The feature extraction of images is a foundational work for image recognition, image data mining, and content-based image retrieval, and it is also the key issues of pattern recognition and classification. Feature extraction based on gray-scale histograms is a typical algorithm for the medical image feature extraction. For features of liver hydatid CT images that is extracted by using different gray-scale histograms are normalizing scale by uniform quantization, the noise is removed by using a median filter, the contrast is enhanced by limited adaptive histogram equalization; and then the gray-scale histograms is used to get the features of the image. The main features of the image classification are obtained by using statistical and maximum classification distance analysis on the histogram features, and then the classification ability of features is evaluated by discriminant analysis. The result shows that there is a certain discrepancy of statistical analysis for the features extracted by gray-scale histograms; features selected by maximum classification distance enhance the accuracy of image classification. This study would lay a solid foundation for the content-based medical image retrieval and the computer-aided diagnosis system to a certain extent.