In order to classify the pavement cracks in the pavement crack images, a new chain code based classification method are proposed. In this method, the preprocessing of the crack images is used to remove the unwanted object, noise and enhance the cracks. Secondly, the max entropy thresholding method is used to get the binary pavement crack images. After this, the morphological filtering operation, especially the thinning filtering operation is used to get the skeleton of the cracks in the pavement images. Then the crack tracing based on the predefined chain code are carried on to get the tracing array. Last, the cracks are classified according to the statistic parameters of the tracing array. The result shows this method is simple and reliable.