Research Of Style Enhancement Technology For Building Images

With the development of computer graphics (CG), people are not satisfied to the realisticeffects generated by CG any more. As the needs for graphical expression are diverse moreand more, computer graphics is diverged into two directions, namely realistic rendering andnon-photorealistic rending. Realistic rendering mainly pursuits for the simulation of the realworld, using physical modeling method to simulate the natural illumination with light,material, shadow and so on. In this way CG generated images are close to reality, looking likephotos taken with camera. However, realistic rending is too precise, that some importantdetails are not properly stressed. So realistic rendering does not always provide the bestexpression. In some cases, people often don’t need accurate rendering. For example inmedical, architecture and other fields, people hope to ignore minor details, highlight some keyinformation, leading to non-photorealistic rending. Non-photorealistic rending is based onhuman perception, combined with the rules of art and technology, to generate images withhand-painted style, mainly illustrating art appearance. In medical field, architecture andeducation area it is playing a more and more important role.The thesis focuses on the effect of stylization of building images, based on the study ofthe generation algorithm of line drawings, according to the characteristics of buildings.The thesis mainly includes:(1) study the characteristics of the realistic rendering anddeficiency, introduce the concept, its characteristics and the importance and necessity ofexistence of non-photorealistic rending, and points out the research background andsignificance, and analyzes present situation in domestic and international ofnon-photorealistic rending;(2) discuss several available style effects of2D non-photorealisticrending;(3)study the line drawing algorithm based on edge tangent flow (ETF) combined withanisotropic Gaussian filtering;(4)design a novel line drawing algorithm specialized forbuilding images, which enhances the abstract buildings characteristics. The method uses theline segment detector algorithm to detect the important edge information of building, and thencombining with the original image gradient field to construct a new edge tangent flow field.Based on the edge tangent flow field the anisotropic Gaussian filtering technique is used toextract the coherent line drawings.(5) The building line drawings and the original image bilateral filtering method for image fusion stack, making building abstract style image.