RPBA-Net: An Interpretable Residual Pyramid Bilateral Affine Network for RAW-Domain ISP Enhancement

TL;DR

RPBA-Net is an interpretable neural network that enhances RAW images by unifying demosaicing and enhancement, achieving state-of-the-art results with low complexity suitable for mobile deployment.

Contribution

It introduces a residual pyramid bilateral affine network with hierarchical modeling and regularization for interpretable RAW-domain ISP enhancement.

Findings

Outperforms existing RAW-to-sRGB methods in fidelity and perceptual quality.

Maintains low model complexity suitable for mobile deployment.

Achieves state-of-the-art results in experiments.

Abstract

To address module fragmentation, uninterpretable mappings, and deployment constraints in RAW-domain demosaicing, color correction, and detail enhancement, this paper proposes RPBA-Net, an interpretable residual pyramid bilateral affine network for RAW-domain ISP enhancement. Given packed RAW as input, the method performs residual affine base reconstruction by estimating a base RGB representation and learning identity-guided residual affine corrections, thereby unifying demosaicing and enhancement. It further builds pyramid bilateral affine grids and combines guide-driven autoregressive adaptive slicing with adaptive cross-layer fusion to hierarchically model global tone restoration and local texture enhancement. In addition, smoothness, cross-scale consistency, and magnitude regularization terms are introduced to improve model stability, controllability, and structural interpretability. Extensive experiments demonstrate that RPBA-Net surpasses representative RAW-to-sRGB methods and achieves state-of-the-art performance in reconstruction fidelity and perceptual quality, while maintaining low model complexity and strong deployment potential for mobile and embedded platforms.