Abstract: Based on the analysis of the characteristics of microscopic image of macerals in the exinite of coal, in view of the difference of texture and directional information between them, an automatic classification method for macerals in the exinite of coal, which is based on contourlet transform and extreme machine (ELM), was proposed.Firstly, the microscopic image of exinite were decomposed into multi-scale and multi-direction with the contourlet transform, and statistical features were extracted from each sub-band to form a feature set.Then, a classifier named extreme learning machine was built into the classification of the macerals.Experimental results showed that comparing with other methods for texture feature extraction, the classifier trained with the proposed feature set has obvious advantages in classification accuracy, with an accuracy of 97.64%.Comparing with support vector machine (SVM) classifier, the accuracy of classification for exinite macerals with ELM is 2% higher than that of SVM, and the time of classification is markedly reduced.