The Research and Realization of SFS Based on OpenGL
|School||Huazhong University of Science and Technology|
|Keywords||Radial Basis Function Shape Reconstruction Shape from Shading OpenGL Tabbed Dialog|
The real reconstruction for 3D Models with computer has been a hot research of digital image processing and computer graphics. More attention is paid to methods of restoring true three-dimensional objects accurately and vividly. This paper described development of the problem for shape from shading and four classical solutions, namely, minimization approaches, propagation approaches, local approaches and linear approaches. It introduces the calculus of variations in the application of the problem of the shape from shading. With the radial basis function introduced into the calculus, the radial basis function network model is achieved and the flow of learning flow as well as algorithms of the model.This paper designed the shape reconstruction system by using the RBF network model under the help of the MFC single document and dialog frame. It described the function of each module of the system as well as the basic knowledge of OpenGL. Therefore, this paper provided the essentials of the system implementation: the design of the CMyOpenGL class, including the effect and details of eight member functions; the design of graphical user interface of the shape reconstruction .The GUI included four parts, namely, toolbar, drawing area, functional bar and status bar. In order to meet different practical needs, this paper designed two approaches to change the material properties, color table approach and tabbed dialog approach. Finally, classic pictures are tested such as vases, Mozart’s head and the realistic images of objects such as human face by using the shape reconstruction system. The recover results are analyzed and compared with each other.The observation mode designed can satisfy most needs of practical applications, with a very good expansion. It can easily be reformed into the binocular system and medical reconstruction system. However, only shape similarity is achieved and the accuracy of any analysis of result is still lacked. We will take further research in the direction of algorithms and accuracy of the SFS.