The method they use for vectorization is cool (vectorizing Y, U and V independently) and also illustrates why subsampling UV yields such good results (for stand

The method they use for vectorization is cool (vectorizing Y, U and V independently) and also illustrates why subsampling UV yields such good results (for standard raster-based codecs). Ultimately however vectors are far more expensive to rasterize than bitmaps, especially if you want antialiasing along edges. There was a bunch of noise a while back about NVIDIA's path rendering extension -- they do antialiasing via 16x multisampling (!!!) NOT a mobile friendly approach.