Abstract: In order to realize an automatic recognition and rapid separation of the coal and refuse, the texture features of the coal and refuse, grayscale grade of th e profile surface and other large difference features were applied and the single colure grayscale image of the coal and refuse were treated with the smooth filtering, no ise reduction and other method. A differential box dimensional method was applied to calculate the grayscale image fractal dimension of the coal and refuse in order to distinguish the surface texture roughness of the coal and refuse.The experiment results showed that there were obvious differences in the fractal dimensions of the coal and refuse, the fractal dimension of coal was larger than that of gangue, take the value 2.450 0 as the dividing point, and the fractal dimension of coal was above the di viding point, the fractal dimension of gangue was below the dividing point. According to the differences, an automatic recognition of the coal and refuse could be realize d and could set up a theoretical basis to realize a new coal and refuse separation system.