Image Fusion Based On Focus Region Detection

Image fusion technology uses complementary information in multiple images to fuse multiple images of the same scene collected by various sensors into an image that can more fully express the scene information.The fused image is more conducive to human visual observation and machine learning.Multi-focus image fusion is an important branch of image fusion.It involves many fields such as electronic sensor science,signal processing,machine learning,etc.,and has been widely used in many fields such as digital image,military,and surveillance.This thesis focuses on the research of region-based multi-focus image fusion algorithm.Firstly,aiming at the fact that multi-focus image fusion methods in the spatial domain cannot accurately identify and extract the small focus-region,a multi-focus image fusion method based on the gradient-focus-measure and random walk is proposed.The proposed method can accurately identify and extract small focused-regions of the source images.And there is no boundary effect in the fused image,which improves the quality of the fused image.Secondly,for the case that the region-based fusion methods in the spatial domain are easy to misjudge the focus-degree of the smooth region in the source images,an image fusion method based on edge-focus-region detection(EFD)and guided filtering(GF)is proposed.The proposed method can avoid the influence of the smooth region in the focus-region,and can obtain a more accurate focus-region,so that more detailed information in the source image can be preserved in the fused image,which improves the quality of the fused image.At the same time,the proposed method can enhance the spatial continuity and improve the visual perception of the fused image.Finally,the traditional image fusion methods based on non-subsampled Shearlet Transform(NSST)domain lose a lot of detail information of the source images during the fusion process.Then a sub-band coefficients fusion rule based on focus-region detection in the NSST domain is proposed to improve this situation.Compared with traditional image fusion methods,the proposed method can make full use of the correlation between the sub-band images in different levels and directions and consider the image from a global perspective.So that the fused image retains the valuable original information in the source images to the greatest extent,and improves the mutual information of the image and the fault tolerance of the fusion methods.At the same time,in order to further reduce the information loss in the fusion process of the fusion methods in transform domain,an image fusion method based on NSST and focus-region detection is proposed,which combines the method in transform domain with the method in spatial domain.The proposed method combines the advantages of the methods in spatial domain and transform domain,so that the fused image contains more focus information in the source images and has good spatial structure characteristics.