An infrared image enhancement method based on Multi-Channel Feature Fusion Network

With the rise of computer vision, there is an urgent demand for high-quality infrared images in various fields, characterized by appropriate contrast, high brightness, and detailed texture. However, the challenge in acquiring infrared images of high quality lies in how to effectively improve contour and detail information while eliminating noise interference. Therefore, an infrared image enhancement method is proposed and based on Multi-Channel Feature Fusion Network (MCFFNet), which consists of three channels and two fusion modules. First, the contour enhancement channel extracts foreground information from the original infrared images to separate the contour from the background. Second, the detail enhancement channel is designed to extract intrinsic information from the input, enriching texture details. Third, the noise processing channel is utilized to restrain background noise and improve brightness and contrast. Finally, the enhanced infrared image is obtained through two fusion modules, which integrate the information obtained by the three channels. Extensive subjective and objective comparative experiments have demonstrated significant improvements in contrast, brightness, and texture details of the infrared images processed by this method. Compared to original image, standard deviation (STD) and average gradient(AG) produced by the proposed method are up to 53.4645 and 14.2594, increased by 37.27% and 105.42% respectively, which shows its efficiency for infrared image enhancement.