Dengue Virus Infected Leukocyte Classification on Microscopic Images with Image Histogram Based Support Vector Machine

Dengue virus detection using blood smear with staining method known as immunocytochemistry streptavidin biotin peroxidase complex has early detection problem caused by the low in number of leukocyte in early day of infection. In order to help the detection of virus infection, this research develop automated system to count and classify infected leukocyte from the microscopic image using image histogram based support vector machine. Image processing is for the detection of the cells by using Gram-Schmidt orthogonalization for converting image into grayscale high contrast grayscale image, median filter for image smoothing, Otsu threshold for image segmentation, image morphology and circularity filter for cleaning up unwanted noise after segmentation. Furthermore, support vector machine for the infected cell classification using image histogram of the cells as feature vector. The result of the research is the image histogram based suport vector machine can classify the infected leukocyte with 83.94% accuracy.