Recondtruction And Texture Acquisition Based On SOR

How to restorate three-dimensional structure from two-dimensional image is referred to as three-dimensional reconstruction ,which is an important area of human vision. The problem of three-dimensional reconstruction from the single uncalibrated image is crazy researched nearly decade, which is widely used in virtual reality, computer animation, military simulation, medical imaging and other areas.In this paper,the problem were discussed and studied.First of all, this paper analysis single-axis movement and geometric properties of SOR. From the research, it shows that using only a single uncalibrated view of surface of revolution(SOR)'s apparent contour and two distinct imaged CrossSections can provide enough information for calculating the invariant and the 3D reconstruction of SOR. Based on this, the paper realized the 3D reconstruction algorithm based on SOR. Experiments show that the algorithm can be applied to any SOR'3D reconstruction which contains two CrossSections and smooth surface of SOR.The algorithm is simple operation,high precision and need lowly quality of image. The experiments show that the algorithm can avoid texture distortion result from sampling the SOR's texture,also can maintain brightness and luster of the image of the original SOR surface.Secondly, this paper realized and improved the algorithm for extracting SOR's textrue. Using this algorithm can expand SOR's surface texture into a plane. In the process of extracting texture, due to joining the bilinear interpolation algorithm, it eliminat the block of the edge of texture pattern,which is more realistic.Then, this paper analysis and research the scale-invariant feature extraction (SIFT) algorithm, then proposed and improved feature extraction and matching algorithms for SOR's texture patterns sequence. By changing description structure of feature point, the number of dimension of feature vector is reduced in this paper. In addition, under the disparity gradient constraintst, In the process of matching, the algorithm uses Least-Median-Squares estimation to eliminate false matching pairs. With experiments, it is obvious that this algorithm has good robustness as well as can reduce the matching time in the circumstances with more deformation and distortion. So, this algorithm can achieve the real-time requirement.Finally, this paper realized the algorithm for SOR's textur sequence mosaic and fusion. In the mosaic process, comparison and analysis the weighted smooth, Laplace pyramid transform and wavelet transform mosaic and fusion algorithmand, then use of indicators such as information entropy, cross entropy, variance, etc. to the evaluation of the algorithm.Experiments show that for SOR's texture image sequence, mosicing and fusion algorithm based on wavelet transform is superior to other algorithms, which can eliminate crack in the surface texture image sequence.