Research On Fusion Of Infrared And Visible Images

Fusion of infrared and visible images is an important branch of image fusion area. Many domestic and abroad research institutes have paid great attention to it for its application value in the military field and as a result, it developed rapidly. Complementarity exists between infrared images and visible images due to the different imaging principle of the infrared sensor and the visible sensor. As infrared imaging mainly depends on the emissivity and temperature distribution of the object, the gray values of the infrared image fluctuate according to the emissivity of the object. It can overcome some of the obstacles to discover the target and can work day and night. However, infrared images have low spatial resolution, so the visual effect of infrared images is usually poorer than that of visible images. On the opposite side, the imaging capability of visible sensors is poor on bad weather conditions, especially in the night. So the fusion of infrared and visible images can help us to take advantage of complementary information of both. It can help us obtain more detailed and accurate information from the scene, so that we can more accurately determine the location of the thermal target and achieve the identification of camouflage, night vision or some other purposes. It has high application value in military operations, electronic product testing, resource exploration and many other areas.The main purpose of this research project is to find a suitable fusion mothod for infrared and visible images in order to take full advantage of the complementary information between infrared and visible image . With the method, the fused image should maintain outstanding thermal target and clear background. Base on theories of image fusion , we developed a novel method for fusion of the infrared and the visible images on the basis of the complementary information. The algorithm mainly includes three parts:①The extraction of thermal target. Brightness of the thermal target area in infrared image is usually higher than the background area, according to this we proposed a novel algorithm for infrared image segmentation which integrated edge detection and region growing. Firstly, the edge of infrared image is detected by using a simple edge detection algorithm, and then seed pixel region growing is performed based on the connectivity and the results of edge detection. The segmentation is the basis of the first fusion. ②The first fusion of target image and visible image. At present, most researches on image fusion have focused on the pixel level, and ignored the natural characteristics in different different regions which result in poor fusion Performance. So we proposed a novel fusion method which is between pixel-level fusion and feature-level fusion to fuse the thermal target with the visible image .③The second fusion of the result of the first fusion and the two source images. In order to fully join the original information of the two source images and avoid loss of important information, we fused the two source images and the fused image of the first fusion with the method which is based on lifting wavelet decomposition. We first decomposed three original images to get their low frequency component and high frequency component. Then, we fused low frequency component and high frequency component using different fusion rules. At last, we used lifting wavelet reconstruction to get the fused image.Experimental results show that it is an efficient algorithm to fuse infrared and visible images. With the proposed method, the fused image maintains the unambiguous location of the target which is represented in the thermal image as well as the high resolution of the visual image, and the intelligibility of the images was improved. The method has strong robustness.