Font Size: a A A

Research On Medical Image Fusion Method Based On Multi-scale Transform

Posted on:2015-02-26Degree:DoctorType:Dissertation
Country:ChinaCandidate:Y C YangFull Text:PDF
GTID:1108330464474443Subject:Intelligent Transportation Systems Engineering and Information
Abstract/Summary:PDF Full Text Request
With the development of medical image engineering and computer technology, medical image has become an indispensable part of modern medical technology. Medical images with different imaging principle can provide different diagnostic informations for doctors, but doctors will give the one-sided treatment with single medical image. Through the research of medical image fusion, the advantages with different medical images can be integrated together, from which doctors can get more scientific judgment with high efficiency in the practice of medical diagnosis and treatment.The processing mode of multi-scale transform accords with the visual feature of human being and has characteristics, such as multi-resolution analysis,direction and so on,it can capture more details and edges information of image, so the multi-scale transform has been widely used in medical image fusion. In order to better meet the needs of clinical diagnosis and treatment, medical image fusion methods based on multi-scale transform have been proposed in this dissertation. The main research work is as follows:(1) A weighted local contrast method of medical image fusion based on wavelet transform has been proposed to deal with the problem of the fusion image edge distortion in the medical image fusion methods by wavelet transform.The fusion rule based on region standard variance is adopted in low frequency sub-band coefficient.The fusion rule of weighted local contrast is adopted according to the human visual characteristic and the directional characteristics in high frequency sub-band coefficients. The proposed method has better fusion effects and quantifying indicators than the traditional fusion methods.(2) Aiming at some problems of poor real-time,high computation complexity and large memory demand in medical image fusion. This dissertation proposes a method of medical image fusion based on lifting wavelet transform and adaptive PCNN(Pulse Coupled Neural Network). The fusion rule based on region energy is adopted in low frequency sub-band coefficient according to regional correlation.The fusion rule based on adaptive PCNN is adopted in high frequency sub-band coefficients.Through the application of the simplified PCNN model, the adaptive PCNN uses gradient energy of each pixel as the linking strength of PCNN, so that the linking strength can be chosen adaptively based on the change of pixel gradient energy, the high frequency sub-band coefficients are determined by the times of firing. The proposed method can make the information of detail edges in the fusion images richer and get the better fusion effects.(3) For the problem of the blurring details and edge fusion image in the transition region.The dissertation proposes a novel method for medical image fusion based on NSCT(Nonsubsampled Contourlet Transform) and regional fire PCNN. The fusion rule based on regional fire PCNN is adopted in low frequency sub-band coefficient. Characteristics of regional fire is analysed by the regional fire intensity, the low frequency sub-band coefficient is determined by characteristics of regional fire. When choosing the bandpass directional sub-band coefficients,the dissertation makes the best use of directional characteristics of NSCT.The fusion rule based on improved energy of Laplacian is presented in bandpass directional sub-band cosfficients. The proposed method can make the information of transition region in the edges of fusion images rich and natural, and get clearer fusion image.(4) A method of colorful medical image fusion for anatomical structure image and functional image based on NSCT has been proposed by using the models of RGB and IHS. Firstly,the IHS transformation of the functional image is conducted to extract the I component.Then,after the NSCT of I component and anatomical structure image,the fusion rule based on regional average gradient is adopted in low frequency sub-band coefficient, and local contrast fusion rule is adopted in high frequency sub-bands coefficients. Finally, the fused I’ component,hue component and saturation component are conducted by inverse IHS. The fused result is obtained by inversed IHS.This method can effectively retain the information of colors and functional metabolism of PET or SPECT.
Keywords/Search Tags:Medical Image Fusion, Multi-scale Transform, Wavelet Transform, Lifting Wavelet Transform, Nonsubsampled Contourlet Transform
PDF Full Text Request
Related items