BSRAW: Improving Blind RAW Image Super-Resolution
This addresses the problem of improving image quality for smartphone and compact camera users by working directly with RAW data to avoid complications from ISP transformations, though it is incremental as it builds on existing RAW super-resolution methods.
The paper tackles blind image super-resolution in the RAW domain by designing a realistic degradation pipeline that considers sensor noise, defocus, and exposure issues, resulting in models that can upscale real-scene RAW images and improve quality, with a new DSLM dataset and benchmark introduced.
In smartphones and compact cameras, the Image Signal Processor (ISP) transforms the RAW sensor image into a human-readable sRGB image. Most popular super-resolution methods depart from a sRGB image and upscale it further, improving its quality. However, modeling the degradations in the sRGB domain is complicated because of the non-linear ISP transformations. Despite this known issue, only a few methods work directly with RAW images and tackle real-world sensor degradations. We tackle blind image super-resolution in the RAW domain. We design a realistic degradation pipeline tailored specifically for training models with raw sensor data. Our approach considers sensor noise, defocus, exposure, and other common issues. Our BSRAW models trained with our pipeline can upscale real-scene RAW images and improve their quality. As part of this effort, we also present a new DSLM dataset and benchmark for this task.