Segmentation of an image plays a very important role in the analysis of an image. It acts as a fundamental step in medical image analysis, detection of objects, face recognition, iris recognition, fingerprint recognition, content-based image retrieval, clustering or classifying the contents of an image etc. Images are segmented considering different aspects like intensity information, similarity among the pixels, discontinuity information etc. Sometimes hybrid techniques are used which considers more than one characteristic. The intensity-based or threshold-based approach is the simplest approach which considers different levels of intensity to obtain the thresholded image of the original image. Discontinuity based methods detect edge boundaries considering variations in pixel intensities. Region based segmentation methods assume that inside a specific region all the pixels have same characteristics and in two different regions pixels have varying property. Clustering-based methods groups the pixels in the image considering the similarity in their characteristics. These segmentation methods can also be combined to generate hybrid methods (Khan & Ravi, 2013).
Motivation
Segmentation plays a vital role in medical image processing in the study of internal structure of body, planning of surgery, detection of blood vessels, study of brain development, analysis of different body organs, diagnosis of tumors and affected parts of a body like mussels, nerves, bones, blood vessels, etc. For generating the images of internal body parts by using different imaging techniques like x-ray, computed tomography (CT), ultrasound, and magnetic resonance imaging (MRI) is used. Out of these, x-rays are one of the oldest techniques and the MRI is a modern imaging technique. The MRI technique is popular because of its capability to detect soft tissues like cartilage and ligaments, give information about movement of blood through blood vessels and organs, show changes in internal body structure effectively. It can also generate three dimensional and cross-sectional view of the body. The disadvantage of this approach is that it generates noise and is very expensive.
Genetic algorithm is widely used in different image processing techniques due to its ability to avoid local optima. Convergence speed of genetic algorithm is very fast as compared to traditional search methods. In traditional methods, searching starts from a single point whereas genetic algorithm searches from a number of points in parallel. Genetic algorithms also give good optimization results in noisy environments.
Fuzzy logic is simple to implement and easy to interpret. It has the ability to solve nonlinear problems, handle uncertain and imprecise data or information. So, it is a good idea to use fuzzy logic with genetic algorithm to generate selection rules using the concept of probability. Combining the advantages of both techniques it is possible to obtain an optimal threshold value which can be helpful in segmentation of images. The need of thresholding is very much essential for medical images since it helps in proper analysis of ultrasound, x-ray and MRI images. This motivates us to apply the probability based optimal values for medical image thresholding and perform a comparative study to conclude the best-observed technique of probability estimation.