Adaptive infrared and visible image fusion method by using rolling guidance filter and saliency detection

Abstract

Infrared and visible images have good complementarity; thus, infrared and visible image fusion methods are extensively used in military, target detection, pattern recognition, and other applications. However, problems such as halo effect, poor background texture, and contrast reduction are encountered in image fusion methods. In addition, when the visible image is disturbed by smoke and high brightness, it will affect the quality of the fused image, making image fusion a challenge. To address these problems, an adaptive image fusion method is proposed in this paper. First, the image decomposition method based on rolling guidance filter and saliency detection (RGFSD) is proposed. Next, the source image is decomposed into three layers by using RGFSD: detail layer, salient layer, and base layer. The detail layer is then fused using the phase congruency strategy. Subsequently, the local entropy and gradient (LEG) method is proposed to simulate the perception of significant information by the human visual system to assign the weights for salient layer fusion. Next, the base layer is fused using the averaging method. Furthermore, an image enhancement method based on rolling guidance filter and contrast-limited adaptive histogram equalization is proposed to enhance the detail and contrast of the fused image. Finally, the proposed method is compared qualitatively and quantitatively with eight fusion methods. The experiment results show that compared with the existing methods, the proposed method can better enhance the contrast and retain details of the source images.