A dual-stream fusion network for contrast enhancement of infrared images in complex scenarios

This paper proposes a contrast enhancement method for infrared images in complex scenarios, addressing the limitations of traditional deep learning methods in high-bit infrared image enhancement. We design a Two-Stream Enhancement Fusion Module (TSEFM) that effectively adjusts dynamic range and preserves image details by independently processing the detail and base layers of images. Unlike existing 8-bit image enhancement algorithms, our method is specifically designed for 12-16 bit infrared images, employing one-dimensional histogram adjustment and attention mechanisms to balance contrast enhancement and detail preservation. Experiments on multiple public datasets validate the superiority of this method in visual effects and quantitative metrics, providing a new technical solution for infrared image enhancement.