Photon　mapping　can　efficiently　render　high-quality　photorealistic　images　with　full　global　illumination.　It　stores　many　photons　in　photon　map　for　improving　reconstruction　qualities　of　complex　illumination,　which　leads　to　high　memory　overhead　and　long　reconstruction　time.　This　paper　presents　a　geometry-bias-based　reconstruction　of　global　photon　maps.　We　divide　the　bounding　box　of　photons　into　a　number　of　grids.　Based　on　the　number　of　photons　and　the　curvature　variation　of　surface　in　every　non-empty　grid,　we　take　a　fraction　of　photons　in　grid　to　reconstruct　global　photon　map,　and　ensure　conservation　of　total　energy　of　photons　in　grid　before　and　after　reconstruction.　Our　method　has　low　computation　cost　and　keeps　view-independence　of　photon　map.　The　experiments　show　that　the　differences　of　reconstructed　illumination　effects　between　two　photon　maps　are　very　small　even　if　the　photon　number　of　photon　map　before　reconstruction　is　20　times　the　one　after　reconstruction.　Compared　with　the　non-reconstructed　photon　map　with　the　same　number　of　photons,　the　reconstructed　photon　map　has　far　better　reconstructed　illumination　effect,　and　can　effectively　reduce　illumination　error　due　to　all　kinds　of　geometry　biases.　(C)　2011　Published　by　Elsevier　Ltd.　Selection　and/or　peer-review　under　responsibility　of　Harbin　University　of　Science　and　Technology