Infrared and Visible Image Fusion Based on Image Enhancement and Target Extraction

• Published in: IEEE access Vol. 13; pp. 61862 - 61875
• Main Authors: Zhu, Haoran, Zhang, Wenying
• Format: Journal Article
• Language: English
• Published: Piscataway IEEE 2025; The Institute of Electrical and Electronics Engineers, Inc. (IEEE)
• Subjects: Accuracy; background subtraction; Brightness; Compression ratio; Computer vision; Data mining; Decomposition; Feature extraction; High frequency; Image coding; Image compression; Image enhancement; Image fusion; Image reconstruction; Infrared imagery; Sparse approximation; target extraction
• ISSN: 2169-3536; 2169-3536
• DOI: 10.1109/ACCESS.2025.3557799

In order to improve the detail visibility and target contrast of fused images, an infrared and visible image fusion method based on image enhancement and target extraction is proposed. The proposed method is divided into three parts, including enhancement stage, target extraction stage, and fusion stage. In the enhancement stage, to efficiently decompose the visible images, a decomposition method based on improved guided filter is proposed by using the deep feature as the guided image. Based on the characteristics of low and high frequency layers, brightness correction function and detail adjustment function are designed to improve the global and local contrast, respectively. In target extraction, morphological operation and background subtraction are introduced to achieve coarse target extraction efficiently. The feature distribution of redundant background is optimized to obtain more accurate infrared targets. In the fusion stage, the infrared target is injected into the enhanced visible image by compression ratio. A fused image with clear local details and high contrast thermal target is obtained while the exposure is suppressed. In order to prove the effectiveness of the proposed method, experiments are performed on different datasets. Moreover, the proposed method is compared with several mainstream methods based on six evaluation metrics. The results show that the fusion effect of the proposed method is better as compared to other methods.