Original scientific paper

https://doi.org/10.17559/TV-20250804002881

A Multi-Scale Fusion-Based Method for Ultra-Low-Light Agricultural Image Enhancement

Jun Li ; School of Information Technology, Xichang University, Key Laboratory of Liangshan Agriculture Digital Transformation of Sichuan Provincial Education Department, Xichang 615013, China *
Zhou Wei ; Xichang Cigarette Factory, China Tobacco Sichuan Industrial Co. Ltd., Xichang, China Xichang, Sichuan, 618400, China
Yining Song ; International Telecommunication Union, 1211 Geneva, Switzerland
Shiqiong Chen ; School of Information Technology, Xichang University, Key Laboratory of Liangshan Agriculture Digital Transformation of Sichuan Provincial Education Department, Xichang 615013, China
Miao Yu ; School of Information Technology, Xichang University, Key Laboratory of Liangshan Agriculture Digital Transformation of Sichuan Provincial Education Department, Xichang 615013, China
Wenfeng Wang ; School of Information Technology, Xichang University, Key Laboratory of Liangshan Agriculture Digital Transformation of Sichuan Provincial Education Department, Xichang 615013, China
Chenping Zeng ; School of Information Technology, Xichang University, Key Laboratory of Liangshan Agriculture Digital Transformation of Sichuan Provincial Education Department, Xichang 615013, China
Yanmeng Chen ; School of Information Technology, Xichang University, Key Laboratory of Liangshan Agriculture Digital Transformation of Sichuan Provincial Education Department, Xichang 615013, China

* Corresponding author.

Abstract

Fruits, vegetables, and medicinal herbs are often processed in enclosed spaces with low light. These conditions create uneven lighting, produce dark images, distort colors, and increase noise, all of which make remote monitoring difficult and reduce product quality. This paper uses tobacco leaf curing as an example to introduce a Multi-Scale Fusion-Based Image Enhancement (MSFIE) technique designed to address these problems. Histogram equalization, a CLAHE module, an enhanced dual-gamma correction, and a multi-scale fusion framework are all included in the technique. When combined, these elements improve overall contrast and sharpness, lessen local overexposure, and brighten dark areas. The primary and secondary leaf vein structures are further made clearer by a detail enhancement technique. Experiments on 150 ultra-low-light images demonstrate that MSFIE improves visual quality, brightness uniformity, detail preservation, and color accuracy. Quantitatively, it achieves consistent gains in Shannon entropy, peak signal-to-noise ratio, laplacian-based sharpness, local contrast, frequency-domain sharpness, luminance skewness, and information entropy.

Keywords

color distortion; detail enhancement; low-light image enhancement; MSFIE

Hrčak ID:

337743

URI

https://hrcak.srce.hr/337743

Publication date:

31.10.2025.

Visits: 104 *