Segmentasi Daun Cendana Berbasis Citra Menggunakan Otsu Thresholding

Abstract

The segmentation process is the separation of parts of the object area from the background in an image, so that segmented objects can be processed for other purposes such as pattern recognition. The results of segmentation must be accurate, if it is not accurate in separating objects in the image it will affect the results of further processing. The segmentation process is carried out using the Otsu Thresholding method on sandalwood leaf images by first applying the Median filter to reduce noise. After obtaining the segmented image, then performing performance measurements. The segmentation results from each test are evaluated using the RAE (relative foreground area error) and ME (misclassification error). The segmentation results of 8 sandalwood leaf images from 2 existing conditions show that, sandalwood leaf image segmentation with good leaf conditions obtains the best segmentation results with smaller errors of 5 image data. While the images of sandalwood leaves affected by the disease as many as 3 image data have more diverse areas so that the segmentation results are not good without any morphological process