Research On The Technology And Application Of Image Fusion Based On Improved NSCT

Image fusion use the redundant information can effectively improve image's accuracy and reliability,in addition,integration and complementary information make the images a better description.Therefore,introduce the image fusion technology into the image acquisition,observation system,can effectively reduce the image distortion and anamorphose which caused by the limitation from the hardware of high-precision equipment.and the impact from environment.However,the existing multi-resolution image fusion method unable to express the singularity of a large number of curves and surfaces in two-dimensional images excellently,and ignored the dependence relation between decomposition layer's coefficients,result in the appearance of blurred contours even artifact shadows in fused images.To solve these problems,this paper explored the multi-directional and multi-scale NSCT transformation,and fully considered the advantages of the NSCT transformation and the statistical correlation between sub-band's coefficients,proposed an improved NSCT transformation image fusion algorithm.The content of this paper mainly focus on the following aspects:(1)The pretreatment methods of image denoising and image registration is referred,and the quality evaluation index of image fusion is analysed systematically.Specifically,a local adaptive block wavelet algorithm is used for image denoising,which can effectively suppress Gaussian noise and retain the important features of original image.Furthermore,to ensure the accuracy of the fused image,the approach based on SURF is exploited for precise image registration.(2)Consider the properties of NSCT,multi-scale and multi-directional as well as translation invariance,an adaptive image fusion algorithm based on NSCT was proposed.It adopts compressed sensing technology to compress and measurement NSCT sub-band coefficients,and introduces the Gaussian fuzzy membership function to create the adaptive fusion rule.The algorithm can not only reduce the pressure of equipment storage space,but also highlight the important features and details in images.(3)In view of the existing fusion algorithm based on NSCT ignores the interdependence between the sub-bands of multi-scale space and the sub-bands of the same scale space,an improved image fusion algorithm based on NSCT was proposed.It employs the generalized Gaussian density function(GGD)to fit the edge distribution curves of the high-frequency coefficients of NSCT,and uses the JSD(Jensen-Shannon divergence)to measure the similarity of the sub-bands.Since the proposed algorithm is based on the statistical correlation between NSCT sub-bands,more important information of source images can be preserved,which can enhance the robustness of the fusion algorithm.(4)The image fusion technology was applied to the measurement system of high-precision instruments in this paper.In order to get more accurate sample morphology,the adaptive image fusion algorithm based on NSCT and the improved image fusion algorithm based on NSCT were used to correct the error of images captured by scanning electron microscope respectively.Experiments show that the image fusion methods in this paper have good results in error correction for nanoscale scanned images,and the improved image fusion algorithm based on NSCT performs better.