CPDDNet: Color-Polarization Denoising and Demosaicking Network

2026-07-01

AI Summary

This work addresses the degraded reconstruction quality of color and polarization information from CPFA sensors in single-shot imaging, particularly under low-light conditions where noise and resolution loss are pronounced. To overcome this challenge, the authors propose an end-to-end joint denoising and demosaicking network that unifies both tasks within a single framework for the first time, explicitly modeling their inherent coupling. A dedicated feature fusion module is introduced to preserve shared low-level features from the raw CPFA data while jointly recovering fine texture details and physically meaningful polarization information. Extensive experiments on real-world datasets demonstrate that the proposed framework significantly outperforms existing methods, achieving notable improvements in both reconstructed image fidelity and polarization parameter accuracy.

Abstract

Color-polarization imaging using a color-polarization filter array (CPFA) sensor captures both texture (color intensity) and physical (polarization) information of the scene in a single shot, enabling various applications in computer vision. However, the raw mosaic output from a CPFA sensor often suffers from severe noise and resolution loss, especially under low-light conditions. Existing methods generally focus on either denoising or demosaicking tasks, failing to capture the coupling between them and neglecting shared low-level features. In this paper, we propose a color-polarization denoising and demosaicking network (CPDDNet), which is a joint framework that performs noise removal and CPFA interpolation using a feature fusion module that retains the features from the CPFA raw data at both the denoising and the demosaicking stages. Experimental results demonstrate that CPDDNet significantly enhances image quality and polarization parameter accuracy, outperforming existing approaches on a real dataset.

Problem

• color-polarization imaging
• CPFA sensor
• denoising
• demosaicking
• low-light noise

Innovation

• color-polarization imaging
• joint denoising and demosaicking
• feature fusion module
• CPFA sensor
• deep learning