Enhancing HEVC Compressed Videos with a Partition-masked Convolutional Neural Network

TL;DR

This paper introduces a partition-masked CNN that leverages HEVC encoding partition data to enhance compressed video quality more effectively than previous methods, achieving significant bitrate savings.

Contribution

It presents a novel CNN approach that incorporates coding unit size information for improved HEVC video enhancement, surpassing existing CNN-based methods.

Findings

Over 9.76% BD-rate saving on benchmark sequences

State-of-the-art performance in HEVC video enhancement

Effective utilization of partition information improves quality

Abstract

In this paper, we propose a partition-masked Convolution Neural Network (CNN) to achieve compressed-video enhancement for the state-of-the-art coding standard, High Efficiency Video Coding (HECV). More precisely, our method utilizes the partition information produced by the encoder to guide the quality enhancement process. In contrast to existing CNN-based approaches, which only take the decoded frame as the input to the CNN, the proposed approach considers the coding unit (CU) size information and combines it with the distorted decoded frame such that the degradation introduced by HEVC is reduced more efficiently. Experimental results show that our approach leads to over 9.76% BD-rate saving on benchmark sequences, which achieves the state-of-the-art performance.