Objective To discuss the classification effect of different pretreatment methods combined with shape and texture features on Xinjiang Kazakh esophageal cancer images. Methods Firstly, the images of ulcerated and constricted esophageal cancer were preprocessed by sharpening, median filtering combined with sharpening and median filtering combined with histogram equalization, and then Hu Invariant Moments feature and Gray level Co-occurrence Matrix feature were extracted. The features are finally classified by the KNN classifier to verify the efficiency of the preprocessing method on the image and the classification efficiency of the features. Results The accuracy of Hu Invariant Moments, Gray level Co-occurrence Matrix and mixed feature classification after sharpening preprocessing are 93.27%, 73.35% and 92.91%, respectively. Conclusions ①The sharpening method can highlight the details of the image and improve the representation of the feature.②The classification efficiency of Hu Invariant Moments shape feature is better than that of Gray level Co-occurrence Matrix texture feature classification,It is concluded that the sharpening filter combined with Hu Invariant Moments shape features is more suitable for the classification research of Xinjiang Kazakh esophageal cancer.