Since the classical post-classification comparison(PCC) technique was affected hy a significant cumulative error and high classification precision was needed for single image, a change-detection method based on joint-classification of bi-temporal SAR images was presented according to the correlation of the unchanged information in different temporal images. The proposed method took gray-levels as an input. The similarity of gray-levels relating to two pixels at the corresponding position for bi-temporal images was obtained through similarity operator. Then the global threshold value of similarity was got, which was used to control the joint-classifier based on K-means to classify the bi-temporal images. Finally, The change-detection map was produced by comparing with both classified images. Experimental results confirm that the proposed method not only improves the precision of classification for single image but also accurately classifies the unchanged geographical information in different temporal images into the same class. The proposed method reduces the influence of the cumulative error and improves the performance of change detection.