Diabetes Mellitus (DM) is one of the communal medical problems around the world. Quick increase of diabetes in patient’s health is one of the huge challenges of health fitness maintenance. Increase in development of DM causes difficult like Diabetic retinopathy (DR), it prompts to loss of vision. For that we need to detect diabetes in earlier stage itself to prevent the blindness. Detection of this disease with maximum accuracy and less amount of time span is challenging. A non-invasive method is proposed in this project to deal with recognizing Diabetic Mellitus, Non proliferative Diabetic Retinopathy and Anemia through the usage of geometry, color and texture features for analysis and to find out its stage. Then twelve colors represent features of tongue which are established in color gamut of tongue. Then block’s texture values are located on the surface of tongue, with all the eight blocks extra mean are utilized to recognize nine features of textures. Subsequently, total features of 34 are extracted from tongue images based on size, areas, ratio’s and distances. It represents the geometric features. Main target of this project is to diagnose Diabetes Mellitus (DM), Non Proliferative Diabetic Retinopathy (NPDR) and Anemia by contrasting the patient’s tongue and samples of healthy people tongue which are pre-stacked in the system. Patient’s tongue features are extracted and which are compared with featured pictures of tongue which are in datasets in system. Human tongue is fit for showing ailments in view of its texture, orientation, and color. By utilizing this extraction of features a try is made to recognize anemia and its levels. © 2018, Institute of Advanced Scientific Research, Inc. All rights reserved.