Hacker News Reader

Umm... I feel like there's a huge flaw in this. He's looking at the different _channels_ in this image, but couldn't that mean there's just less blue data in that particular image?

I believe that humans may be worse at seeing blue, but this test seems bogus. By looking at the channels, he's actually cutting out different subsections of data, so it's apples to oranges, right? What if there's just much less blue in this particular photo?

Or am I mistaken?

View on HN

No, You're spot on.

The author states that this is "an unmodified picture, straight off the digicam". Unfortunately for his test, this means the light passed through a Bayer filter, which uses a pattern of "50% green, 25% red and 25% blue": http://en.wikipedia.org/wiki/Bayer_filter

It could, but you can verify that our eyes cannot detect loss of blue information even in the parts of the image that are rather blue (her wristwatch, the car in the background, the edge of the windshield, the stripe on the towel).

In each of those cases fidelity is more damaged by loss of quality in the other

Reply on news.ycombinator.com