Research On Self-Similarity Measurement Of 3D Surface Textures

Self-similarity measurement in textures is hot pursuit in recent years and playing an important role in texture analysis, editing, synthesis and visualization in area of computer vision, image processing and computer graphics, and it is leading a great prospect. Traditional texture self-similarity measurement is restricted in 2D textures and is well studied. But these methods don't work well in measuring the self-similarity of 3D surface textures because of the influence from illumination and viewing conditions.Research on 3D surface texture contains appearance representation under different illumination and view conditions, data collection, analysis, synthesis and visualization. Self-similarity of 3D surface textures is measured by different methods within various 3D surface texture representation methods respectively. The measured results are analyzed and evaluated, and finally applied in different 3D surface texture application domains.This thesis introduces similarity and self-similarity measurement in 2D textures, and discusses the mathematical framework of the 3D surface texture. The photometric stereo and SVD techniques and the original captured images are utilized to develop the self-similarity measurement research. For Lambertian model and SVD method, self-similarity data is obtained from model parameters and then utilized in non-linear magnification editing of the 3D surface textures. For original captured images, self-similarity data is computed directly and then filtered by a filter bank to generate feature data for the experiment of texture classification. The experimental results prove the effectiveness of the proposed method. Finally, the work of self-similarity measurement of 3D surface textures is concluded.