A Learnable Color Correction Matrix for RAW Reconstruction

TL;DR

This paper introduces a lightweight learnable color correction matrix for RAW image reconstruction from sRGB images, improving RAW-domain object detection performance with a simple model.

Contribution

The paper proposes a novel, ultra-lightweight RAW reconstruction method using a single convolutional layer to learn a color correction matrix, simplifying inverse ISP approximation.

Findings

SimRAW images improve object detection performance.

The method achieves comparable results to complex inverse ISP techniques.

The approach is computationally efficient and practical.

Abstract

Autonomous driving algorithms usually employ sRGB images as model input due to their compatibility with the human visual system. However, visually pleasing sRGB images are possibly sub-optimal for downstream tasks when compared to RAW images. The availability of RAW images is constrained by the difficulties in collecting real-world driving data and the associated challenges of annotation. To address this limitation and support research in RAW-domain driving perception, we design a novel and ultra-lightweight RAW reconstruction method. The proposed model introduces a learnable color correction matrix (CCM), which uses only a single convolutional layer to approximate the complex inverse image signal processor (ISP). Experimental results demonstrate that simulated RAW (simRAW) images generated by our method provide performance improvements equivalent to those produced by more complex inverse ISP methods when pretraining RAW-domain object detectors, which highlights the effectiveness and practicality of our approach.