K-mean algorithm for Image Segmentation using Neutrosophy

Image Segmentation is a important step in the major applications such as Image Processing, Recognition Tasks, Object Detection, Medical Imaging etc. Method used for image segmentation is responsible for the quality of resultant segments. High quality segmentation requires a method that segments an image into more accurate and relevant results. This paper introduces a new approach for segmenting an image. It combines two learning algorithms, namely the K-means Clustering and Neutrosophic logic, together to obtain efficient results by removing the uncertainty of the pixels. A Neutrosophic domain is defined to characterize an image into three membership sets: Truth, Falsity and Indeterminacy. The Indeterminacy Set is compared against a threshold value. If Indeterminacy is found to be greater than threshold, which means that the pixel may belong to more than one cluster, we change the intensity of the pixel depending upon the truth value. The K-means Clustering algorithm is then employed on modified pixels to obtain hard clusters. Experimental Results verify that the results obtained are more accurate, thereby improves the quality of segmentation.