The Design Of The Real-time Image Anti-aliasing Algorithm Based On Gpu

Anti-aliasing is a very important core operator in digital image processing field, which is vastly adopted in image enhancement, image reconstruction, digital image synthesis, digital photography and graphics transmission over internet. In signal processing discipline, aliasing phenomenon refers to an effect that causes different signals to become indistinguishable when sampled and converted to digital format. Anti-aliasing technology generally includes several operations, analyzing the characteristic of image, locating the aliasing parts of image, improving the quality of image by removing the aliasing partially or even completely, reserving the primary characteristic of image.In this thesis, we first introduce the application and developments of digital image synthesis field based on the computer graphics theory, and then indicate that computer graphics has been vastly used in many fields, as the emergence of enormously powerful Graphics Processor Unit (GPU] and the maturation of interactive graphics rendering theory. The aliasing phenomenon of raster graphics has been the one of most dominant issues which affects the quality of digital image.Based on the background of topic just mentioned, we briefly review the basic theory and fundamental concepts about aliasing, including signal processing and filter theory, computer graphics and graphics rasterization. Then we analyze the hardware-based anti-aliasing algorithms, GPU-based anti-aliasing algorithms which utilize the programmability of GPU, and image super-resolution algorithms which inspire our works.After all the analysis and research above, we present a new anti-aliasing algorithm based on the geometric pattern of pixels. This algorithm scans the geometric pattern of potential aliased pixels, filter the pixels locally as pre-defined rules, ease the aliasing phenomenon of digital image. We specially consider the applicability of this algorithm in real interactive graphics application, and make good balance between performance and quality. The experiments show that we can gain the performance boost about4x-10x.