Assessment of pavement surface distress is an important component of the pavement management process. Pavement surface distresses characterize failures and distortions of the pavement surface structure. A large number of highway surface images have been collected through the application of a video image system. We present an automated approach that detects pavement surface cracks from a forward viewing video camera system. Initially the oblique imagery is transformed to a rectified one which supports quantitative measurements of the crack patterns. For the detection and extraction of pavement surface distress elements, we propose to use a scale-space image approach, where the image scale is defined based on the level of detail of the image structure to be detected. Finally, the detection of crack patterns is performed considering the image as a 3D intensity surface where the bright and dark lines are considered as ridges and valleys. The approach is based on determining the local directions of the image curvature along the curvilinear lines, and determining where along these directional profiles the second derivative of the line profile reaches its maximum absolute value.