Human visual system deep learning enhancement algorithm based on optimization of physical information similarity

Hangying Zhang; Liangcai Cao; Feng Yang

doi:10.1117/12.2597933

18 June 2021 Human visual system deep learning enhancement algorithm based on optimization of physical information similarity

Hangying Zhang, Liangcai Cao, Feng Yang

Author Affiliations +

Proceedings Volume 11850, First Optics Frontier Conference; 1185002 (2021) https://doi.org/10.1117/12.2597933
Event: First Optics Frontier Conference, 2021, Hangzhou, China

Abstract

The performance of the imaging system under low light intensity will be affected by shot noise, and the shot noise will become stronger as the power of the light source decreases. Aiming at the impact of shot noise, this paper applies the principle of deep learning to low-light image enhancement. To improve the generalization ability of deep neural networks in different scenarios, a block matching solution based on BM3D is proposed to optimize the data of the Retinex network model. In the training process of the network, the consistency of the reflection component and the smoothness of the illumination component of the low-light image and the normallight image are used to constrain, without the need for real data of the reflection component and the illumination component. Experimental results show that this method can obtain a satisfactory low-light enhancement effect, and can significantly improve the reconstruction results of low-light images affected by noise.

Citation Download Citation

Hangying Zhang, Liangcai Cao, and Feng Yang "Human visual system deep learning enhancement algorithm based on optimization of physical information similarity", Proc. SPIE 11850, First Optics Frontier Conference, 1185002 (18 June 2021); https://doi.org/10.1117/12.2597933

ACCESS THE FULL ARTICLE

INSTITUTIONAL
Select your institution to access the SPIE Digital Library.

SELECT YOUR INSTITUTION

PERSONAL
Sign in with your SPIE account to access your personal subscriptions or to use specific features such as save to my library, sign up for alerts, save searches, etc.

PERSONAL SIGN IN

No SPIE Account? Create one

PURCHASE THIS CONTENT

SUBSCRIBE TO DIGITAL LIBRARY

50 downloads per 1-year subscription

Members: $195

Non-members: $335 ADD TO CART

25 downloads per 1 - year subscription

Members: $145

Non-members: $250 ADD TO CART

PURCHASE SINGLE ARTICLE

Includes PDF, HTML & Video, when available