In　this　paper,　we　propose　an　improved　Largest　Coding　Unit　(LCU)　level　rate　control　algorithm　for　the　3D　video　coding　extension　of　the　high　efficiency　video　coding　(3D-HEVC).　In　the　proposed　algorithm,　bit　allocation　for　extended　views　is　investigated　in　combination　with　temporal　and　inter-view　Mean　Absolute　Difference　(MAD)　predictions.　Upon　prediction　weights　generated　by　MAD　prediction　errors,　accurate　bitrate　for　LCU　can　be　obtained.　Then,　a　weight-based　R-lambda,　model　is　designed　based　on　the　prediction　weights　by　improving　the　model　parameters　consistent　with　video　content.　Experimental　results　show　that　the　proposed　algorithm　can　achieve　better　rate　distortion　performance　and　bitrate　accuracy　in　comparison　with　the　R4,　model　based　rate　control　benchmark　algorithm　in　HTM.