Image texture analysis is one of the effective methods for improving the accuracy of image interpretation and classification. This paper deals with textures of terraced fields and features of frequency images. The linear texture of image was converted into the distribution pattern with frequency lines crossing the center of the frequency space and the orientation vertical to the linear texture of the image. Taking the IKONOS image of 1 m resolution for extracting information of terraced fields as an example，the authors used the ratio of the maximum orientations as the main texture extraction index of terraced fields in 5 pixel×5 pixel window. The result shows that the accuracy of image classification of terraced fields is up to 81.3%.