Research On The Multi-source Image Fusion Algorithm

With the rapid development of the modern science and technology,the single sensor cannot have satisfied the needs of people.Through the analysis of the performance of different sensors,the combination of a variety of image data with different features,image processing technology arises at the historic moment.Image fusion is affiliated with an important branch of image processing technology,will be two or more integrated sensors to obtain a scene image,the purpose is to generate images with clear and rich information.The development of multi-source image fusion technology enlarged the application range of the image data,makes the image of the original information is highly concentrated,convenient for storage and transmission.For the research of image fusion algorithm,the core problem is how to get an efficient representation.Since the 1980s,multi-source image fusion become the important subject in the field of image processing researchers worldwide,and is widely used in military,computer vision,medical and remote sensing etc.This paper introduces the concept of multi-source image fusion,processing and evaluation.Multi-scale transform is a hot research topic in recent years,image fusion,it for image segmentation in different scales and different resolution,so as to improve the contrast of the image.Consider multi-source images of different imaging mechanism,from the perspective of multi-scale tool to research,the main results of this paper are as follows:(1)Aiming at the limitation of single pixel fusion rules,and the shortcomings of the edge profile is not clear.To solve this problem,a novel image fusion algorithm is proposed based on nonsubsampled contourlet transform(NSCT)and regional features.Based on the characteristics of high and low frequency domain,the fusion rule based on regional average energy and the average energy matching degree is used to fuse the low frequency subband components.The high frequency subband components by using improved sum of Laplace energy.The experimental results show that the algorithms in this paper compared with the traditional algorithms at different area of the window and the number,visual subjective and objective evaluation of the results are superior to the traditional fusion,fusion efficiency is improved.(2)The disadvantage of external excitation and link strength in traditional PCNN model,consider the human visual sensitivity to the edge of the image detail and information on the parameters of the adaptive improvement.The support value of image classification as an external stimulus,improved sum of Laplace energy intensity as a link strength,fully retained edge information.The algorithm combines the advantages of NSCT and PCNN models.The experimental results show that the improved algorithm fusion better than traditional methods,and more efficient.(3)For the integration of remote sensing spectral image distortion problem,a new algorithm combines improved IHS and NSCT transformation.The IHS transformation of the multispectral images is conducted to extract the I component.Then it is enhanced to gain a new component I of multispectral images.The panchromatic images and new I component are decomposed by NSCT.For low frequency component of NSCT using fuzzy logic function weighted rules,the high frequency component with multiple decision fusion rules,finally fused image is obtained by inverse transformation reconstruction.The fusion results show that the proposed algorithm can ensure the spectral information enrich,and effectively improve the spatial resolution.