Low light combining multiscale deep learning networks and image enhancement algorithm

Journal Title: Modern Innovations, Systems and Technologies - Year 2022, Vol 2, Issue 4

Abstract

Aiming at the lack of reference images for low-light enhancement tasks and the problems of color distortion, texture loss, blurred details, and difficulty in obtaining ground-truth images in existing algorithms, this paper proposes a multi-scale weighted feature low-light based on Retinex theory and attention mechanism. An image enhancement algorithm is proposed. The algorithm performs multi-scale feature extraction on low-light images through the feature extraction module based on the Unet architecture, generates a high-dimensional multi-scale feature map, and establishes an attention mechanism module to highlight the feature information of different scales that are beneficial to the enhanced image, and obtain a weighted image. High-dimensional feature map, the final reflection estimation module uses Retinex theory to build a network model, and generates the final enhanced image through the high-dimensional feature map. An end-to-end network architecture is designed and a set of self-regular loss functions are used to constrain the network model, which gets rid of the constraints of reference images and realizes unsupervised learning. The final experimental results show that the algorithm in this paper maintains high image details and textures while enhancing the contrast and clarity of the image, has good visual effects, can effectively enhance low-light images, and greatly improves the visual quality. Compared with other enhanced algorithms, the objective indicators PSNR and SSIM have been improved.

Authors and Affiliations

Xia Yu, Lin Bo, Chen Xin

Keywords

Related Articles

Model for assessing the investment attractiveness of regions

In this paper, a comprehensive model for evaluating the investment attractiveness of regions is presented. This model incorporates the influence of key indicators of investment attractiveness on investment risk, enabling...

Modeling the water cycle in a bio-regenerative life support system

The work is devoted to modeling the water cycle in the bio-regenerative life support system. The scheme of closed hydrological cycle, including human and plant links, modules of food preparation, waste treatment, water a...

Performance comparison of PostgreSQL and its TimescaleDB extension

This article considers the PostgreSQL database and its extension TimescaleDB. The purpose of this article is to compare the performance of databases when dealing with big data. Since nowadays, when developing application...

Calculation models for the assessment of deflected mode in the surface layer of parts during surface plastic deformation by running and smoothing

The choice of a calculation model for the assessment of the stress state during finishing-hardening processing, in particular, during shot-impact processing are substantiated in the work. Analytical dependencies for calc...

Modeling of fuzzy logic for control of the process of bacterial oxidation of concentrates in reactors with a mixer

The multiplicity of chemical, biological, technological parameters and the nonlinear dependence in the extraction of metals in bacterial oxidation media complicate the mathematical modeling of these systems. This study w...

Download PDF file
  • EP ID EP711099
  • DOI 10.47813/2782-2818-2022-2-4-0215-0232
  • Views 104
  • Downloads 0

How To Cite

Xia Yu, Lin Bo, Chen Xin (2022). Low light combining multiscale deep learning networks and image enhancement algorithm. Modern Innovations, Systems and Technologies, 2(4), -. https://europub.co.uk/articles/-A-711099