Spatio-temporal progressive optimization network for video bit depth enhancement

• Published in: Multimedia systems Vol. 30; no. 5
• Main Authors: Li, Qingying, Lin, Xin, Liu, Jing, Su, Yuting, Ma, Rui
• Format: Journal Article
• Language: English
• Published: Berlin/Heidelberg Springer Berlin Heidelberg 01.10.2024; Springer Nature B.V
• Subjects: Algorithms; Computer Communication Networks; Computer Graphics; Computer Science; Cryptology; Data Storage Representation; Frames (data processing); Ghosting; Image enhancement; Image quality; Image reconstruction; Machine learning; Multimedia Information Systems; Operating Systems; Optimization; Regular Paper; Spatiotemporal data; Deep learning; Spatio-temporal feature fusion; Video bit-depth enhancement
• ISSN: 0942-4962; 1432-1882
• DOI: 10.1007/s00530-024-01474-x

The advancement of bit-depth enhancement has yielded remarkable outcomes in the realm of reconstructing high-quality images, yet its application to video enhancement has been hindered by structural distortions present in non-aligned low bit-depth frames. The structural distortion impedes efficient spatio-temporal modeling and gives rise to pronounced ghosting and blurring artifacts, particularly evident during motion across consecutive frames. In response, this paper introduces a two-stage Spatio-Temporal Progressive Optimization network tailored for video bit-depth enhancement, aiming to deliver superior performance on High Bit Depth devices. The impact of structural distortions from neighboring frames is progressively eliminated in a coarse-to-fine reconstruction approach. Spatio-Temporal Attention and Temporal Attention are designed to initially enhance the neighboring frames and subsequently supplements the lost details of the target frame. Experiments demonstrate that the proposed algorithm outperforms other traditional and deep learning methods in terms of both subjective and objective evaluations on multiple datasets.