Texture synthesis based hybridisation for images and geometry

Abstract

Texture synthesis deals with the example based creation of highly repetitive information. It is an important and widely studied problem in the computer graphics community due to the ever increasing need for art. Texture synthesis offers a way to automate the creative process so that artists are free to focus on the core aspects of their project, defining a look and style, and free from the mundane and tedious creative tasks of producing many unique instances of similar, repetitive content. The work presented in this thesis takes as input some form of example data and attempts to automatically imagine what more of it could look like. Specifically, we explore the idea of hybridisation, taking multiple unique objects and mixing them to create new instances of those objects which share similar statistical features both locally and globally with the input. Hybrids occur in most textures but thus far have not been studied as a texture synthesis problem. In this thesis hybridisation is identified as a sub-problem of texture synthesis and hybrids are synthesised independent of the texture they would exist in. A new texture synthesis based hybridisation solution is proposed that generates near infinite unique instances of a hybrid from just a small handful of examples. Texture, being a kind of information (highly repetitive information), can exist for any type of multimedia data (e.g. images, audio, geometry, animation, etc.). While some work has been done with regards to customizing texture synthesis algorithms into other content domains such as audio, animation and geometry, the bulk of the work done for texture synthesis has focused on images. In fact the challenges specific to the image domain have heavily influenced the direction texture synthesis research has taken over the past years. We explore an interesting and practical new geometry synthesis approach that allows us to perform texture synthesis and hybridisation directly on 2D curves and 3D meshes, a previously open problem with vast academic and commercial appeal. Given a handful of unique but similar images, 2D curves or 3D meshes, nearly infinite more similar but unique instances are automatically created. While we do not claim that such an algorithm reproduces the phenomena known as human imagination, we do believe that it mimics the final result and offers an autonomous approximation that is immediately useful to industry for the purpose of art creation. We hope that this thesis opens up an interesting new direction for texture synthesis, hybridisation and artificial imagination.