The pre-processing is done by getting input images. Gray scale, Gaussian filter Effects, Otsu Threshold is applied to the input images and Lesion is Traced and unwanted pixels are been removed from Threshold Image. The Traced Lesion image is got passed to Texture Distinctiveness. Texture Distinctiveness (TD) consists of Sparse texture distributions and Texture Distinctiveness (TD) metric. The Sparse texture distributions and Texture Distinctiveness metric are done by applying kmeans clustering on image. Region classifications of the images are done by using some SRM segmentation on the image and the Sobel edge detection of the image is done. The image got segmented by the above methods with Distinctiveness based Segmentation. The Haralick GLCM Texture Feature Extraction Technique of the segmented image is calculated and passed to the densenet 101 classification and classified whether image is Benign or Malignant .The proposed system achieved the Accuracy of 89.08%, Recall of 89.58%, Precision 88.00% and F1 score as88.78%.

Built With

Share this project:

Updates

Submission history