COIN: COmpression with Implicit Neural representations

TL;DR

This paper introduces a novel image compression method that encodes images as the weights of a neural network fitted to pixel data, outperforming JPEG at low bit-rates without complex entropy coding.

Contribution

The paper presents a simple neural network-based image compression approach that stores network weights instead of pixel data, demonstrating promising results at low bit-rates.

Findings

Outperforms JPEG at low bit-rates without entropy coding

Uses a neural network to implicitly represent images

Offers attractive properties for neural data compression

Abstract

We propose a new simple approach for image compression: instead of storing the RGB values for each pixel of an image, we store the weights of a neural network overfitted to the image. Specifically, to encode an image, we fit it with an MLP which maps pixel locations to RGB values. We then quantize and store the weights of this MLP as a code for the image. To decode the image, we simply evaluate the MLP at every pixel location. We found that this simple approach outperforms JPEG at low bit-rates, even without entropy coding or learning a distribution over weights. While our framework is not yet competitive with state of the art compression methods, we show that it has various attractive properties which could make it a viable alternative to other neural data compression approaches.