Lutsyk O. Information technology to data pre-processing, with non-ideal lighting problems in photometric stereo

The thesis is devoted to the improvement of information technology data pre-processing tasks in three-dimensional reconstruction methods of photometric stereo through the development and modification of pre-treatment and analysis. Particular attention is paid to the not ideal lighting, which occurs in the reconstruction of real surfaces and makes a significant error in the final result of the reconstruction of three-dimensional surface. To achieve this goal a method for correcting lighting is developed. Preprocessing methods for image filtering in the problem of metal surfaces reconstruction are proposed. A method for the localization of the secular regions in the images of the surface is developed. The improved method of integration normals field is proposed. It is based on the proposed algorithm for splitting and local areas on the basis of Schwartz equations. It is improved Kohonen neural network through consideration of entropy neurons weights. This improved preliminary classification problem in the task of the segmentation in grayscale image areas. Micro program modules which implement neural network and new methods for preliminary processing of images on a homogeneous computing environment are developed. It gives a possibility to speed up the information technology of three-dimensional reconstruction by photometric stereo in the case of not ideal lighting.