nm
6th February 2006, 14:36
It would be best to get someone more experienced in metric testing to figure out a good test. I can't think of any good way to include resizing in the PSNR calculations where the values are supposed to be compared with PSNR values from a non-resized encode.
Perhaps visual comparison with a low resolution (< 320x240) and very low bitrate (100-200 kbps) would be better. Still, resizing would be hard to do without affecting the judgement because of scale differences. You could do the resize twice, first to a higher resolution and then back to the original, but that wouldn't be fair.
Perhaps visual comparison with a low resolution (< 320x240) and very low bitrate (100-200 kbps) would be better. Still, resizing would be hard to do without affecting the judgement because of scale differences. You could do the resize twice, first to a higher resolution and then back to the original, but that wouldn't be fair.