The recognition method of three kinds of rice diseases (blast,sheath blight,and bacterial leaf blight) was studied by image processing technique and the Bayes discrimination method. Firstly,the disease spots were segmented by color features and their outlines,and 63 parameters of shape,color and texture features were extracted. Secondly,the stepwise discriminant analysis was used to select effective recognition parameters from four parameter sets of shape,color,texture feature and all of them. Finally,the Bayes discrimination method was applied to classify and recognize the three kinds of rice diseases. The results showed that the number of texture parameters decreased to 35.2% and the highest recognition rates of four parameter sets was 97.2%. This method could also be applied to recognize diseases of other crops.