LOW LIGHT IMAGE ENHANCEMENT USING DYNAMIC HETEROGENEOUS AND 3D-EMSM RGB COLOR REFLECTION METHODS

Abstract

     The traditional Retinex method is being utilized for image enhancement in order to improve the visual quality of low light video images, uncover hidden information in photos, and address the issue of low expressiveness brought on by color distortion and low saturation. In order to approximate the visual system perception intensity to physical reflectance, the input image is divided into two parts: the illumination component and the reflection component. Based on Retinex theory, a logarithmic transformation and Gaussian filtering are applied to the illumination component of HSV Color space. The preliminarily improved low light image of deep learning algorithms based on Retinex theory is obtained by adjusting the multi-scale reflection components of the input image using the estimated illumination value of the scene. The corruptions that the light-up process conceals are thought to be represented by the Retinex model. To create the final greyscale photographs, a color correction factor is applied to the first improved image. The brightness of nighttime and low light images, as well as color distortion, can be successfully improved with the first suggested Retined histogram equalirism technique. The second project is to use integral imaging from a point light source to create an efficient three-dimensional (3-D) display. This display is made up of a transparent liquid crystal display (LCD) panel without a back polarizer, a polarizer array, and an edge-lit light guiding plate. In addition to reconstructing the three-dimensional image, the light guide plate and polarizer array work together as a transparent Dynamic Heterogenous Method (DHA). The study acquisition describes how, as compared to typical picture enhancement techniques, the suggested solutions perform better in terms of both objective and subjective indications when handling low light RGB colorRetinex displays. Determine the night image's angular deviation as well. This study describes the proposed efforts, which include 3D effective multi-scale modules (3D-EMSMs) and RGB Retinex Histogram Equalirism, to effectively exploit angular correlations approach for deep learning low light picture enhancement system.