This paper presents a method to characterize asphalt pavement macrotexture using the gray level co-occurrence matrix (GLCM). Data collected at 37 field sites are included in the analysis, representing 6 types of asphalt surface layers such as dense asphalt concrete (DAC), stone matrix asphalt (SMA), rubber asphalt concrete (RAC), ultra-thin wearing course (UTWC), micro-surfacing (MS), and open graded friction course (OGFC). This paper documents the investigation into the differences of GLCM indicators under various pixel pair spatial relationships. Then, the average of each GLCM indicator in some pixel pair spatial relationships is selected for mean texture depth (MTD) correlation. The correlation analysis shows there are 2 GLCM indicators, f_8 and f_9, have strong relationship with MTD, which are entropy of the gray level sum distribution and the gray level combination distribution of pixel pairs of pavement macrotexture respectively. The larger the values of f_8 and f_9, the more complex of the pavement macrotexture. The correlation coefficients between MTD and f_8, f_9 are 0.9601 and 0.9493 respectively. The exponential models are better choice for connecting f_8 and f_9 with MTD, which are highly significant. The mean square errors (MSE) of the exponential models with f_8 and f_9 are 0.00343 and 0.00351 respectively.