Determination of Optimum K Value for K-means Segmentation of Diseased Tea Leaf Images

Detecting diseases from the leaf images of a plant is an important and challenging task. Various image processing techniques like pre-processing, segmentation, classification, etc., are performed to detect plant diseases from its leaf images. Image segmentation is one of the important steps in the process of disease detection in leaf images of plants. A well segmented image increases the accuracy of prediction. In this paper, we have implemented the K-means algorithm to segment leaf images of tea infected with red rust disease caused by algae. The value of K in K-means needs to be set manually. Determining the optimum value of K is crucial to obtain a well segmented image. So, the elbow method and silhouette coefficient determination are employed for this purpose.