High-Frequency First: A Two-Stage Approach for Improving Image INR

TL;DR

This paper introduces a two-stage training method for image implicit neural representations that emphasizes high-frequency details early on, improving reconstruction quality by addressing neural spectral bias.

Contribution

It presents a novel two-stage training strategy that adaptively emphasizes high-frequency details, offering a new approach to mitigate spectral bias in image INRs.

Findings

Improves image reconstruction quality over baseline INRs.

Effectively emphasizes high-frequency details during training.

Complementary to existing INR methods.

Abstract

Implicit Neural Representations (INRs) have emerged as a powerful alternative to traditional pixel-based formats by modeling images as continuous functions over spatial coordinates. A key challenge, however, lies in the spectral bias of neural networks, which tend to favor low-frequency components while struggling to capture high-frequency (HF) details such as sharp edges and fine textures. While prior approaches have addressed this limitation through architectural modifications or specialized activation functions, we propose an orthogonal direction by directly guiding the training process. Specifically, we introduce a two-stage training strategy where a neighbor-aware soft mask adaptively assigns higher weights to pixels with strong local variations, encouraging early focus on fine details. The model then transitions to full-image training. Experimental results show that our approach consistently improves reconstruction quality and complements existing INR methods. As a pioneering attempt to assign frequency-aware importance to pixels in image INR, our work offers a new avenue for mitigating the spectral bias problem.