Most super-resolution and image enhancement methods rely on large, paired training datasets and predefined distortion models, limiting their effectiveness on real-world images acquired under unknown, variable, or suboptimal conditions. This technology, termed zero-shot super-resolution (ZSSR), introduces an unsupervised approach that trains a convolutional neural network directly at test time using internal patterns extracted from the input image itself. Leveraging cross-scale recurrence within a single image enables adaptive, image-specific enhancement of low-resolution, noisy, or distorted data without prior training or external information, consistently outperforming existing methods in non-ideal scenarios.Optical and dimensional limitations of images can be partially overcome today using signal enhancement technologies which typically rely on supervised, deep-learning methods. However, as existing methods are restricted to specific training images and distortion types, they provide poor results for any practical case, unless huge set of data-pairs exist and all share the same exact distortion type. The current invention enables signal enhancement of low-resolution images, by exploiting deviations from expected internal patch recurrences detected within the image itself. More specifically, the approach leverages cross-scale internal repetition of image-specific information, which is trained, at test time, on internal examples extracted solely from the test image. This first unsupervised conventional neural network (CNN)-based super-resolution approach allows for signal enhancement of real-world images acquired under suboptimal, unknown or image-specific conditions and has been shown to outperform state-of-the-art technologies.
- Enhances low-resolution images acquired under unknown or mixed distortion conditions, improving usability without retraining
- Restores quality in noisy, blurred, or artifact-heavy images, including legacy and degraded visual data
- Improves medical and biological imaging outputs (e.g., fMRI) where acquisition conditions are constrained
- Enables enhancement of video and audio sequences using internal data redundancy within signals
- Zero-shot approach requiring no external datasets or pre-training
- Image-specific learning from internal cross-scale patch recurrence
- Robust to unknown and non-ideal imaging conditions
- Unified framework addressing multiple distortions (super-resolution, denoising, deblurring, dehazing)
A) Super-resolution of a suboptimal low-resolution image using the ZSSR approach versus state-of-the-art (SotA) approaches.
B) Super-resolution using ZSSR as compared to state-of-the-art methods, of a low-resolution image generated under ideal, supervised conditions.
The technology has been experimentally validated across multiple image enhancement tasks, including super-resolution, dehazing, and artifact removal, demonstrating performance advantages over state-of-the-art supervised methods in suboptimal conditions.
