Lightweight Unpaired Smartphone ISP Transfer with Semantic Pseudo-Pairing

2026-05-08

AI Summary

This work addresses the challenge of transferring mobile image signal processing (ISP) pipelines across unpaired RAW and RGB images, where misalignment in scene content and color hinders effective learning. To overcome this, the authors propose a lightweight ISP method based on semantic pseudo-pairing. By reconstructing high-resolution images to preserve global context and leveraging DINOv2 for semantic embeddings, they establish reliable pseudo-supervision through Gromov–Wasserstein optimal transport at both image and patch levels to achieve semantic alignment. Built upon this framework, a compact CNN with only 7K parameters is designed to focus specifically on color mapping, significantly enhancing training stability and reducing artifacts. Evaluated on the NTIRE 2026 Challenge hidden test set, the method achieves a PSNR of 22.569, SSIM of 0.675, and ΔE of 8.067, ranking third in both SSIM and ΔE, thereby demonstrating its effectiveness and competitiveness.

Abstract

Unpaired smartphone ISP is a challenging problem due to the lack of scene and color alignment between RAW and target RGB images. Many existing methods either require paired data or rely heavily on adversarial training, which can become unstable in the unpaired setting. In this work, we present a simple and effective approach developed for the NTIRE 2026 Learned Smartphone ISP Challenge with Unpaired Data. Our method first reconstructs larger images from training patches to recover global context. Then, we extract semantic embeddings with DINOv2, and use fused Gromov-Wasserstein (FGW) optimal transport to build pseudo pairs between RAW and RGB images at both image and patch levels. This semantic matching allows us to partially alleviate the unpairedness of the data and build these pseudo input-target pairs. Based on these pseudo pairs, we train a lightweight CNN with only 7K parameters for color rendering. The network is designed to be compact and focus on color transformation rather than structural change, which helps reduce artifacts and improve training stability. Our challenge submission achieves 22.569 PSNR, 0.675 SSIM, and 8.067 $ΔE$ on the final hidden test set, significantly improving over the baseline and achieving the 3rd best SSIM and $ΔE$ among all challenge entries. Our code is available at github.com/nuniniyujin/Unpaired-ISP .

Problem

• unpaired smartphone ISP
• RAW-to-RGB conversion
• semantic alignment
• image signal processing
• color rendering

Innovation

• semantic pseudo-pairing
• unpaired ISP
• Gromov-Wasserstein optimal transport
• lightweight CNN
• DINOv2