Image-Based Modeling And Rendering Technology Research

Image-based modeling and rendering technology is one of challenging research subject in the field of machine vision and computer graphics. This paper studied the algorithms used in each step of image-based modeling and rendering. First, extracting the SIFT features of images and match them. Second, using structure from motion method solves the structure of target object and the projection matrixes of images. Third, using patched-based multi-view stereo method reconstructs dense points cloud model of target object. Finally, using poisson surface reconstruction algorithm for point cloud model surface mesh. To overcome the disadvantages of structure from motion algorithm, packet processing is adopted to decrease the influence of the error cumulative of the iterative calculations. In each iteration, three images were used to increase the number of scene points which was restored.