECR and saliency feature for blocks at positions (3, 1), (3, 10) and (5, 7) respectively.
For clear visualization, we use 32×32 block size.
Fig 3.
Salience feature after applying GBVS on three different image blocks (GBVS modelling applied on the current image blocks) with different motions and texture of a frame identified in Fig 2; note that for clear visualization we use 32×32-pixel block.
Fig 4.
Illustration of motion and salient areas identification with and without saliency feature based cost function.
Fig 5.
Codebook of the proposed binary pattern templates for subset of inter-mode selection.
Table 1.
Mode selection at 32×32 block level based on the codebook of predefined binary pattern templates.
Table 2.
Subset of mode selection at 16×16 and 8×8 coding depth level based on the patterns of RDMS.
Fig 6.
The representation of motion map for various Th values at QP = 28.
The reddish blocks in (b~d) indicate the motion regions and the bluish blocks indicate the background regions.
Table 3.
Performance comparison of different fast mode selection algorithms compared to HEVC encoder in terms of BD-Bit Rate (BD-BR), BD-PSNR and computational time.
Table 4.
A theoretical analysis- percentage of time savings by the proposed method (against HM) for different sequence types based on average number of Inter-modes selected per CU.
Fig 7.
Illustration of overall average time savings by the proposed method against HM at different bit-rates for SVS.
Fig 8.
Illustration of overall average time savings by the proposed method against HM at different bit-rates for SCVS.
Fig 9.
Illustration of overall average time savings by the proposed technique against HM based on different video categories.
In (a), the numbers of videos used in HD, MV and SD are 4, 2 and 6 respectively, while in (b), these numbers for Class A, B, C, D and E are 1, 4, 2, 4, and 3 respectively.
Fig 10.
HM and proposed method based average mode selection at three coding depth levels.
Fig 11.
Block partitioning for the 12th frame of the Tennis video at QP = 24 with HM and the Proposed method.
Fig 12.
Comparative study on RD performance by HM12.1 and the proposed method for a wide range of bit-rates.
Table 5.
Performance comparison of proposed technique compared to HM12.1 using BD-BR and BD-PSNR for the SVS.
Table 6.
Performance comparison of proposed technique compared to HM12.1 using BD-BR and BD-PSNR for the SCVS.
Fig 13.
Subjective quality evaluation for HM12.1 and the proposed technique for Tennis video sequence.
The figures are achieved from the 20th frame of the Tennis video at the same bit-rate.