| Summary: | 碩士 === 國立清華大學 === 資訊工程學系所 === 105 === We propose a new cloud gaming platform to address the limitations of the existing ones. We study the rendering pipeline of 2D planar maps, and convert it into the server and client pipelines. While doing so naturally gives us a distributed rendering platform, which compresses 2D planar maps for transmission has never been studied in the literature. In this thesis, we pro- pose a compression component for 2D planar maps with several parametrized modules, where the optimal parameters are identified through real experi- ments. The resulting cloud gaming platform is evaluated through extensive experiments with diverse game scenes. The evaluation results are promising, compared to the state-of-the-art x265 codec, our platform: (i) achieves bet- ter perceptual video quality, by up to 0.14 in SSIM, (ii) runs fast, where the client pipeline takes ≤ 0.83 ms to render each frame, and (iii) scales well for ultra-high-resolution displays, as we observe no bitrate increase when mov- ing from 720p to 1080p, 2K, and 4K displays. The study can be extended in several directions, e.g., we plan to leverage the temporal redundancy of the 2D planar maps, for even better performance.
|
|---|
