Example-based Texture Synthesis Research Articles

Multiscale texture synthesis

Example-based texture synthesis algorithms have gained widespread popularity for their ability to take a single input image and create a perceptually similar non-periodic texture. However, previous methods rely on single input exemplars that can capture only a limited band of spatial scales. For example, synthesizing a continent-like appearance at a variety of zoom levels would require an impractically high input resolution. In this paper, we develop a multiscale texture synthesis algorithm. We propose a novel example-based representation, which we call an exemplar graph, that simply requires a few low-resolution input exemplars at different scales. Moreover, by allowing loops in the graph, we can create infinite zooms and infinitely detailed textures that are impossible with current example-based methods. We also introduce a technique that ameliorates inconsistencies in the user's input, and show that the application of this method yields improved interscale coherence and higher visual quality. We demonstrate optimizations for both CPU and GPU implementations of our method, and use them to produce animations with zooming and panning at multiple scales, as well as static gigapixel-sized images with features spanning many spatial scales.

Style-learning with feature-based texture synthesis

The objective of artistic style learning is to synthesize a new image from a source image with the style learnt from example images. Existing example-based texture synthesis (EBTS) techniques model style with low-level statistical properties. These methods work well with some artistic styles such as oil painting, but have difficulties in preserving image details and features for other styles, such as pencil hatching. In this article, an improved artistic style-learning algorithm with feature-based texture synthesis (FBTS) is introduced. Compared with existing EBTS methods, in our FBTS algorithm, image details and features are better defined with a feature field generated from the source image. Also, an improved L2 neighborhood distance metric which provides better measures of perceptual similarity is proposed. Results and comparisons are given to demonstrate the effectiveness of the FBTS algorithm with applications in the areas of stylized shading and artistic style transfer.

Vector field design on surfaces

Vector field design on surfaces is necessary for many graphics applications: example-based texture synthesis, nonphotorealistic rendering, and fluid simulation. For these applications, singularities contained in the input vector field often cause visual artifacts. In this article, we present a vector field design system that allows the user to create a wide variety of vector fields with control over vector field topology, such as the number and location of singularities. Our system combines basis vector fields to make an initial vector field that meets user specifications.The initial vector field often contains unwanted singularities. Such singularities cannot always be eliminated due to the Poincaré-Hopf index theorem. To reduce the visual artifacts caused by these singularities, our system allows the user to move a singularity to a more favorable location or to cancel a pair of singularities. These operations offer topological guarantees for the vector field in that they only affect user-specified singularities. We develop efficient implementations of these operations based on Conley index theory . Our system also provides other editing operations so that the user may change the topological and geometric characteristics of the vector field.To create continuous vector fields on curved surfaces represented as meshes, we make use of the ideas of geodesic polar maps and parallel transport to interpolate vector values defined at the vertices of the mesh. We also use geodesic polar maps and parallel transport to create basis vector fields on surfaces that meet the user specifications. These techniques enable our vector field design system to work for both planar domains and curved surfaces.We demonstrate our vector field design system for several applications: example-based texture synthesis, painterly rendering of images, and pencil sketch illustrations of smooth surfaces.