Research On Multi-band Night Vision Imaging And Visible Image Fusion Recovery Technology

Infrared and visible imaging sensor has important applications in surveillance,detection,intelligence gathering and security.When infrared and visible sensors record the same scene,some of the information they provide is redundant or complementary.Image fusion technology combines visible and thermal infrared images into one image.The fused new image clearly shows the target and background,and the fused image provides more information than each individual source image.On the basis of analyzing the deficiencies of the existing image fusion method,several infrared and visible image fusion algorithms are researched in the dissertation.The main research work is as follows:The edge and texture preserving ability of fusion algorithm will be affected under the interference of noise.To solve this problem,an infrared and visible image fusion method based on iterative guided filtering and multi-vision weight information is proposed.Firstly,iterative guided filter with scale-aware and edge-preserving properties is applied to decompose the source images into base layer and a series of detail layers.The iteration process can not only better separate the spatially overlapping image feature information,but also effectively reduce noise and suppress edge artifacts.Then,in order to retain more important details of the source image,binary weight maps are synthetically determined by deep texture information and edge information.The detail layer is optimized by the weighted least square method to reduce noise.Finally,the fusion image is obtained by linear combination of the decomposition information.Experiments show that proposed method is superior to other methods in suppressing noise.In the dark environment,there are some hidden or blurred details in the fused image.To solve this problem,an infrared and visible image fusion method based on contrast enhancement and multi-scale edge-preserving decomposition is proposed.A multi-scale contrast adaptive enhancement algorithm is used to improve the visibility of night visible images before the fusion.The clarity and contrast of details can be effectively enhanced with this method.Then,fusion rules are designed according to the properties of different layers to preserve the restored information as much as possible.In order to retain more texture information of source images,the parameter-adaptive model is applied to fuse the small-scale detail layer.In order to avoid mutual interference of different characteristic images,a new edge detection method is used to fuse large-scale detail layer.In order to improve the contrast of fusion results,the weighted averaging technique based on visual saliency map is used to fuse the basic layer.Finally,a clear image is obtained by combining the fused layers.The experimental results show that the proposed method has better visual effect than other fusion methods based on contrast enhancement.It is superior to other methods in average gradient,spatial frequency,performance of edge preservation and total loss of information,leading by 11.86,12.98,0.04,0.05.The fusion method based on the multi-scale decomposition idea has high computational complexity and can not meet the real-time requirements.In order to solve the above problems,an infrared and visible image fusion method based on intensity transformation and saliency information extraction is proposed.Firstly,the visible image is transformed by Sigmoid and exponential function to highlight the background details.Secondly,infrared saliency information is obtained by reconstructing infrared background image.Thirdly,two suppression factors are used to optimize infrared salient information,reduce redundant information and prevent overexposure.Finally,the fusion image is obtained by combining the visible image of gray transformation with the infrared saliency information.Subjective and objective analysis shows that,compared with other fast fusion methods,the proposed method can retain more visible detail information and more salient thermal targets.It is superior to other methods in average pixel intensity,perceptual saliency and mutual information.In addition,the computation time of the proposed method is less than 0.03 seconds,which meet the real-time requirements.