Color feature of an image is the most common used feature in content-based image retrieval (CBIR) systems. Combining the traditional color-histogram with some spatial information will greatly increase the accuracy of matching two color images. One advantage of color feature is that it is independent to the distortion of an image. However, with the introducing of spatial information, this advantage will disappear. In this paper, we propose a new method to compare two images, which based on symmetrical color-spatial feature (SCSF). Our experiments show that the SCSF is independent to the symmetrical distortion of an image, such as rotation and flip.