Research On Image Fusion Algorithm Based On Modified Shearlet

Image fusion refers to fuse images from two or more sensors of the same scene together,so that the fused image is more abundant with information and contains more redundant information than the source images,more consistent with human visual characteristics,and be easier to image processing and analysis further.With the development of digital image processing technology,image fusion technology based on multiscale analysis theory has become the focus of the research,and has been widely used in military,medical,remote sensing and other fields.Shearlet transform,a multiscale analysis method,is more popular in recent years,not only has the simple local mathematical structure,good directional sensitivity,good time-frequency characteristics,low computational complexity,but also a multidimensional function near optimal sparse representation tool,which favored by many researchers.However,due to the subsampled operation during shearlet discretization,shearlet does not have translation invariance,and it is easier to generate pseudo phenomenon near the singular points when it is applied to image denoising and image fusion.In order to overcome the shortcomings of shearlet,the improved shearlet transform has been proposed,such as the non-subsampled shearlet transform(NSST)and the fast finite shearlet transform(FFST).With the help of the theory of the improved shearlet transform,according to the characteristics of different types of images,mainly to improve the quality of the image edges,keep clear,easy to detect,identify,analyze and deal with the target information for the purpose of researching image fusion method.Any research can not be carried out without the support of theory.The paper mainly to improve the quality of the image edges,keep clear,easy to detect,identify,analyze and deal with the target information for the purpose of image fusion method research in the light of the characteristics of polarization and multi-focus image with the improved shearlet transform and guide filtering theory.The improved shearlet transform are proposed based on shearlet transform,which overcome the shortcomings of the shearlet transform,has better directionality,time-frequency characteristics,and the unique ability in representing image details.The specific shearlet transform process is described detailedly in the following chapters.Compared with the traditional filters,guided filter can enhance image details,preserves the overall characteristics of the input image while preserving image edge gradient,which is widely used in image denoising,image defogging and image fusion.This paper applies it to image fusion,mainly to keep the edge of image clear while enhancing the detail information of image.In addition,the selection of the application of several other popular multiscale geometry analysis methods(wavelet,curvelet,contourlet,etc.)in image fusion,this paper carried out comparative experiments to analyze the performance,which has some guiding significance for the research of the following image fusion algorithms.According to the characteristics of polarization images,a fusion algorithm about polarization image based on region energy and NSST was proposed in this paper.In this algorithm,firstly the source images were decomposed into low frequency and high frequency coefficients by using NSST;low frequency coefficients of the decomposed were fused based on the fusion rule of region distance energy,the high-frequency coefficients were fused based on the regional energy fusion rule,and then the detail enhancement was performed by using the guided filter to highlight the detail features of the image;finally,the fused image was obtained by inverse NSST transform.The algorithm was characterized by introducing the guided filter to enhance the high frequency coefficients of the source images,preserving the edges of the image while enriching the details of the image and improving the clarity of the image,so as to be more in line with human visual characteristics.According to the feature of multi-focus images,this paper proposed a multi-focus image fusion algorithm based on guided filter and FFST transform.Before fusion,the source images were decomposed into low frequency coefficients and high frequency coefficients by FFST.In the fusion of the low frequency coefficient,defined a new improved Laplace energy(NSML),and designed a selection scheme for the fusion of low frequency coefficient based on region NSML.High frequency coefficients were rich in detail information,this paper put forward a region energy weighted fusion algorithm based on guided filter.Finally,the final fused image obtained by inverse FFST.According to the subjective visual effect,the fused image not only has prominent details,but also has a clear edge;the objective evaluation indicators have improved by a big margin compared to other algorithms.