Interpreting CNN for Low Complexity Learned Sub-pixel Motion Compensation in Video Coding

TL;DR

This paper introduces a neural network-based method for low-complexity sub-pixel motion compensation in video coding, achieving significant bit savings while reducing neural network complexity.

Contribution

It presents a novel approach to interpret learned interpolation filters, reducing neural network complexity in video coding applications.

Findings

Up to 4.5% BD-rate savings in VVC sequences.

Significant reduction in neural network complexity.

Improved interpolation accuracy for fractional motion compensation.

Abstract

Deep learning has shown great potential in image and video compression tasks. However, it brings bit savings at the cost of significant increases in coding complexity, which limits its potential for implementation within practical applications. In this paper, a novel neural network-based tool is presented which improves the interpolation of reference samples needed for fractional precision motion compensation. Contrary to previous efforts, the proposed approach focuses on complexity reduction achieved by interpreting the interpolation filters learned by the networks. When the approach is implemented in the Versatile Video Coding (VVC) test model, up to 4.5% BD-rate saving for individual sequences is achieved compared with the baseline VVC, while the complexity of learned interpolation is significantly reduced compared to the application of full neural network.