Study On Fusion Algorithm Of Visible And Infrared Image

Due to the different imaging principles,optical and infrared sensors have different descriptions of the same scene.Image fusion can integrated the complementary information of visible and infrared images to create a new image whose information is more comprehensive and complete.Existing algorithms of fusion often suffer from loss of texture details,incomplete retention of significant targets and contours,etc.,which are not conducive to the perception and understanding of scene.In view of the above problems,three new infrared and visible image fusion algorithms are proposed based on the analysis of source images characteristics.The main contents and results are as follows(1)In order to fully retain the information,an adaptive weighted average fusion method for visible and infrared images based on NSCT is proposed.Firstly,the source images are decomposed into high and low frequency components of different scales and directions by NSCT.Secondly,the weight map of low-frequency is adaptively constructed by Gaussian fuzzy logic,and the high-frequency sub-band is fused by max-absolute.Experiments show that the proposed algorithm can better preserve the salient features and improve the contrast of the ffused image.(2)In order to avoid the"ringing"effect,while fully retaining the target and texture details of the source images,this paper proposes a new two-scale fusion algorithm based on saliency detection.Firstly,the source images are decomposed by bilateral filtering to obtain the base layer and the detail layer.Secondly,the weight map of base layer is constructed by the saliency detection,and the detail layer is processed by max-absolute.The results show that the proposed algorithm can overcome the"ringing"effect of the fused image and improve the quality of the fused image.(3)So as to make the retention of target and contour of the images more completely and the improve the image consistency,this paper proposes an image fusion algorithm based on super-pixel segmentation and maximum variance.The entropy rate super-pixel segmentation is performed on the base layer,and the weight map is constructed by global saliency detection,and the max-regional variance fusion strategy is designed for the detail layer.Experiments indicate that the proposed algorithm can make the targets and contour more complete,improve the image consistency,and be more in line with the visual characteristics of the human eye.