Lightweight Infrared and Visible Image Fusion Technique: Guided Gradient Optimization Driven

Infrared and visible image fusion technology aims to combine data from several spectral bands in order to increase target identification, processing capabilities, and image quality. With the rapid development of consumer electronic products for imaging, there is an urgent need for a lightweight and efficient fusion technology that ensures efficient information extraction and fusion while maintaining image quality. Existing algorithms designed to achieve accurate information extraction, noise reduction, artefact suppression, and edge preservation need to be simplified and more challenging to meet the requirements of lightweight imaging consumer electronic products. We propose a lightweight method for the fusion of infrared and visible images by exploiting the properties of the Anisotropic Guided Filter and the Gradientlet Filter. This method achieves significant feature texture extraction, effectively reduces gradient texture and noise, minimizes halo artifacts, and enhances edge contours while preserving overall image brightness and edge gradients. Furthermore, the explicit stage processing and concise algorithmic structure design of this method contribute to its optimal time efficiency. Experimental results demonstrate its superiority in both subjective visual effects and objective metrics over nine other existing image fusion methods.

I. Introduction

Extracting important feature data from different spectral bands and creating a comprehensive fused image is the main objective of image fusion. This technology is widely applied across a spectrum of fields, including biomedicine, modern agriculture, digital photography, environmental monitoring, military reconnaissance, and meteorological forecasting, among others [1], [2]. In particular, infrared and visible images play a crucial role as fusion data sources, leading to technological breakthroughs and innovative applications in these domains. The application of the fusion technology of the infrared and visible pictures in the area of security and surveillance is illustrated in Fig. 1. It can be observed that the information captured by the (a) and (b) exhibits significant differences. The fusion result (C) is capable of combining the information from both sources, providing a more enhanced visual effect. The aim of fusing the infrared and the visible images to achieve mutual complementarity and maximise the use of the information available. Infrared sensors transform objects’ thermal properties into high-contrast grayscale images, enhancing target visibility against diverse backgrounds. However, these images can suffer from poor resolution and noise. By contrast, images captured with visible light offer high resolution and detail, but are sensitive to environmental conditions. Thus, by merging the advantages of both modalities—infrared and visible light—a more thorough and accurate depiction of the scene can be achieved. Fig. 1.

(a), (b) and (c) show a pair of infrared and visible images and the fusion result.