In　rock　engineering,　fracture　measurement　is　important　for　many　applications.　This　paper　proposes　a　novel　method　for　rock　fracture　skeleton　tracing　and　analyzing.　As　for　skeleton　localizing,　the　curvilinear　fractures　are　multiscale　enhanced　based　on　a　Hessian　matrix,　after　image　binarization,　and　clutters　are　post-processed　by　image　analysis;　subsequently,　the　fracture　skeleton　is　extracted　via　ridge　detection　combined　with　a　distance　transform　and　thinning　algorithm,　after　which　gap　sewing　and　burrs　removal　repair　the　skeleton.　In　regard　to　skeleton　analyzing,　the　roughness　and　distribution　of　a　fracture　network　are　respectively　described　by　the　fractal　dimensions　D-s　and　D-b;　the　intersection　and　fragmentation　of　a　fracture　network　are　respectively　characterized　by　the　average　number　of　ends　and　junctions　per　fracture　N-average　and　the　average　length　per　fracture　L-average.　Three　rock　fracture　surfaces　are　analyzed　for　experiments　and　the　results　verify　that　both　the　fracture　tracing　accuracy　and　the　analysis　feasibility　are　satisfactory　using　the　new　method.