レンダリング品質予測モデルに基づく高品質ノーマルマップ圧縮

概要

論文の詳細を見る
Normal maps play an important role in realistic 3D image rendering. They are used to express pseudo roughness of a 3D object's surface using a small amount of polygon data. A fast and efficient normal map compression algorithm is presented in this paper that uses vector quantization, which is based on a quality estimation model of the rendered images. An optimal encoding strategy is mathematically obtained from the quality estimation model. Our experimental results demonstrated that this algorithm produces better results than the other previously reported approaches, especially in the higher PSNR regions. As a result, it is possible to encode normal maps by taking into consideration the quality of the resultant rendered images without actually rendering the images. A reasonable encoding time is achieved by taking into consideration the limited signal space of normal maps and by employing an inner product in the nearest vector search. In addition, the decoding time is very small, which is also important for practical usage.
社団法人映像情報メディア学会の論文
2007-08-01