The　digital　image　processing　system　for　pavement　distress　assessment　benefits　from　good　segmentation　technologies　because　a　accurate　segmentation　result　leads　to　correct　defect　recognition　and　accurate　assessment.　In　this　paper,　an　improved　C-V　segmentation　method　based　on　a　narrow　band　of　the　active　contour　line　is　proposed.　The　method　substitutes　the　local　region　with　a　narrow　band　of　the　active　contour　line　in　the　local　region　C-V　model.　Further,　in　the　iteration　equation,　the　original　image　is　replaced　by　the　gradient　image　to　process　pavement　distress　images.　The　experimental　results　show　that　the　inhomogeneous　objects　could　be　segmented　effectively　and　achieve　good　results.　©2010　IEEE.