Infrared and Visible Image Fusion Method Based on Entropy‐Weighted Optimisation and Image Enhancement

ABSTRACT To address the issues of severe loss of edge detail information and insufficient contrast in infrared–visible image fusion algorithms under low‐light conditions, resulting in poorly distinguished targets, this paper proposes an infrared–visible image fusion method based on entropy‐weighted optimisation and image enhancement. Firstly, a multiscale decomposition is performed using a rolling guided filter combined with a Gaussian filter, dividing the image into detail and base layers. Subsequently, the base layer undergoes weighted fusion based on background reconstruction techniques and visual saliency maps. For detail layer fusion, the finest detail layer is optimised through luminance feature enhancement combined with an entropy‐weighted strategy, while the remaining detail layers are fused using a bilateral filter smoothing mask and entropy‐driven adaptive weighting. Finally, a dual denoising enhancement algorithm combining non‐local mean denoising and wavelet denoising is employed to improve the contrast and visual quality of the fused image. To validate the effectiveness of the proposed method, extensive testing was conducted on the publicly available KAIST and TNO datasets. The results demonstrate that the proposed method significantly outperforms seven comparative approaches across objective metrics, including information entropy, joint entropy, edge strength, standard deviation, and spatial frequency. Furthermore, the fused images exhibit visually clearer target edges, heightened contrast, and enhanced detail in shadow regions.