3 ms·
Looking at the result and judging is actually the best and most important metric. There are numerical metrics, like SSIM, but they are always worse than human c
by TD-Linux 12y ago
Looking at the result and judging is actually the best and most important metric. There are numerical metrics, like SSIM, but they are always worse than human comparison trials. You're optimizing for what looks best to a human in the end, after all.
- pornel 12y agoIt's very hard to create a fair test. The problem with human judgement that it's very imprecise. You're unlikely to notice difference between JPEG at Q=90 and Q=95, but you can't say it doesn't matter, because that can cause 40% difference in file size. OTOH objective metrics can easily spot that. "Looks the same to me" leaves a lot of room for error and you could be unfairly telling one format to save much more detail than the other. And almost by definition these are the details you're least likely to be able to see. There's also a pitfall of choosing image that looks subjectively "nicer" rather than closer to the original. Codecs that are better (e.g. faithfully preserve noise) may be judged as worse than another codec that's an accidental Instagram filter.