紀要論文

JVCSR : VIDEO COMPRESSIVE SENSING RECONSTRUCTION WITH JOINT IN-LOOP REFERENCE ENHANCEMENT AND OUT-LOOP SUPER-RESOLUTION

eng

24368083

https://doi.org/10.15002/00025362

法政大学大学院理工学研究科

法政大学大学院紀要. 理工学研究科編

63

1

7

2022-03-24

Version of Record

Recently, deep learning-based video compressive sensing reconstruction (VCSR) technologies have achieved tremendous success in improving the quality of the reconstructed video. However, the reconstructed video quality at low bit rates or high compression ratios still dissatisfies with the requirement in practice. In this paper, a video compressive sensing reconstruction method with joint in-loop reference enhancement and out-loop super-resolution (JVCSR) is proposed, which focuses on removing reconstruction noises, blocking artifacts and increase the resolution simultaneously. As an in-loop part, the enhanced frame is utilized as a reference to improve the recovery performance of current frame. Furthermore, it is the first work that realizes out-loop super-resolution for VCSR to obtain high quality image at low bit rates. As a result, our JVCSR can improve average of 2.53 dB PSNR by comparing with state-of-the-art compressive sensing methods at the similar bit rate.

Article