Retinex convolutional neural network and low-light image enhancement method

Through the image decomposition and fusion module of the Retinex convolutional neural network, the low-light image quality problem is solved, the image details and color improvement is achieved, and it is suitable for image processing applications in multiple fields.

CN118967486APending Publication Date: 2024-11-15TIANJIN UNIV OF SCI & TECH +1
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202410939229.5
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-07-15
Publication Date
2024-11-15

AI Technical Summary

Technical Problem

Image quality is affected under low light conditions, resulting in insufficient brightness, loss of detail and increasing noise. The existing low-light image enhancement methods are complex in calculations and sensitive in parameter selection, making it difficult to deal with complex scenes.

Method used

Convolutional neural network is constructed based on Retinex theory and deep learning framework, including image decomposition module, reflectivity recovery module and illuminance adjustment module. The image is decomposed into reflectivity map and illuminance map through convolutional neural network, and fuses and adaptive adjustments. Light changes to improve image detail and color authenticity.

Benefits of technology

It effectively solves the problems of color distortion, excessive noise and insufficient detail recovery of low-light images, improves the natural brightness and clarity of the image, and is suitable for image processing, medical images and information quality detection of industrial products.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118967486A_ABST
    Figure CN118967486A_ABST
Patent Text Reader

Abstract

The invention provides a Retinex convolutional neural network and a low-light image enhancement method, and belongs to the technical field of image processing and image analysis. The Retinex theory is used as the background, a deep learning framework is used as the basis, a network structure is built by using a convolutional neural network, a low-light image enhancement network model and a training process are provided, and the problems of color distortion, detail loss, complex network structure and the like of the low-light image are solved. According to the method, the Retinex theory is combined with the convolutional neural network, and image features are extracted through convolution of reflectivity and illumination, so that the enhanced image is highlighted in details, natural in brightness and real in color. According to the method, the real information of the low-light image can be obtained, the details of the low-light image can be further optimized, the enhancement problems of noise, detail loss and the like of the low-light image are solved, and the method has good application potential and value in the fields of safety and monitoring, night photography, medical imaging, military affairs, aviation and the like.
Need to check novelty before this filing date? Find Prior Art