DEM (digital elevation model) data can provide useful information on topographic features and fault distribution. For clarifying regional fracture system hidden by soil and vegetation covers, this study aims to develop an algorithm of automatic lineament extraction from several criterion of DEM-based topographic features. Four features selected are rectilinear valley, break of slope angle and orientation, col, and linear piedmont, which are generally used in visual interpretation of lineament. In the visual lineament extraction, segments as fundamental component of a lineament are detected through two steps: understanding of macroscopic, regional geographic structures at first and then, interpretation of detailed topographic features. This procedure is implemented for automatic segment extraction by the Laplacian calculation, large depression feature of DEM over local area, and regional slope map that eliminates small scale topography. Next step is connection of the segments by considering macroscopic topographic features and using a search distance for segment pair which is changeable with location and direction. Reliability of lineament consisting of connected segments is evaluated through suitable weighting the four topographic features and assigning the total score to each lineament: lineaments with high scores are regarded to exist certainly. This algorithm, ALDES (Automatic Lineament extraction using DEM-based topographic featureS), was applied to the Abukuma mountains area composed mainly of the Cretaceous granite in northeast Japan. It was clarified that the extracted lineaments with high scores have similar patterns of direction and distribution to those of the visually interpreted lineaments from an ASTER satellite imagery. Consequently, regional fracture systems can be characterized through the high score lineaments.