Research on IR-RGB Image Fusion Method Based on Target Perception and Hue Correction

In this paper, a hue-corrected IR-RGB fusion method is proposed based on object perception. We use the salient object detection network to detect the original IR image to obtain the salient regions in the image and then use the object bounding box in the dataset label to remove the noise information and retain the object area of interest. In order to implement different fusion schemes for the target area and the background area in the model, this paper trains a generator to synthesize the target fusion image, trains the target discriminator to identify the pixel intensity and hue information of the target area.And trains the background discriminator to identify Gradient information for the background. To improve the performance of downstream object detection tasks, an object detection network is added to the generative model, and an object detection loss is used to guide image fusion.