Segmentation of melanoma is a tedious process to separate the region of interest from an image. The existing methods fail to segment weak or fine details in an image. To overcome this drawback, a new segmentation technique known as Triclass thresholding technique is introduced. This method is proposed by Otsu’s method. The proposed method classifies the images into three regions namely, foreground, background and to-be-determined region (TBD). TBD area is considered as the suspicious region and this region is segmented repeatedly for extracting the fine or weak objects details of an image. This process will stop only when the iteration is capable of separating the Region of Interest into foreground and background regions only. Finally, the combination of foreground and background is done using a logical union operation. The proposed method gives better performance and low computational complexity while comparing to other existing methods.
Volume 11 | 06-Special Issue
Pages: 1603-1609