Tomato (Lycopersicon esculentum L.) is one of the most widely grown crops in the world. This crop is easily prone to various diseases. One such disease is late blight, caused by the fungus Phytophthora infestans. The first symptoms of late blight on tomato leaves are irregularly shaped, water soaked lesions, which are typically found on the younger leaves of the plant canopy. During humid conditions, white cottony growth may be visible on the underside of affected leaves. As the disease progresses, lesions enlarge causing leaves to brown, shrivel and perish. Hence in the present paper, a novel computer vision system has been proposed for detection and analysis of late blight disease. The proposed system implements thresholding algorithm to classify the leaf as diseased or healthy. Later it uses K -means clustering algorithm for analyzing late blight disease. The experiment was carried out on leaves of tomato collected from various plantations. The accuracy, sensitivity and specificity of the developed system in analyzing the late blight disease are 84%, 85% and 80% respectively.