Towards Lossless Implicit Neural Representation via Bit Plane Decomposition

TL;DR

This paper introduces a bit-plane decomposition method for implicit neural representations (INRs) that enables lossless 2D image and audio fitting, even at high bit depths, by reducing the model size upper bound and accelerating convergence.

Contribution

The paper proposes a novel bit-plane decomposition technique that allows INRs to predict individual bits, achieving lossless representation and extending applications to compression and quantization.

Findings

Achieved lossless 2D image and audio fitting at 16-bit depth.

Demonstrated faster convergence with reduced upper bound on model size.

Identified the significance of the most significant bit (MSB) bias in INRs.

Abstract

We quantify the upper bound on the size of the implicit neural representation (INR) model from a digital perspective. The upper bound of the model size increases exponentially as the required bit-precision increases. To this end, we present a bit-plane decomposition method that makes INR predict bit-planes, producing the same effect as reducing the upper bound of the model size. We validate our hypothesis that reducing the upper bound leads to faster convergence with constant model size. Our method achieves lossless representation in 2D image and audio fitting, even for high bit-depth signals, such as 16-bit, which was previously unachievable. We pioneered the presence of bit bias, which INR prioritizes as the most significant bit (MSB). We expand the application of the INR task to bit depth expansion, lossless image compression, and extreme network quantization. Our source code is available at https://github.com/WooKyoungHan/LosslessINR