A New Concatenated Method for Deep Curve Estimation Using Low Weight CNN for Low Light Image Enhancement

Abstract

This paper presents a new concatenated method for deep curve estimation. Which is a modified version of Zero-DCE. Which produces an enhanced image from a low light image. With the aid of a deep network, this may be accomplished as an image-specific curve estimate task. Our technique develops a light-weight deep network that can forecast pixel wise and high order curves for an image’s dynamic range correction. Unlike majority of GAN based and CNN based methods, which requires paired images for training it can be trained without paired/ reference images. The potential advantages of our technique for reading license plates of vehicles at night are also discussed.