Transform domain texture synthesis on surfaces

In the recent past application areas such as virtual reality experiences, digital cinema and computer gamings have resulted in a renewed interest in advanced research topics in computer graphics. Although many research challenges in computer graphics have been met due to worldwide efforts, many more are yet to be met. Two key challenges which still remain open research problems are, the lack of perfect realism in animated/virtually-created objects when represented in graphical format and the need for the transmissiim/storage/exchange of a massive amount of information in between remote locations, when 3D computer generated objects are used in remote visualisations. These challenges call for further research to be focused in the above directions. Though a significant amount of ideas have been proposed by the international research community in their effort to meet the above challenges, the ideas still suffer from excessive complexity related issues resulting in high processing times and their practical inapplicability when bandwidth constraint transmission mediums are used or when the storage space or computational power of the display device is limited. In the proposed work we investigate the appropriate use of geometric representations of 3D structure (e.g. Bezier surface, NURBS, polygons) and multi-resolution, progressive representation of texture on such surfaces. This joint approach to texture synthesis has not been considered before and has significant potential in resolving current challenges in virtual realism, digital cinema and computer gaming industry. The main focus of the novel approaches that are proposed in this thesis is performing photo-realistic texture synthesis on surfaces. We have provided experimental results and detailed analysis to prove that the proposed algorithms allow fast, progressive building of texture on arbitrarily shaped 3D surfaces. In particular we investigate the above ideas in association with Bezier patch representation of 3D objects, an approach which has not been considered so far by any published world wide research effort, yet has flexibility of utmost practical importance. Further we have discussed the novel application domains that can be served by the inclusion of additional functionality within the proposed algorithms.