Fast inter mode decision process for HEVC encoder

Efficient prediction of CU depth and PU mode for fast HEVC encoding using statistical analysis

“…The second part shows the performance of only the proposed fast PU mode decision method, which has been explained in Section 3. The state-of-the-art method we compared with is [15], denoted as REF2, since it is also a HEVC encoding method with a fast PU decision. Finally, the test results of the overall algorithm including both the proposed fast CU size decision and fast PU modes decision methods are shown and compared with REF1, REF2, and another state-of-the art work [16], denoted as REF3.…”

“…Comparison results between the proposed fast PU mode decision method in Section 3 and the REF2 [15] (also a fast PU algorithm) are shown in Table 7 with LP Configuration profile and Table 8 with RA Configuration profile.…”

“…This method reduces encoding time by 46.5%. In [15], a fast inter mode decision method utilizing correlated tendency of PU mode is introduced to decrease encoding time; for each CU depth, when the best mode of inter 2N Â 2N_PU is selected as skip, the Advanced Motion Vector Prediction (AMVP) process is skipped in sub-partition PUs of current CU, and only Merge mode decision process will be used. Another fast inter mode decision algorithm for HEVC is proposed in [16] by jointly using the inter level correlation of the quad-tree structure and the spatiotemporal correlation, which contains early skip mode decision, prediction-size-correlation-based mode decision and RDcost-correlation-based mode decision.…”

“…Meanwhile, the values of BDBR (Bjontegaard Delta Bit Rate) are only 1.0517% and 0.9918%, respectively, which means that the degradation of video quality is acceptable. Compared with state-of-the-art algorithms [11,15,16], and two built-in fast methods in HM, which are HM with ESD (Early Skip Decision) (ESD set to 1) and HM with CBF (Coded Block Flag) (CFM, CBF Fast Mode, set to 1), our algorithm has higher time saving, and it maintains good bitrate performances.…”

A novel motion classification based intermode selection strategy for HEVC performance improvement

“…Their tested results affirm the time savings approximately 30% with 0.5% quality loss. Yang et al [16] propose a fast intermode decision method for HEVC by checking whether the best prediction mode of their 2N×2N is skip mode or not by utilizing the correlated tendency of PU mode. This method reduces 39% computational time by sacrificing 0.8% bit-rate.…”