Substitutional Neural Image Compression

TL;DR

SNIC is a novel neural image compression enhancement technique that creates substitutional images to improve compression quality and enable bit-rate control without additional data or tuning.

Contribution

It introduces a method to enhance neural image compression by generating substitutional images through differentiable optimization, improving performance and control.

Findings

Improves compression quality across various models and metrics.

Enables effective bit-rate control with a single trained model.

Demonstrates competitive rate-distortion performance.

Abstract

We describe Substitutional Neural Image Compression (SNIC), a general approach for enhancing any neural image compression model, that requires no data or additional tuning of the trained model. It boosts compression performance toward a flexible distortion metric and enables bit-rate control using a single model instance. The key idea is to replace the image to be compressed with a substitutional one that outperforms the original one in a desired way. Finding such a substitute is inherently difficult for conventional codecs, yet surprisingly favorable for neural compression models thanks to their fully differentiable structures. With gradients of a particular loss backpropogated to the input, a desired substitute can be efficiently crafted iteratively. We demonstrate the effectiveness of SNIC, when combined with various neural compression models and target metrics, in improving compression quality and performing bit-rate control measured by rate-distortion curves. Empirical results of control precision and generation speed are also discussed.