Research And Application On Image Fusion Algorithm

With the rapid development of image processing technology and the wide applicationof various sensors, more and more image data are obtained from sensors of differentphysical characteristics,therefore, image fusion technology comes into being.Image fusionsynthetically makes use of complementary and surplus information of different sensors togenerate more complete, clearer and more reliable images. Image fusion can improve theutilization rate of image information and provide basis for visual observation and computerrecognition and processing. At present, image fusion is widely applied in medical science,information encryption, remote sensing and machine vision, therefore, image fusiontechnology is playing a significant role.Image fusion is based on pixel level, feature level and decision level. Pixel level is themost elementary level. It is the precondition of the fusion of feature level and decisionlevel, meanwhile can supply richer and more reliable detailed information than other levels.This paper is centered with pixel level. Main study is as follows:The basic concept, level structure and application status of image fusion areintroduced. The basic principle and method are stated. The key technology, imageregistration is introduced. The evaluation method of image fusion effect is studied andsummed up.Typical image fusion method is studied. And emphasis is put on the image methodbased on IHS.Aiming at heavy spectral loss, an improved method is proposed in this paper.Boundary extraction of high resolution images is according to the characteristic of eyessensitive to boundary. For critical point and non-critical point, different fusion strategiesare taken to reflect the contribution of two images. As for critical point, marginalinformation is reserved as much as possible; while as for non-critical point, multi spectralinformation is reserved as much as possible. The experiment result show that the methodproposed in this paper can improve the boundary definition and better reserve spectralcharacteristics of multi spectral images.Pyramid fusion method is studied. Fusion quality of different pyramid methods areanalyzed and compared. The influence of decomposition on fusion quality is studied. Thefusion method based on wavelet transform is the key point of the study. Because thecorrelations of sensor images of different physical characteristics are different, highfrequency fusion strategy regarding correlation coefficient as threshold and based onstandard deviation weighted average is proposed in this paper. This strategy is compared with usual regional variance and regional energy fusion strategies at present. The abovestrategies are put in simulation experiments, and objective evaluation system is used toanalyze the results quantitatively to prove the efficiency of the strategy proposed in thispaper.Sum up the advantages and disadvantages of this paper, and look forward to the futuredevelopment of image fusion.