DCNGAN: A Deformable Convolutional-Based GAN with QP Adaptation for Perceptual Quality Enhancement of Compressed Video

TL;DR

This paper introduces DCNGAN, a deformable convolutional GAN that adaptively enhances the perceptual quality of compressed videos by leveraging multi-frame information and QP adaptation, outperforming existing methods.

Contribution

The paper presents a novel deformable convolution-based GAN that efficiently aligns multiple frames and adapts to QPs for superior compressed video quality enhancement.

Findings

DCNGAN outperforms state-of-the-art algorithms in quality enhancement.

Deformable convolutions effectively align multiple frames for better results.

Adaptive QP handling improves perceptual quality across different compression levels.

Abstract

In this paper, we propose a deformable convolution-based generative adversarial network (DCNGAN) for perceptual quality enhancement of compressed videos. DCNGAN is also adaptive to the quantization parameters (QPs). Compared with optical flows, deformable convolutions are more effective and efficient to align frames. Deformable convolutions can operate on multiple frames, thus leveraging more temporal information, which is beneficial for enhancing the perceptual quality of compressed videos. Instead of aligning frames in a pairwise manner, the deformable convolution can process multiple frames simultaneously, which leads to lower computational complexity. Experimental results demonstrate that the proposed DCNGAN outperforms other state-of-the-art compressed video quality enhancement algorithms.