As far as I remember, chroma subsampling is lossy. The idea is that we're less sensitive to red and blue light and so rather than sampling every pixel for it's chroma value, an average is taken, and when the average is displayed over adjacent pixels we will tend not to notice the error (as long as the number of pixels averaged is small of course...) But there's no way to correctly get back what has been lost. I guess you can make a good guess, and with certain images, eg test patterns, maybe even get it right, but you are also going to get it wrong as well, particularly in complex images (ie non test patterns) and so I'd personally rather not interpolate at all and simply display what has actually been sampled as that will more often that not be the more accurate image.

I'd love to proven wrong.

Mark

My Music : My Cinema