Prior Metadata-Driven RAW Reconstruction: Eliminating the Need for Per-Image Metadata

Wencheng Han, Chen Zhang, Yang Zhou, Wentao Liu, Chen Qian, Cheng-zhong Xu, Jianbing Shen

Abstract: While RAW images are efficient for image editing and perception tasks, their large size can strain camera storage and bandwidth. Techniques exist to reconstruct RAW images from sRGB data, but these methods typically require additional metadata from the RAW image, which adds to camera processing demands. To address this problem, we propose using Prior Meta as a reference to reconstruct the RAW data instead of relying on per-image metadata. Prior metadata is extracted offline from reference RAW images, which are usually part of the training dataset and have similar scenes and light conditions as the target image. With this prior metadata, the camera does not need to provide any extra processing other than the sRGB images, and our model can autonomously find the desired prior information. To achieve this, we design a three-step pipeline. First, we build a pixel searching network that can find the most similar pixels in the reference RAW images as prior information. Then, in the second step, we compress the large-scale reference images to about 0.02% of their original size to reduce the searching cost. Finally, in the last step, we develop a neural network reconstructor to reconstruct the high-fidelity RAW images. Our model achieves comparable, and even better, performance than RAW reconstruction methods based on metadata.

Primary Subject Area: [Experience] Art and Culture

Secondary Subject Area: [Experience] Art and Culture

Relevance To Conference: The proposed technique for reconstructing RAW images from sRGB data using Prior Meta is crucial for multimedia, offering a balance between image quality and storage efficiency. By removing the need for per-image metadata and significantly reducing file sizes, this approach enhances multimedia applications with better image quality and editing flexibility without straining storage and bandwidth. It's particularly valuable in digital photography, content creation, and streaming, where high fidelity and efficiency are key.

Submission Number: 489