Reconstruction performances of curvelet transform for magnetic resonance images

Abstract

Reconstruction performances of transforms are deeply associated with imageprocessing, scientific computing and computer vision. This research focuses onthe performance of Curvelet Transform for Magnetic Resonance Images. The mainoutcome of this technique includes the removal of non-homogeneous noise usingCurvelet based de-noising methods. Curvelet Transform belongs to the family ofdirectional Wavelets. Curvelet Transform not only contains translations, dilations butalso the rotations, which can enhance the reconstruction of curve objects. This researchinvolves multi-scale reconstruction of objects with edge discontinuities. Experimentalresults show that Curvelet Transform has superior reconstruction capability for animage with curve objects. Another phase of this research covers the segmentationof de-noised images using Fuzzy C-Means Clustering (FCM) algorithm. The clusterformation in FCM algorithm is based on the Euclidian distance between pixels withsimilar intensities. Experimental results show that segmentation of reconstructedimages is adversely affected by the noise bursts.