Rethinking Skip Connections: Additive U-Net for Robust and Interpretable Denoising

TL;DR

This paper introduces the Additive U-Net, a novel architecture replacing concatenative skip connections with learnable additive ones, improving interpretability and robustness in image denoising tasks.

Contribution

It proposes a new skip connection method that enhances interpretability and efficiency, avoiding channel inflation and enabling better control over information flow.

Findings

Achieves competitive denoising performance on Kodak-17 benchmark.

Demonstrates robustness across various noise levels and kernel schedules.

Learns a natural progression from high-frequency to low-frequency features without explicit hierarchical design.

Abstract

Skip connections are central to U-Net architectures for image denoising, but standard concatenation doubles channel dimensionality and obscures information flow, allowing uncontrolled noise transfer. We propose the Additive U-Net, which replaces concatenative skips with gated additive connections. Each skip pathway is scaled by a learnable non-negative scalar, offering explicit and interpretable control over encoder contributions while avoiding channel inflation. Evaluations on the Kodak-17 denoising benchmark show that Additive U-Net achieves competitive PSNR/SSIM at noise levels {\sigma} = 15, 25, 50, with robustness across kernel schedules and depths. Notably, effective denoising is achieved even without explicit down/up-sampling or forced hierarchies, as the model naturally learns a progression from high-frequency to band-pass to low-frequency features. These results position additive skips as a lightweight and interpretable alternative to concatenation, enabling both efficient design and a clearer understanding of multi-scale information transfer in reconstruction networks.