A general framework of robot manipulation planning for door opening task is proposed in this paper. First grasp pattern, including grasp point and method for various door handles, is defined. Second door-opening pattern, including task planning and trajectory planning, is also proposed. This paper utilizes Kinect as a vision sensor to collect point cloud data of the environment. Viewpoint Feature Histogram (VFH) descriptors are used to recognize the type of the door handle. Then corresponding parameters of grasp pattern and door-opening pattern are extracted using Point Cloud processing algorithms. The robot follows the corresponding grasp pattern to grasp the door handle and repeats the motion trajectory to finish each sub operation. Two experiments were conducted to validate the proposed framework.