Abstract
Storing high-resolution digital photos and video requires substantial storage space on a user device and/or a cloud-based image library. This disclosure describes the use of generative artificial intelligence to achieve storage savings via image downscaling with little or no perceptible impact. With user permission, a customized model such as a LoRA patch is trained based on images in the user’s library. Images in the library are stored as downscaled versions. When the user requires a high-resolution version, a generative artificial intelligence model, optionally with the customized LoRA patch, is used to generate the target resolution version from the downscaled image. Users can choose to implement the techniques for their entire library and/or specific images, and can select the target compression that achieves adequate upscaling quality. The described techniques can enable users to store more images on-device and in cloud storage without the added expense to purchase storage capacity.
Creative Commons License
This work is licensed under a Creative Commons Attribution 4.0 License.
Recommended Citation
Kamali, Amir; Talebi, Hossein; Panpalia, Prachi; and Torres, Nick, "Efficient Photo Compression and Reconstruction Using Generative Artificial Intelligence", Technical Disclosure Commons, (August 20, 2025)
https://www.tdcommons.org/dpubs_series/8493