ResNet-Driven Joint Decision-Making for VVC Optimization via Gradient Search
Year: 2025
Type: article
Abstract: The latest video coding standard, H.266/VVC, introduces novel intra-coding techniques, advancing intra-coding technology while significantly increasing its complexity.Compared to H.265/HEVC, H.266/VVC introduces a new QTMT structure for CU partitioning and nearly doubles the number of intra-prediction modes, leading to a substantial increase in computational complexity and encoding time.To mitigate the exponential computational demands inherent in quad-tree based partitioning, this research establishes a rate-distortion aware optimization framework. The proposed dynamic decision-making mechanism coordinates coding unit hierarchy configuration with spatial prediction modalities, achieving Pareto efficiency in computational resource allocation.The algorithm consists of two main components: a fast CU partitioning module and an intra-mode selection module.In the first component of the proposed algorithm, we present a rapid CU partitioning technique built upon the ResNet model. This method effectively extracts and learns the distinctive features of CU blocks, enabling it to predict partition structures with high accuracy, thus streamlining the CU decision process.In the second component, we introduce an intra-mode selection algorithm based on gradient descent, which utilizes block gradient features to optimize and determine the final intra-prediction mode.Experimental results demonstrate that the proposed algorithm significantly reduces encoding time by 53.17%.At the same time, the BD-BR only increases by 1.38%, showcasing an optimal trade-off between video quality and encoding efficiency.
Source: IEEE Access
Authors: Fangmei Liu, Jiyuan Wang, Qiuwen Zhang
Institution: Zhengzhou University of Light Industry
Language: English
Subfield: Signal Processing
Field: Computer Science
Domain: Physical Sciences
Open Access status: gold
Estimated APC paid: $2,075
OpenAlex ID: W4411406298