H.265 / HEVC intra prediction fast algorithm (1)

Recently, the Papers of the Fast Algorithm in the H.265 / HEVC intraframe is looking for it, I want to find IDEA used in H.266 / VVC. (I wrote this article in November 2018, but I have not sent it. At present, the fast algorithm of VVC has been completed. In exchange for 46% of the coded time in 0.91% BDR LOSS, the coded time is declining, and the MMM2020 is submitted, and this algorithm is shared. .) The idea of ​​HEVC intra prediction optimization has the following two: When CTU is divided, reduce the number of inspections - Skip certain divisions, terminate / divide in advance; mode decisions, reduce the IPM number of inspections (intra predes). When pattern decisions, reduce the IPM number of inspections (Intra Prediction Modes). Idea 1: When CTU is divided, terminate / divide in advance The characteristics of this algorithm and video sequences are related to the texture characteristics of the CTU. In general, flat, homologous areas are usually divided into "large blocks"; and the textured areas are often divided into "small blocks". In some papers, the authors put forward some ways to measure the complexity of CTU / CU to terminate or divide in advance. If a CTU is determined as simple block, it can be directly divided, thereby avoiding the next series of division overhead, saving time; if it is determined to be a complex block, you can directly divide, no need for COST calculations, save time. [ISIEA2013] This indicator of variance is used to measure, and it is better to understand that the CTU / CU brightness variance is smaller, lower than the threshold, and the homogeneous, and early termination. The experimental results ΔT_enc = 55%, BDR is only 0.63%. The effect is excellent. [DICTAP2014] The calculated index-λmode is used, which reflects COST, and the smaller the prediction mode, the better. The authors believe that when λmode is small enough, it can stop dividing when it is below a threshold. The experimental results ΔT_enc = 31%, BDR is only 0.70%. The effect is better. [ICalip2014] is the paper published by the laboratory brother, and he uses the Cu information encoding the corresponding position of the frame adjacent frame. The experimental results ΔT_enc = 36%, BDR is only 1.20%. General effect. In addition, there are some papers, here is not listed. Idea 2: Mode decision, reduce the number of IPMs In HEVC, when the brightness component intra prediction, each PU has 35 prediction mode, the perspective mode in PLANAR, DC, and 33. There are also a lot of papers to reduce the encoding time.

Our other product:

Professional lightning protection system for Tower	Professional FM Radio Station Equipment Package	FMUSER FU-1000C 1000W FM Transmitter With USB Audio Input Interface
FMUSER 200W FM Radio Broadcasting Package	Hotel IPTV Solution	FMUSER FSN-1000T V5.0 1KW FM Radio Broadcast Transmitter