Using Global Optimization Techniques to Segmentation of Magnetic Resonance Images (MRI)

Abstract

The Otsu method is one of the segmentation methods that work by finding an appropriate threshold to segment the image. This paper focuses to improve this method by using global optimization precisely the filled function method. The proposed method has been applied to various MRI images, revealing that a segmentation time was reduced by 80% as an approximate percentage. Then applying the single peak quality value MRI segmentation assessment criteria including power-to-noise ratio (PSNR), mean square error (MSE), and signal-to-noise ratio (SNR), appears to result shows the proposed method took a shorter period than the Otsu approach. Then we hybridized the proposed method with K-means cluster and fuzzy c-means methods. The calculating results with the same above criteria show an improvement in image segmentation by hybrid k-means and fuzzy c-means methods of the proposed method in comparison with the traditional methods.