A fast CU decision using image variance in HEVC intra coding

Abstract

As the next generation standard of video coding, high efficiency video coding (HEVC) achieves high coding efficiency. The coding unit (CU) was adopted as the processing unit for HEVC to improve the coding efficiency. However, determining the optimal distribution of CU sizes requires lengthy calculation. To reduce the calculation time in intra prediction processing, we developed a method that determines the CU sizes using the variance value of the input image. Experimental results show that the proposed method reduces encoding time by about 40-70% compared to that of a conventional HEVC test model.

Fast Algorithm with Early Termination CU Split and Mode Decision

“…Figure 5 shows the rate distortion curve of the four test sequences including "PeopleOnStreet", "BQTerrace", "PartyScene" and "BasketballPass" with three algorithms. We can observe that, compared with the original algorithm HM13.0 and the fast algorithm proposed in literature [7], our proposed algorithm has almost the same coding efficiency with them. At the same time it can also reduce the computational complexity, save the coding time, and improve the encoding speed.…”

“…The proposed algorithm can reduce coding time, on average, about 22.8% with the maximum of 26% in "PeopleOnStreet (2560×1600)" and the minimum of 20% in "BasketballPass (416×240)". At the same time, the proposed fast prediction algorithm achieves better RD performance, where the average coding efficiency loss in term of PSNR is only 0.04dB and in term of bit rate is about 0.34% compared with the reference algorithm in the [7]. Figure 5 shows the rate distortion curve of the four test sequences including "PeopleOnStreet", "BQTerrace", "PartyScene" and "BasketballPass" with three algorithms.…”

“…The proposed algorithm is evaluated with QPs 22, 27, 32 and 37 [13]. In the above-mentioned test conditions, the proposed algorithm in this paper is compared with HM13.0 and the fast algorithm proposed in literature [7] in the three aspects consisted of the encoding time, PSNR, and bitrate [14]. The experimental results are shown in Table 2, Table 3.…”

“…Therefore, the proposed algorithm can reduce the coding time, and effectively improve the coding efficiency while maintaining the video quality unchanged as HM13.0. Table 3 shows performances of proposed fast prediction algorithm compared to the fast algorithm in the literature [7]. The proposed algorithm can reduce coding time, on average, about 22.8% with the maximum of 26% in "PeopleOnStreet (2560×1600)" and the minimum of 20% in "BasketballPass (416×240)".…”

“…In addition to select N candidates, Zhao Liang, et al, [6] made full use of the correlation between neighboring pixels in the HM2.0 and added the MPM (the best prediction mode of the upper and left PU) to the candidate sets, which improved the video quality. Jiang, et al, [7] proposed a fast intra coding algorithm, detected whether the current coding unit belongs to the smooth region. According to the detection result, it can help determine how to adaptively skip the coding process of the coding unit sizes not suitable for the texture features.…”