Computer-Assisted Diagnosis of Digestive Endoscopic Images Based on Bayesian Theory

Computerized processing of medical images can ease the search of the representative features in the images. The endoscopic images possess rich information expressed by color and texture. In this paper, a novel scheme is developed to provide an objective, rapid and exactitude analysis of endoscopic images which could provide a valuable tool for diagnosis. Firstly, a new color quantization method which is more realistic is introduced. Then, we extract color-based features from the color histogram and texture-based features from the texture spectra. Finally, a Bayesian classifier is used for the classification. For large amount of endoscopic images, proper diagnosis results corresponding with unique medical features can be acquired, which suggests that the unsupervised endoscopic image diagnosis is applicable.