Deep learning improves image clarity for myopic vision
A Gradient-Regularized Deep Learning Method for Restoring Image in Myopic Vision
Computer Vision and Pattern Recognition
Summary
Myopic vision makes it hard to see fine details because of how the eye focuses. The authors developed a deep learning method that changes images so people with myopia see them more clearly without needing to adjust their eyes. They trained a neural network using a special model of how vision blurs images and tested it on different photo collections. Their method improves image sharpness and maintains important details better than standard techniques.
What this means in practice
- •For mobile app developers: Develop apps that enhance image clarity for myopic users viewing photos or scenes on fixed screens.
- •For camera system engineers: Integrate computational correction to improve image output perceived by users with myopic vision in fixed-focus devices.
Authors
Kebin Contreras, Luis Toscano-Palomino, Rafael Torres, Jorge Bacca
Abstract
Myopic vision degrades the quality of human vision due to spherical eye defocus aberrations that impair the perception of fine spatial details. To address this limitation, this work proposes a deep learning-based approach to computationally design transformed images that allow individuals with myopia in fixed distance to perceive scenes with enhanced clarity. The method trains a neural network to counteract defocus aberrations modeled using the second-order Zernike polynomial $Z_2^0$. The approach employs a U-Net architecture trained with a loss function combining reconstruction error minimization with spatial difference structural preservation through $\ell_1$ regularization. Training is carried out exclusively on the KITTI dataset, while testing is performed on DIV2K and ImageNet datasets, demonstrating generalization across diverse image distributions. Under myopic conditions modeled by $Z_2^0$, baseline performance shows visual degradation with PSNR values ranging from 19.47 to 27.42 dB across datasets and severity levels. The proposed gradient-regularized method achieves consistent improvements, delivering PSNR gains ranging from 1.48 to 7.34 dB and significantly lower perceptual compared to classical baselines.