Transactions on Mass-Data Analysis of Images and Signals (ISSN:1868-6451)
Volume 9 - Number 1 - September 2018 - Page 19-36
Principal Component Analysis Based Multimodal Medical Image Fusion of MRI and CT in Wavelet Domain
Yosef Jbara
Buraydha Colleges, Al Qassim, Kingdom of Saudi Arabia
Abstract
Image fusion of multi-model bio-medical images assists the general medical practitioner in diagnostics and therapeutic planning of the patient. The process of image fusion merges different images to make into one image so that the resultant image poses all the complementary information of the original image. During diagnostics and therapeutic planning of the patient, image fusion provides the complementary anatomic details of an organ, from the images, which are captured from different modalities. Our effort in this research paper is to present an effective fusion method for medical images, which enhances both spectral and spatial quality of the fused image with adopting the principal component analysis in the wavelet domain. The PCA technique Karhunen-Loeve Transform procedure (KLT) is applied to reduce the noise and dimension of both MRI and CT images, with improved geometrical resolution, which generates several principal component images. Then, the first principal component image of source images, are split into frequency band using discrete wavelet transform. After, the average fusion rule is applied to the coefficients of the low frequency band and gradient weighted maximum fusion rule is performed to the coefficients of the high frequency band. Further using inverse wavelet transform and inverse KL transform, the fused image is generated. In addition, the fused image tested quantitatively with spatial and spectral quality measures. The spatial quality assessment is accessed by standard deviation (SD), the average correlation coefficient (ACC), and the root mean square error (RMSE). Similarly, the spatial quality assessment is accessed by the peak signal to noise ratio (PSNR), the average gradient (AG) and the entropy. The investigation shows that our presented algorithm, PCA-based wavelet technique for image fusion scheme has the best measure in view of spectral and spatial qualities.
Keywords: Image Fusion, Karhunen-Loeve Transform, Multi resolution analysis, Discrete Wavelet Transform, Gradient operator, Quality Assessment
Download Paper
(582 KB)