Medical Image Fusion Algorithms Based On Multi-scale Geometric Analysis

With the fast development of sensor technology, numerous different modality medicalimages are available. however because of the different imaging mechanism of the sensors, thesingle-modal medical images only can provide the local detailed information. In order toovercome the limitations, multi-modal medical image fusion technology is proposed byacademics. The proposed multi-modal medical image fusion technology can obtain the picturewhich can presents human organs and lesions more accurate, more clear, more comprehensivethrough extracting and integrating information from different modal medical images. Thus it canprovide a reliable basis for doctors to diagnose disease and to establish reasonable treatmentmethods. First, in this paper we introduce the basic theory of multi-modal medical image fusion,then we focus on multi-scale medical image fusion algorithm of the closed-loop control research.In this paper, the main research contents are summarized as follows:1) Pointing at the problem of multi-modal medical image fusion, a new algorithm formedical image fusion is proposed based on lifting wavelet transform. Firstly, source multi-modalmedical images after registration are decomposed into low frequency sub-band and highfrequency sub-bands by applying lifting wavelet transform. Secondly, image fusion rules aredesigned respectively according to different features of low frequency sub-band and highfrequency sub-band. A fusion rule based on weighted region average energy is adopted in lowfrequency sub-band coefficients. For high frequency sub-bands coefficients, weighedbox-counting method is applied in the fusion rule of low-rise sub-bands with low content ofnoise and the fusion rule of weighed local area energy of image gradient is used for high-risesub-bands with higher content of noise. And combining with the feedback theory, to obtain eachthreshold value adaptively by. using the evaluation index of image edge retention degree QABFas performance index.2) According to the characteristics of multi-modal medical image and the human visualfeatures, a new medical image fusion algorithm in nonsubsampled coutourlet domain based onclosed loop feedback is proposed. Firstly, source images after registration are decomposed lowand high frequency sub-bands using NSCT. According to the low frequency sub-bandsconcentrating the majority energy of the source image and determining the image coutour, afusion rule based on weighted region average energy combined with average gradient is adoptedin low frequency sub-band coefficients. Moreover, according to human visual system being more sensitive to contrast and edge, texture of image, the fusion strategy based on the directivecontrast integrated with the improved sum-modified-laplacian and PCNN are used to fusehigh-frequency sub-bands. Furthermore, closed loop feedback is introduced into low and highfrequency sub-bands to obtain optimal fused weights adaptively by using improved weightedstructure similarity (WSSIM) which highly consistent with the HVS as objective function.In this paper, a lot of experiments of fusion of images including gray images and colorimages are conducted. The experiment results show that the proposed algorithm can significantlyimprove the performance of fusion image in terms of quantity of information, dispersed grayscale, and visual quality. The experiment results are compared and analyzed in terms of visualquality and objective evaluation. The proposed algorithm can effectively preserve edge andtexture information.