Utilize Neighboring LCU Depth Information to speedup FVC/H.266 Intra Coding
Autor: | Cheng-Hwa Lee, Jiann-Jone Chen, Yi-Ying Chiu, Yao-Hong Tsai |
---|---|
Rok vydání: | 2019 |
Předmět: |
Binary tree
Speedup Artificial neural network business.industry Computer science 020206 networking & telecommunications 02 engineering and technology Coding tree unit 0202 electrical engineering electronic engineering information engineering Wireless Quadtree 020201 artificial intelligence & image processing business Time complexity Algorithm Coding (social sciences) |
Zdroj: | 2019 International Conference on System Science and Engineering (ICSSE). |
DOI: | 10.1109/icsse.2019.8823436 |
Popis: | With the advance of the 5th generation (5G) wireless system technology, the JVET started to develop a FVC (Future Video Coding) standard for the ultra-high definition video (UHDV) since 2016. We study how to speed up the H.266 coding without degrading the coding quality. The FVC/H.266 adopts QuadTree plus Binary Tree (QTBT) structure for Coding Units (CU). We proposed to reference the average depth information of neighboring Largest Coding Unit (LCU) to determine whether to early terminate CU decomposition or not. By utilizing the coding modes of neighboring CUs, it can effectively eliminate unnecessary rate-distortion optimization (RDO) operations. Correlations of coding modes between neighboring LCUs are not strong and the mode prediction rule of a current CU is developed based on heuristic approaches. Experiments showed that the proposed method can save up 25.42% of processing time, while the BDBR rises only 0.31%, as compared to the JEM system program. How to reduce the time complexity of HEVC/H.265 and FVC/H.266 can be formulated as a problem solving through neural network models, which is expected to yield much more time complexity reduction under the same video coding quality. |
Databáze: | OpenAIRE |
Externí odkaz: |