Crack inspection of bridges is indispensable to maintain the public traffic safety. It is conventionally carried out by specialists through visual inspection manually. Considering the efficiency and security, digital image processing technique is used to inspect the bridges. After analyzing many representative images of cracks, this paper presents the particular region feature, gray value feature and contour feature of cracks on the bridges’ surface and obtains hundreds of cracks’ feature values. We select some representative data to identify cracks from images according to these cracks’ feature values. The software has been developed to distinguish whether cracks exist in an image, and to extract automatically the existent cracks by means of MATLAB. We present experimental results obtained to prove these methods.