Color dependent K-means clustering for color image segmentation of colored medical images

In the present work, a neoteric image segmentation technique has been framed, which is stood on color of the image using an unsupervised K-means clustering. The color image is converted into Lab (L=luminocity layer; a=chromaticity layer 1; b = chromaticity layer2) in various computational steps and each layer has its own importance. Clustering is a process to distinguish different kind of objects in an image and K Means clustering partitions the image, such that within each cluster same type of objects are as close as possible and each cluster must be distinguished. Having several clusters we segment the nuclei into a separate image by recalling L layer. The proposed technique enables detection and analysis of objects without feature computation of every pixel in the image. Support vector machine (SVM) classified the disease by comparing the selected clustered image which is very much close for detection of disease with the existing standard data. The results so obtained with the proposed technique have been shown, which clearly depicts the segmentation of the complex colored medical images.