We　present　a　scatter　point　cloud　denoising　method,　which　can　reduce　noise　effectively,　while　preserving　mesh　features　such　as　sharp　edges　and　corners.　The　method　consists　of　two　stages.　Firstly,　noisy　points　normal　are　filtered　iteratively;　second,　location　noises　of　points　are　reduced.　I-low　to　select　proper　denoising　neighbors　is　a　key　problem　for　scatter　point　cloud　denoising　operation.　The　local　shape　factor　which　related　to　the　surface　feature　is　proposed.　By　using　the　factor,　we　achieved　the　shape　adaptive　angle　threshold　and　adaptive　optimal　denoising　neighbor.　Normal　space　and　location　space　is　denoising　using　improved　trilateral　filter　in　adaptive　angle　threshold.　A　series　of　numerical　experiment　proved　the　new　denoising　algorithm　in　this　paper　achieved　more　detail　feature　and　smoother　surface.