Detection of cervical cancer by using thresholding, watershed segmentation and SVM classifier in image processing

Abstract

Cervical cancer is an abnormal growth of cervix tissue. We use algorithm to detect cervical cancer. Noise removal is applied on the image to remove the noise. Image sharpening is applied to detect boundary of the image. Segmentation used to divide the image into regions. Two types of segmentation are used they are thresholding and watershed segmentation. Thresholding value convert grayscale into binary images. Here main selection is the threshold value. Watershed segmentation is the best method to group the pixels on the basis of intensity. After watershed segmentation apply morphological operation. It’s main aim to separate tumor parts from an image. Next phase of the proposed system is multiclass classification that is based on these extracted features. This method uses multiclass Support Vector Machine (SVM) to classify the tumor. After classifying the tumor as stage IA,IB,IIA,IIB,IIIA,IIIB,IVA,IVB, using SVM classifier the overall accuracy is found to be 80 – 90%. The fine detection of tumourous area is a challenging task in medical image science. Using this proposed method it is possible to detect infected area