A Novel Vector Field Convolution-Radix4 FFT for MRI Image Segmentation

  • R.Rajeswari , G.Nirmala Priya

Abstract

Vector Field Convolution (VFC) snakes are having the advantages of enhanced capture range,  convergence at the contour of the images and  initialization insensitivity and concavity convergence. But the disadvantage of VFC is that it  has heavy computation load. In this paper, a novel Radix4-FFT-based with modified edgemap for VFC segmentation model is proposed which is fast, accurate and robust to noise for MRI brain segmentation and classification. Firstly, Canny operator was applied for the detection of edges of brain MRI images, and the result was used as the edge map which has modified for GVF snake model. Then calculation of external force is done by vector field convolution with the modified edge map derived from the image by using Radix 4-FFT with reduced  cost of computation, and also gives superior noise suppression and initialization. After that, by applying a Hamming window with GVF snake-Radix4-FFT, the leakage at the boundary of tumour has been eliminated which gives enhancement of MRI images. The main advantage of Hamming window , it  used to eliminate the possible leakage problem because of choosing inappropriate parameters. Then, Hamming window with Radix4-FFT applied to compute the spectral leakage of the MRI brain image which accurately defines the boundary region of tumour or boundaries of WM, GM and CSF tissues in the brain. Finally, the proposed VFC-based Radix4-FFT–Hamming window segmentation model is used to refine the rough segmentation. The proposed segmentation method applied on both synthetic and real images of MR brain indicate that the method can be used for accurate ,robust  and  fast segmentation.

Published
2020-06-01
How to Cite
R.Rajeswari , G.Nirmala Priya. (2020). A Novel Vector Field Convolution-Radix4 FFT for MRI Image Segmentation. International Journal of Advanced Science and Technology, 29(11s), 246 - 260. Retrieved from https://sersc.org/journals/index.php/IJAST/article/view/19975
Section
Articles