Image Fusion Algorithms Based On Multi-Scale Geometric Analysis

The image fusion is to combine several images with the same scene to generate an image which contains all of the source images' important features.The fused image is more suitable for human and machine to understand,and is more convenient to complete the image processing tasks,in the meantime to obtain a more effective image fusion algorithm is of great application value and research significance.This paper proposed the image fusion algorithm for different images on the basis of the study of the existing literature.The main research work and innovations are as follows:(1)The beginning of this article introduces the background and significance of image fusion,the existing issues and current situation of image fusion in domestic and foreign.Afterwards,detailed expounds the imaging characteristics of integrating source images in proposed algorithm.Finally analyzes the proposed fusion algorithm in spatial domain and transform domain level.(2)Emphatically introducing the multi-scale decomposition tool Contoutlet transformation theory and decomposition effect,at last from principle and structure expounding the shortcoming under the Countourlet transform in comparison with nonsubsampled Contourlet in multi-scale image processing and multiple directions.(3)In the light of the characteristics of CT and MRI medical images,this paper proposes a self-adaptive medical image fusion algorithm on the basis of combining NSCT and PCNN.Firstly Using the nonsubsampled transformation to decompose the strictly registered images to be fused for multi-directional and multi-scale,to obtain low frequency sub-band coefficients and different directions of the high frequency sub-band coefficients.The selected value based on the edge of coefficient of the low frequency sub-band and high frequency sub-band coefficients are regarded as external excitation imput for PCNN.To fuse low-frequency sub-band refer to the first ignition timing of the corresponding point.(4)According to the multi-focus image characteristic and human visual characteristics,this paper proposed a multi-focus image algorithm based on combination of NSCT and ICM.obtaining the high frequency sub-band coefficients and low frequency sub-band coefficients from different directions by using NSCT to decompose source images,the low frequency sub-band focus the energy and background information of the source images,and at the same time keep the non-focused part as much as possible.The Laplace energy Which calculate the improvement of low frequency sub-band coefficient is regarded as the external motivation of input of ICM.Comprehensively select low frequency coefficients refer to the corresponding points' first ignition time and the Sum Modified Laplacian.The high frequency sub-band mainly reflects the edge and contrast of the source images.The spatial frequency of high frequency sub-band coefficients of improvement is regarded as the external motivation of input of ICM.Finally to select the high frequency sub-band coefficients based on the corresponding points' first ignition time and modified spatial frequency.This paper carries out the corresponding simulation and experiments to both two fusion algorithms,and comparison with other fusion algorithms at the same time.The results showed the obtained image by using proposed fusion algorithm have a good visual effect and more detailed information,and carrying on the analysis comparison under the objective evaluation index for different fusion algorithms,that demonstrates the proposed fusion algorithm has better ability to retain the image edge and texture information,the important conclusion got from the analysis has important practical guiding significance for the subsequent image processing work.