The Study, Based On The Three-dimensional Modeling Of The Multi-view Images

The subject of three-dimension reconstructions is in relation to Computer Vision, Computer Reality, Image Processing and Virtual Reality etc. It is a research hotspot these years. A lot of ways that may solve the problem have been introduced. Because we know little about Computer Vision and Image Understanding, a few accidence theories frame have been brought out. Different researchers have their own model to explain the object. It will affect the universality.The thesis reconstructs the object's 3D model based on several picture taken by a digital camera, and tries three different ways. The first method, make a model in the software 3ds max. This way is free manipulation, but low efficiency. The second method, according to the arithmetic which reconstructs from point to line to face, it get a high-exactitude model, but without good third dimension. The third modeling method is based on model dividing. First make a model by Geometry Based Modeling, and divide it into several curved faces. Then get the textures corresponding to the curved faces in pictures. In this procedure, I present a new arithmetic of image division based on contour linking. This way colligates the strongpoint of the two method mentioned previously, and the effect is good, the workload is low. It is a good modeling way for the object with smooth faces, such as cars.