CNN-based driving of block partitioning for intra slices encoding

TL;DR

This paper introduces a CNN-based method for optimizing block partitioning in intra slice encoding, balancing complexity and coding efficiency, and demonstrating significant speed-ups with minimal quality loss.

Contribution

It presents a novel deep learning approach to replace heuristic partitioning in intra slice encoding, enabling faster processing with controlled quality trade-offs.

Findings

Achieves 2x speed-up without BD-rate loss

Attains over 4x speed-up with less than 1% BD-rate loss

Demonstrates effectiveness in JVET intra coding scenarios

Abstract

This paper provides a technical overview of a deep-learning-based encoder method aiming at optimizing next generation hybrid video encoders for driving the block partitioning in intra slices. An encoding approach based on Convolutional Neural Networks is explored to partly substitute classical heuristics-based encoder speed-ups by a systematic and automatic process. The solution allows controlling the trade-off between complexity and coding gains, in intra slices, with one single parameter. This algorithm was proposed at the Call for Proposals of the Joint Video Exploration Team (JVET) on video compression with capability beyond HEVC. In All Intra configuration, for a given allowed topology of splits, a speed-up of $\times 2$ is obtained without BD-rate loss, or a speed-up above $\times 4$ with a loss below 1\% in BD-rate.