Fast intra algorithm based on texture characteristics for 360 videos

Abstract

With the rapid progress of virtual reality technology, 360 videos have become increasingly popular. Given that the resolution of a 360 video is ultra-high (generally 4K to 8K), the encoding time for this type of video is considerably high. To reduce encoding complexity, this study proposed a fast intra algorithm that is based on image texture characteristics. On the one hand, the proposed algorithm determines whether to terminate the coding unit partition early on the basis of texture complexity. On the other hand, the proposed algorithm reduces the number of candidate modes in mode decision according to texture directivity. Experimental results showed that the proposed algorithm can obtain an average time reduction rate of 53% and a Bjontegaard delta rate increase of only 1.3%, which is acceptable for rate distortion performance.

Exploring ERP Distortions to Reduce the Encoding Time of 360 Videos

“…When it is combined with a previous work [11] (IntraTech line), the encoding time is reduced by 23.09% with a BD-BR of 1.21%. When compared to related works, it is visible that IntraTech is outperformed by [14] and [15]. However, both related works are mainly based on generic complexity reduction techniques that are applied in the same manner for the entire frame, irrespective of projection distortion.…”

“…The average results for each algorithm are presented in Table 1, where InterTech is the result of employing all interframes prediction algorithms simultaneously (i.e., Early Skip, Reduced SR, and Reduced FME), and IntraTech is the combination of Reduced Intra Sizes and a previous work published at [11]. Table I also presents the average results of related works: Liu [12] and Ray [13] present proposals to accelerate the interframes prediction of ERP 360 videos, whereas Wang [14] and Zhang [15] propose to accelerate the intraframe prediction of ERP 360 videos. Liu [12] uses a KNN classifier to discard some unlikely block sizes during interframes prediction; Ray [13] discards unlikely block sizes during interframes prediction and reduces FME precision based on block position; Wang [14] discards some prediction modes and block sizes during intraframe prediction based on block position; and Zhang [15] evaluates the texture of the blocks to evaluate a subset of prediction modes and block sizes.…”

“…Table I also presents the average results of related works: Liu [12] and Ray [13] present proposals to accelerate the interframes prediction of ERP 360 videos, whereas Wang [14] and Zhang [15] propose to accelerate the intraframe prediction of ERP 360 videos. Liu [12] uses a KNN classifier to discard some unlikely block sizes during interframes prediction; Ray [13] discards unlikely block sizes during interframes prediction and reduces FME precision based on block position; Wang [14] discards some prediction modes and block sizes during intraframe prediction based on block position; and Zhang [15] evaluates the texture of the blocks to evaluate a subset of prediction modes and block sizes. It must be noted that Ray [13] uses the encoder Joint Exploration Model (JEM) [16], an exploratory encoder built on top of HEVC to evaluate new encoding tools for future video coding standards.…”

“…It must be noted that Ray [13] uses the encoder Joint Exploration Model (JEM) [16], an exploratory encoder built on top of HEVC to evaluate new encoding tools for future video coding standards. The works [12] [14] [15] use the same HEVC encoder as this work -the HM [7].…”

“…In this context, the contribution of this work is twofold: provide evidence on the impacts that ERP distortion causes in the encoding, and propose algorithms to exploit such evidence and accelerate the encoding. [14] 24.50% 0.24% Zhang [15] 53.00% 1.30% *** This work uses a different encoder…”