Abstract: We have developed a genetic algorithm approach for automatically generating procedural textures. Our system, known as GenShade, evaluates evolutionarily generated procedural textures by comparing their rendered images with single or multiple target images of real textures. It uses a multiresolution image querying metric to automatically prioritize parents for breeding. GenShade simulates several key factors in natural selection. It employs a multiple generation breeding population, a notion of gender, and the concept of aging to maintain diversity while providing many breeding opportunities to highly successful offspring. The approach is also especially efficient running in a multiple processor, multiple selection-strategy mode using multiple settings. This paper discusses and evaluates these Genetic Algorithm techniques.

Abstract: Partition of unity parametrics (PUPs) are a recent framework designed for geometric modeling. We propose employing PUPs for procedural texture synthesis, taking advantage of the framework's guarantees of high continuity and local support. Using PUPs to interpolate among data values distributed through the plane, the problem of texture synthesis can be approached from the perspective of point placement and attribute assignment. We present several alternative mechanisms for point distribution and demonstrate how the system is able to produce a variety of distinct classes of texture, including analogs to cellular texture, Perlin noise, and progressively-variant textures.

Abstract: A procedural texture relates texel coordinates to color values through an arbitrary function, herein called a texel shader. The procedural texture is defined by a dimension, size, texel format and the texel shader. Texel coordinates are an input to the texel shader, which generates a color value for those texel coordinates. A renderer can be implemented either in hardware, such as part of a graphics processor, or in software as a computer program executed by a processor. The renderer samples from the procedural texture in response to texel coordinates, and evaluates the texel shader on demand. Filtering also can be applied automatically to results. The results of the texel shader invocations are stored in a texture cache to take advantage of spatial and temporal locality. Results are shared among threads, processes and the like through the texture cache.

Abstract: Taking advantage of the raw processing power offered by today's graphic processing units (GPUs) has become a major research topic. The amount of repeated independent computation that goes into the simulation of physical systems such as wind and snow simulations makes them great candidates for GPU parallelization. The HPC-Lab at the Norwegian University of Science and Technology (NTNU) has had several master students contributing to a real-time snow simulation. This thesis presents the work done to implement a number of real-time rendering techniques to improve the realism of this snow simulator. The main implementation of our snow simulator, which this thesis is based on, runs on the GPU using CUDA.The rendering is done in OpenGL, so the use of CUDA/OpenGL interoperability has been integral to achieve the performance needed to render the simulator output in real-time. The additional rendering techniques introduced by this thesis work are mesh texturing and lighting, triplanar texturing, scalar texture mixing, Perlin noise texture blending, shadow mapping, distance fog, skybox, billboarding and procedural texturing. Keeping the strict real-time constraint of the snow simulator has been the most important factor in choice and implementation of the rendering techniques. Our results show that scalar mixing, shadow mapping, distance fog and skybox all give significant visual improvements to the snow simulator, at a relatively low cost, keeping the frame rate above 24 frames per second (fps) for terrains of resolution 1024x1024 vertices. Triplanar texturing, however, turns out to not be that well suited for the snow simulator because of the added computational cost of doing several texture samples for each fragment, and the lack of visual improvement due to the relatively flat terrain height maps used. Procedural texturing of the snow particles using Perlin noise are tested and shown to be as fast as using image textures, with nearly 25 fps when rendering 5 million particles using a wind field with resolution 128x32x128 over a 768x768 terrain on a PC with a NVIDIA's GTX480 card. It significantly improves the realism of the rendered snowfall. Ideas for further improvements are also included.

Abstract: Procedural textures have long been a staple of off-line rendering, and impressive creativeresults havebeen accomplished by using procedural methods to their advantage. As GPU speeds and computational capabilities continue to increase, procedural texturing will likely become a useful tool also for real time rendering. In fact, it is already possible to generate procedural patterns of considerable complexity at real time frame rates on a modern GPU. Even on current (2013) low-end and mobile GPUs, the programmable shading system offers considerable processing power that often remains largely unused. Such untapped resources could be used for procedural patterns and procedural geometry computed entirely on the GPU. This article presents the state of the art in the field. Most of this is yet to be implemented in commercial projects like games, VR and visualization applications. Code for the shader examples in this article is available under permissive open source licenses or as public domain software and is collected on the address: http://www.itn.liu.se/˜stegu76/sigrad13

Abstract: This paper discusses different method of texturing techniques such as photorealistic and procedural texturing e.g. Dynamic Pulse Function (DPF). This technique can be utilized to deal with the problems of reflections, shadows, disturbing objects, leaning geometries, huge data size, and low quality of photorealistic texturing. Based on a predefined XML schema, unlimited number of layers can be created for windows, doors and background. The system can use many textures for a single layer to increase Level of Realism (LoR) in 3D virtual model. The advantage of this technique is the geospatial database of each layer in XML-schema. It can include the name of the layer, width, height, geometry and starting point of the object in the layer with respect to upper-left corner of the façade which can be queried. The concept behind DPF, is to use logical operations to project the texture on the background image which is dynamically proportional to real geometry. The process of projection is based on two vertical and horizontal dynamic pulses starting from upper-left corner of the background in down and right directions respectively based on image coordinate system. The logical one/zero on the intersections of two vertical and horizontal dynamic pulses projects/does not project the texture on the background image. The priority of the door layer is higher than window layer which means that window texture cannot be projected on the door layer. Orthogonal and rectified perpendicular symmetric photos of the 3D objects that are proportional to the real façade geometry were utilized for the generation of the output frame for DPF. Texture enhancement can be done by leftright or up-down symmetrical transformation. Higher radiometric adjustment was utilized for raised surfaces to represent the removed geometry on the 3D virtual plane surface. Disturbing objects and obstacles such as pedestrians, flying birds, etc. were removed according to the form of the façade, observation of the operator and the method of texture tessellation. The DPF produces very high quality and small size of output image compare with the photorealistic texturing method. The disadvantage of DPF is preprocessing and generating output image file rather than online processing and generating the texture on the 3D environment such as CityGML from the textures in small data sizes. Complex geometries such as coconut or palm trees can be designed as a single implicit geometry and utilized in CityGML environment via URL to deal with rendering and lagging problems of visualization. The problem of missing faces and objects as exist in CityGML has been resolved and described in this paper. Some external code lists for CityGML were defined for vegetation and trees for Malaysian Three Dimensional Spatial Data Infrastructure (M3DSDI) and approved by OGC. The new code list will be defied for other street furniture and heritage and religious landmarks in near future. In the current work the texture of the façade was created based on preprocessed procedural DPF technique and the output image was utilized in 3D modeling as a texture. We plan to further develop a graphic user interface (GUI) for the DPF package based on online processing rather than preprocessed texture generation – as the outlook of this work.

Abstract: In various embodiments, an ray-marched-tangent space shader is provided which uses adaptive, curved ray marching of an implicit weave/thread procedural texture to create the appearance of individual cloth yarns complete with sub-fibers which separate rather than stretch over the surface. The volumetric surface shader shades cloth by performing adaptive curved ray marching of an implicit tangent space distance field.

Abstract: An algorithm is introduced for rapidly synthesizing solid textures from a set of 2D example images. In contrast to previous techniques, this method restricts synthesis to a subset of the voxels, while enforcing spatial determinism. A major difficulty lies in reducing the dependency chain of neighborhood matching, so that each voxel only depends on a small number of other voxels. The key idea is to synthesize a volume from a set of 3D candidates, each being a triple of interleaved 2D neighborhoods. An efficient algorithm is presented to carefully select in a pre-process only those candidates forming consistent triples. This significantly reduces the search space during subsequent synthesis. The result is a new parallel, spatially deterministic solid texture synthesis algorithm which generates high resolution solid textures on surfaces within seconds. With a GPU implementation, this method can interactively synthesize new textures for surfaces that appear when breaking or cutting objects.

Abstract: Color map textures applied directly to surfaces, to geometric microsurface details, or to procedural functions (such as noise), are commonly used to enhance visual detail. Their simplicity and ability to mimic a wide range of realistic appearances have led to their adoption in many rendering problems. As with any textured or geometric detail, proper filtering is needed to reduce aliasing when viewed across a range of distances, but accurate and efficient color map filtering remains an open problem for several reasons: color maps are complex non-linear functions, especially when mapped through procedural noise and/or geometry-dependent functions, and the effects of perspective and masking further complicate the filtering over a pixel's footprint. We accurately solve this problem by computing and sampling from specialized filtering distributions on-the-fly, yielding very fast performance. We filter color map textures applied to (macro-scale) surfaces, as well as color maps applied according to (micro-scale) geometric details. We introduce a novel representation of a (potentially modulated) color map's distribution over pixel footprints using Gaussian statistics and, in the more complex case of high-resolution color mapped microsurface details, our filtering is view- and light-dependent, and capable of correctly handling masking and occlusion effects. Our results match ground truth and our solution is well suited to real-time applications, requires only a few lines of shader code (provided in supplemental material), is high performance, and has a negligible memory footprint.

Abstract: In this article, we present a method for rendering dynamic scenes featuring translucent procedural volumetric detail with all-frequency soft shadows being cast from objects residing inside the view frustum. Our approach is based on an approximation of physically correct shadows from distant Gaussian area light sources positioned behind the view plane, using iterative convolution. We present a theoretical and empirical analysis of this model and propose an efficient class of convolution kernels which provide high quality at interactive frame rates. Our GPU-based implementation supports arbitrary volumetric detail maps, requires no precomputation, and therefore allows for real-time modification of all rendering parameters.