Spatial fuzzy c-mean sobel algorithm with grey wolf optimizer for MRI brain image segmentation
Segmentation is the process of dividing the original image into multiple sub regions called segments in such a way that there is no intersection between any two regions. In medical images, the segmentation is hard to obtain due to the intensity similarity among various regions and the presence of no...
Saved in:
| Main Author: | |
|---|---|
| Format: | Thesis |
| Language: | English |
| Published: |
2017
|
| Subjects: | |
| Online Access: | http://eprints.utm.my/id/eprint/79123/1/ImanOmidvarTehraniPFC2017.pdf |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | Segmentation is the process of dividing the original image into multiple sub regions called segments in such a way that there is no intersection between any two regions. In medical images, the segmentation is hard to obtain due to the intensity similarity among various regions and the presence of noise in medical images. One of the most popular segmentation algorithms is Spatial Fuzzy C-means (SFCM). Although this algorithm has a good performance in medical images, it suffers from two issues. The first problem is lack of a proper strategy for point initialization step, which must be performed either randomly or manually by human. The second problem of SFCM is having inaccurate segmented edges. The goal of this research is to propose a robust medical image segmentation algorithm that overcomes these weaknesses of SFCM for segmenting magnetic resonance imaging (MRI) brain images with less human intervention. First, in order to find the optimum initial points, a histogram based algorithm in conjunction with Grey Wolf Optimizer (H-GWO) is proposed. The proposed H-GWO algorithm finds the approximate initial point values by the proposed histogram based method and then by taking advantage of GWO, which is a soft computing method, the optimum initial values are found. Second, in order to enhance SFCM segmentation process and achieve higher accurate segmented edges, an edge detection algorithm called Sobel was utilized. Therefore, the proposed hybrid SFCM-Sobel algorithm first finds the edges of the original image by Sobel edge detector algorithm and finally extends the edges of SFCM segmented images to the edges that are detected by Sobel. In order to have a robust segmentation algorithm with less human intervention, the H-GWO and SFCM-Sobel segmentation algorithms are integrated to have a semi-automatic robust segmentation algorithm. The results of the proposed H-GWO algorithms show that optimum initial points are achieved and the segmented images of the SFCM-Sobel algorithm have more accurate edges as compared to recent algorithms. Overall, quantitative analysis indicates that better segmentation accuracy is obtained. Therefore, this algorithm can be utilized to capture more accurate segmented in images in the era of medical imaging. |
|---|