Tongue diagnosis plays a significant role in Traditional Chinese Medicine. Owing to its experience-based nature, an increasing number of researchers have been studying on automatic tongue diagnosis methods. However, most methods fail to calibrate the input tongue images which rely on the quality of tongue images. In this paper, we present an optimised computer-aided tongue diagnosis method based on Android platform which makes up the drawbacks and insufficiency of our previous work. First, it calibrates the colour of the input tongue images and makes them standardised. Then the hue range of the brim pixels of the sample images is evaluated based on maximum likelihood estimate method. This can help us obtain edges and link them into a complete outline to finish tongue segmentation. Finally, we realise the new algorithm on Android platform which can give the classification of the tongue and make diagnosis. Compared to existing calibration algorithms, our system shows better robustness, comprehension and accuracy.
(8.0+极速模式)