Ing. Jaroslav Sloup

This work presents a Vulkan-based implementation rendering volumetric clouds and atmosphere. We combine previously published solutions to produce a single unified look. We use Raymarching as the main method to render both the atmosphere and clouds. Furthermore, we use multiple precomputed look-up tables (LUTs) proposed by Hillaire to speed up the rendering of the atmosphere. We enhance these methods with the option to render volumetric clouds using a precomputed three-dimensional texture setup storing procedurally generated noise. With our final solution, we can render images in a high dynamic range. We apply post-processing effects and use adaptive luminance to transform the image into a low dynamic range for presentation.

Ray tracing of trimmed non-uniform rational basis spline (NURBS) surfaces is a must in the CAD industry, and has been widely applied in 3D modeling software. While most CAD software uses tessellation followed by rasterization to display surfaces, we show the performance of the 3D models by ray tracing can reach real-time performance even for large-scale models. We show the performance of ray tracing can reach from 50 to 1,000 mega rays per second for primary and random rays on the contemporary GPU RTX 3090, depending on the scene complexity between 10K to 1800K NURBS surfaces and 37K to 10,000K of trimming NURBS curves. We utilize the computing power of the GPU and new algorithmic ideas to accelerate the trimming in parametric space. The achieved performance of the ray tracing engine can be used for non-trivial rendering of high-quality images, including the global illumination with interactive and real-time rates.

The representation of geometric models by trimmed NURBS surfaces has become a standard in the CAD industry. In CAD applications, the rendering of surfaces is usually solved by tessellation followed up by z-buffer rendering. Ray tracing of NURBS surfaces has not been widely used in industry due to its computational complexity that hinders achieving real-time performance in practice. We propose novel methods achieving faster point location search needed by trimming in the context of ray tracing trimmed NURBS surfaces. The proposed 2D data structure based on kd-trees allows for faster ray tracing while it requires less memory for its representation and less preprocessing time than previously published methods. Further, we show the current state of the art for ray tracing trimmed NURBS surfaces on a GPU. With careful design and implementation, the number of rays cast on a GPU may reach real-time performance in the order of tens to hundreds of million rays per second for moderately to large complex scenes containing hundreds of thousands of NURBS surfaces and trimming curves.

We propose an extension to the state-of-the-art text rendering technique based on sampling a 2D signed distance field from a texture. This extension significantly improves the visual quality of sharp corners, which is the most problematic feature to reproduce for the original technique. We achieve this by using a combination of multiple distance fields in conjunction, which together provide a more thorough representation of the given glyph's (or any other 2D shape's) geometry. This multi-channel distance field representation is described along with its application in shader-based rendering. The rendering process itself remains very simple and efficient, and is fully compatible with previous monochrome distance fields. The introduced method of multi-channel distance field construction requires a vector representation of the input shape. A comparative measurement of rendering quality shows that the error in the output image can be reduced by up to several orders of magnitude.

This paper is about interest point detection and GPU programming. We take a popular GPGPU implementation of SIFT - the de-facto standard in fast interest point detectors - SiftGPU and implement modifications that according to recent research result in better performance in terms of repeatability of the detected points. The interest points found at local extrema of the Difference of Gaussians (DoG) function in the original SIFT are replaced by the local extrema of determinant of Hessian matrix of the intensity function. Experimentally we show that the GPU implementation of Hessian-based detector (i) surpasses in repeatability the original DoG-based implementation, (ii) gives result very close to those of a reference CPU implementation, and (iii) is significantly faster than the CPU implementation. We show what speedup is achieved for different image sizes and provide analysis of computational cost of individual steps of the algorithm. The source code is publicly available.

We present a simple method for visual simulation of refraction phenomena in the Earth's atmosphere. The method divides the atmosphere into a set of spherical layers and the rays are approximated by circular arcs within these layers. The curvature of arcs is related to the variations of the refractive index. The path of the ray through the atmosphere is found by intersecting the circular arcs with the layer boundaries.