Image processing technique has been implemeted to detect the crack on road surface. However, the accuracy of the detection is still low due to the difficulties in segmentation between crack and non-crack area. This research proposes the implementation of Sauvola technique to perform automatic local segmentation of crack. The methodology involves preprocessing, image segmentation, feature extraction and classification step. In segmentation step, in addition to Sauvola, other techniques, i.e. manual thresholding, Otsu and Bernsen, are also implemented for comparison purpose. The result shows that Sauvola technique produces consistent segmentation results on high, medium and low quality images. Sauvola method also perform the best accuracy detection of 96% among them.