Real time volumetric rendering optimization approach for cloudscapes in virtual environment

• Main Author: Zamri, Muhamad Najib
• Format: Thesis
• Language: English
• Published: 2022
• Subjects: QA75 Electronic computers. Computer science
• Online Access: http://eprints.utm.my/id/eprint/101544/1/MuhamadNajibZamriPSC2022.pdf.pdf; http://eprints.utm.my/id/eprint/101544/; http://dms.library.utm.my:8080/vital/access/manager/Repository/vital:150615

There is a great demand for visualizing natural atmospheric clouds in a realtime environment. However, the challenge is maintaining the cloud visualization system performance due to the high computational cost, high sampling cost, intensive main memory consumption, and the deterioration of system performance. Therefore, this research aimed to optimize the rendering performance of a real-time atmospheric cloudscape visualization system via a set of efficient methods considering the volumetric cloud data as the input, the fly-through navigation as the user interaction, and the realistic real-time features as the output of the generated image. First, the volumetric-based path tracing method was enhanced by reducing the number of ray castings during the intersection test and determining the new pixels via the frame-toframe coherence mechanism. Second, the multiple light scattering method was improved by limiting the maximum number of sampling steps during the light bouncing procedure when calculating the lighting effects in the virtual environment. Third, the new hybrid visibility acceleration method was proposed. It took into account the view-dependency and multi-resolution cloud level of detail, focusing on distancebased view volume segments and the optimized view frustum culling method to remove the invisible regions of clouds in a real-time environment. Based on the experiment conducted, the proposed approach in the prototype system was capable of achieving about 3.83 to 18.58 times faster computational time. This showed an increase of approximately 2.53 to 19.21 times in rendering speed and consumed about 20.63% of the total computer’s main memory space compared to previous systems. Furthermore, all the domain experts have reviewed and verified that the prototype system has effectively produced high visual realism of the generated images. Based on the results of this study, an improved real-time volumetric cloudscape rendering optimization approach has been successfully developed to address the inability of realtime rendering used in previous methods.