Previous　ray　tracing　methods　usually　treat　point-cloud　models　with　attributes　including　coordinates,　normals　and　radius　of　points.　While　3D　coordinates　of　points　can　be　precisely　acquired　by　equipments,　normals　and　radius　of　points　need　to　be　computed　before　ray　tracing　algorithm.　Such　computation　always　takes　a　long　time,　and　produces　various　errors.　This　paper　proposes　a　novel　ray　tracing　method　of　point-cloud　models,　based　on　K-nearest-neighbors　of　iterative　points.　The　method　locates　a　finite　number　of　nearest　points　to　the　iterative　point　in　a　ray,　and　computes　the　normal　vector　of　local　surface　by　using　area-weighted　average　of　normals　of　triangles,　which　consist　of　the　iterative　point　and　the　Knearest　points.　The　intersection　and　its　normal　are　obtained　by　firstly　computing　intersections　between　ray　and　regular　triangles,　and　then　blending　normals　of　regular　sampling　points　of　local　surface,　which　produce　smooth　rendering　effect　and　represent　more　geometric　details.　Numerical　experiments　show　that　our　method　ensures　coherence　of　normals　of　intersections,　especially　for　half-open　or　fragmented　point　clouds.　Moreover,　our　method　can　progressively　render　point　clouds　in　multi-resolution　mode.　We　employ　a　balanced　binary　tree　to　locate　the　nearest　points,　and　employ　grids　to　avoid　unnecessarily　iterative　computation　in　the　space　of　point　clouds.　©,　2015,　Journal　of　Information　and　Computational　Science.　All　right　reserved.