To circumvent the subjective and qualitative problems of traditional tongue diagnosis, we present a novel computer aided tongue diagnosis system (CATDS). In this system, a standard acquisition device as well as a new color correction method is utilized to capture qualified tongue images. The system is constituted by five components: User Interface Module, Acquisition Module, Tongue Image Database, Image Preprocessing Module and Diagnosis Engine. In contrast to existing CATDS, the proposed system aims to establish the relationship between quantitative features and diseases via the Bayesian networks. System tests are carried out on a group of 544 patients affected by 9 common diseases and 56 healthy volunteers. The results show that the system can properly identify six groups: healthy, pulmonary heart disease, appendicitis, gastritis, pancreatitis and bronchitis with accuracy higher than 75%. Moreover, the execution time for the whole diagnosis process including image preprocessing and diagnosis is less than 5 seconds.