Sparse Gradient Regularized Deep Retinex Network for Robust Low-Light Image Enhancement

Main Article Content

Resham Taluja

Abstract

An algorithm for improving images was created in this procedure. It is founded on the linear domain simultaneous estimate of light and reflectance. For single picture low-light enhancement, many image priors have been used. However, the most effective technique is to evenly enhance the illumination directly. By extending the dynamic pixel intensity of a picture, histogram equalisation (HE) can solve the issue of lighting and make dark images visible. Instead than changing the lighting, HE seeks to improve the contrast. As a result, the findings are resilient to noise and improve image quality. The test findings demonstrate the good performance of the suggested approach to provide lighting and reflectance with increased visual outcomes and a promising convergence rate. The suggested technique produces equivalent or superior outcomes in subjective as well as objective evaluations when compared to previous testing methodologies. Image enhancement’s main goal is to treat an input image such that the final output is more suitable for a certain application than the original image. It draws attention to or emphasises visual elements like borders, limits, or contrast to make a graphic presentation more useful for study and display. The improvement enhances the dynamic range of the selected characteristics, making it easier to recognise them even while it does not increase the data's intrinsic information richness.

Downloads

Download data is not yet available.

Metrics

Metrics Loading ...

Article Details

How to Cite
Taluja, R. . (2019). Sparse Gradient Regularized Deep Retinex Network for Robust Low-Light Image Enhancement. Turkish Journal of Computer and Mathematics Education (TURCOMAT), 10(2), 995–1002. https://doi.org/10.17762/turcomat.v10i2.13581
Section
Research Articles