Abstract
Melanoma is a deadly skin cancer which increases the death rate at a faster rate. In order to bring the death rate under control, melanoma should be detected at its earlier stage. To achieve this, researchers have introduced Computer aided diagnosis and adopted the same. In this technique, Segmentation is found to be one of the important steps. Many algorithms exist in practise for segmentation,where one of the important algorithms is Traditional Otsu Segmentation Algorithm. In this algorithm the major drawback is that the segmentation is improper in the presence of variable illumination. This paper proposes an algorithm “Normalised Otsu Segmentation” which overcomes the above mentioned drawback and results in an accurate segmentation. This algorithm first normalises the image to overcome variable illumination and then segments the image using Otsu algorithm. The accurate result given by this algorithm can be used in further steps to detect the lesion accurately which will provide a hand a for reducing the death rate.
1. Introduction
Melanomacan be detected at earlier stage using computer aided analysis in which segmentation is the major and important step1-3. Segmentation is done with the help of Segmentation algorithm to speed up the process. After segmenting the image into subregions measurements on each region can be accomplished4 . Therefore segmentation is a major step for significant analysis of image data. Various techniques have been proposed for segmentation in many literatures. Discontinuity and similarity are the two properties based on which categorization of image segmentation method is done5 . Region based and edge based segmentation are the two categories of image segmentation based on these properties. Based on the property of discontinuity of pixels segmentation methods are classified as edge or boundary based techniques. A binary image is the result of edge based segmentation. The two categories of edge based segmentation methods are gradient based and gray histogram based methods6 .
6. Conclusion
This literature provides an algorithm which can be implemented easily for obtaining better and accurate results in the segmentation process, which provides a better and wider path for melanoma prediction. The proposed algorithm of this literature produces less error rate and high PSNR value which proves that the segmented image quality is high. When this proposed algorithm is used for segmentation, the problem of variable illumination can be overcome. When this variable illumination problem is removed,the result of the segmentation process is more accurate and clear, which provides a wider and clear way for extracting features from the segmented image in more qualified and quantified manner.
