Abstract: In the present study, considering the importance of the quantificational measurement of soil color, a new soil color classification method based on image processing techniques was designed. To verify the feasibility of this method, soil samples derived from six kinds of soil parent materials were measured color using the traditional method and the modified method. The results showed that the R、G、B values of paddy soil derived from different kinds of parent materials were respectively between 166.4 and 177.6, between 146.1 and 153.5, between 96.1 and 124.2. The R values of the soil in plow pan were much larger than that in plough layer (P<0.05). For the same sample, the number of soil color types measured by the visual method were more than that measured by the modified method. For these 12 soil samples, the average number of color types using the modified method was 1.4, while the mean number of soil color types measured by the traditional method was 4.6. So compared with the visual method, the modified method significantly improved the accuracy of soil color classification (P<0.05). The result proved that compared with the traditional recommended method, the accuracy and repeatability of the modified method were significantly improved as the otherness resulted from observers’ vision disparity was overcame and the method was simple in operation.