Rui M. Bastos

Technical Section: A survey of raster-based transparency techniques

Transparency is an important effect for several graphics applications. Correct transparency rendering requires fragment-sorting, which can be more expensive than sorting geometry primitives, and can handle situations that might not be solved in geometry space, such as object interpenetrations. In this paper we survey different transparency techniques and analyze them in terms of processing time, memory consumption, and accuracy. Ideally, the perfect method computes correct transparency in real-time with low memory usage. However, achieving these goals simultaneously is still a challenging task. We describe features and trade-offs adopted by each technique, pointing out pros and cons that can be used to help with the decision of which method to use in a given situation.

Hybrid transparency

Hybrid transparency is an approach for real-time approximation of order-independent transparency. Our hybrid approach combines an accurate compositing, of a few core transparent layers, with a quick approximation, for the remaining layers. Its main advantage, the ability to operate in bounded memory without noticeable artifacts, enables its usage with high scene complexity and image resolution, which other approaches fail to handle. Hybrid transparency is suitable for highly-parallel execution, can be implemented in current GPUs and further improved, with minimal architecture changes. We present quality, memory, and performance analysis and comparisons which demonstrate that hybrid transparency is able to generate high-quality images at competitive frames rates and with the lowest memory consumption among comparable OIT techniques.

Memory-Efficient Order-Independent Transparency with Dynamic Fragment Buffer

Order-independent transparency (OIT) rendering is computationally intensive due to required sorting and sufficient memory to store fragments before sorting. We present Dynamic Fragment Buffer, a revamped two-pass OIT rendering technique, which performs correct blending of a large number of transparent layers at interactive frame rates. Our approach self-adjusts memory allocation to handle a variable number of fragments per pixel without wasting memory. In this paper we perform a detailed analysis of several design decisions which lead to this technique. We present a collection of experiments that illustrate the advantages of our technique with respect to others OIT algorithms in the literature.

Real-time multi-agent path planning on arbitrary surfaces

Path planning is an active topic in the literature, and efficient navigation over non-planar surfaces is an open research question. In this work we present a novel technique for navigation of multiple agents over arbitrary triangular domains. The proposed solution uses a fast hierarchical computation of geodesic distances over triangular meshes to allow interactive frame rates, and a GPU-based collision avoidance technique to guide individual agents. Unlike most previous work, the method imposes no limitations on the surface over which the agents are moving, and can naturally deal with non-planar meshes of arbitrary genus and curvature. Moreover, the implementation is a hybrid CPU/GPU algorithm that explores the current trend of increasing the number of CPU cores and GPU programmability. This approach exploits the best qualities in each processor, thus achieving very high performance.

Transparency and Anti-Aliasing Techniques for Real-Time Rendering

Transparency and anti-aliasing are crucial to enhance realism in computer-generated images, which have a high demand for such effects. Transparency is largely used to denote relationships among objects in a scene, and to render several structures, such as particles and foliage. Anti-aliasing (AA) is also important, since jagged edges can be easily spotted and create disruptive distractions during a scene walkthrough, which are unacceptable in real-time applications. Figure 1 illustrates both effects. In common, they have the fact that they rely on processing discrete samples from a given function, but using the samples for different purposes. In this tutorial we review state-of the-art techniques for transparency and anti-aliasing effects, their initial ideas and subsequent GPU accelerations. We support our presentation with a discussion on their strengths and limitations.

Special Section on Computer Graphics in Brazil: A selection of papers from SIBGRAPI 2012: Memory-optimized order-independent transparency with Dynamic Fragment Buffer

Exact order-independent transparency (OIT) rendering is memory demanding because it requires per-pixel blending of an unknown number of fragments that need to be stored and sorted before compositing. In this paper, we describe the Dynamic Fragment Buffer (DFB) algorithm, which efficiently manages memory to perform correct compositing for pixels with varying numbers of fragments. We present a collection of experiments that illustrate the advantages of the DFB algorithm with respect to other OIT algorithms, and analyze the impact of the proposed variations.

Approximate on‐Surface Distance Computation using Quasi‐Developable Charts

There is a vast number of applications that require distance field computation over triangular meshes. State‐of‐the‐art algorithms have quadratic or sub‐quadratic worst‐case complexity, making them impractical for interactive applications. While most of the research on this subject has been focused on reducing the computation complexity of the algorithms, in this work we propose an approximate algorithm that achieves similar results working in lower resolutions of the input meshes. The creation of lower resolution meshes is the essence of our proposal. The idea is to identify regions on the input mesh that can be unfolded into planar regions with minimal area distortion (i.e. quasi‐developable charts). Once charts are computed, their interior is re‐triangulated to reduce the number of triangles, which results in a collection of simplified charts that we call a base mesh. Due to the properties of quasi‐developable regions, we are able to compute distance fields over the base mesh instead of over the input mesh. This reduces the memory footprint and data processed for distance computations, which is the bottleneck of these algorithms. We present results that are one order of magnitude faster than current exact solutions, with low approximation errors.

Geotextures: A Multi-source Geodesic Distance Field Approach for Procedural Texturing of Complex Meshes

Texture mapping is an important technique to add visual detail to geometric models. As an alternative to traditional image-based texture mapping, procedural textures are described by a function, with interesting properties such as compact representation, resolution independency and parametric adjustment of the visual appearance. Procedural textures are usually defined in the 2D texture space, making the result dependent on texture mapping coordinates assigned to the model, or in the 3D object space, implying in no correlation with the surface model. In this work we introduce GeoTextures, an approach that uses the geodesic distance, defined from multiple sources over the model, as a parameter that is taken into account by time-varying procedural textures. The use of geodesic distances allows the process to be both independent from the mapping of texturing coordinates and also conforming with the model surface. We validate the proposal by applying real-time procedural textures in complex surfaces.

Object-Space Visibility Ordering for Point-Based and Volume Rendering

This paper presents a method to accelerate algorithms that need a correct and complete visibility ordering of their data for rendering. The technique works by pre‐sorting primitives in object‐space using three lists (one for each axis: X, Y and Z), and then combining the lists using graphics hardware by rendering each list to a texture and merging the textures in the end. We validate our algorithm by applying it to the splatting technique using several types of rendering, including point‐based rendering and volume rendering. We also detail our hardware implementation for volume rendering using point sprites.

Geodesic-driven visual effects over complex surfaces

Texture mapping is an important technique for adding visual details to geometric models. Image-based texture mapping is the most popular approach, but it relies on pre-computed images which often limit their use to static effects. For adding dynamic effects, procedural-based texturing is more adequate. Since it rely on functions to describe texturing patterns, procedural texturing allows for a more compact representation and control of visual effects by a simple change of parameters. In this work we describe GeoTextures, an approach that uses geodesic distance fields defined from multiple sources at different locations over a model surface to place, advect, and combine procedural visual effects over complex surfaces. The use of geodesics extends the scope of common procedural textures which are usually limited to using spatial 3D coordinates or 2D texture coordinates. We illustrate the flexibility of our real-time approach with a range of visual effects, such as time-based propagation of weathering phenomena, transparency effects, and mesh displacement over surfaces with smooth silhouettes using hardware based tessellation available in current graphics cards.