3D Object Reconstruction From Raster Projection And Image Compression

At present, the practicable 3D reconstruction technology is scarce. The raster projection technology is one of the most promising technologies in 3D reconstruction. This thesis researches and designs a system of 3D object reconstruction from raster projection, based on spatial geometry, projective geometry, image processing, computer graphics and Virtual Reality Modeling language (VRML). This system combines the advantages of both raster projection and binocular vision, has many merits such as quick and exact projection, no-touching, high precision, fast reconstruction and simple equipment. Because the system deals with a sequence of pictures, the data stream is very huge, image compression is necessary.This thesis researches the processes of factoring the wavelet transform into lifting steps, proposes a new method to deal with the constant item, brings forward a method to treat with the 'g' filter, and solves the problem when the factoring steps is odd, and makes a read rule to get the lifting format. Last, combined with Embedded Zerotree Wavelets (EZW) and Self-adaptive Arithmetic coding, an image compression subsystem with optional precision is given out.A new method contrasting the entirely exposal bright image with spacial binarization is proposed in this thesis. Experiments show that the quality of the image binarization is improved significantly, and the calculation is simple and the speed is high. This method, however, must save the bright image and dark image first, thus the memory requirements become bigger,A method of the grouping maximal square error spacial binarization is proposed.