Your eyes suck at blue (neat image compression trick)

http://nfg.2y.net/games/ntsc/visual.shtm

by yummyfajitas • 17 years ago

101 25 17 years ago

nfg.2y.net

ilamont• 17 years ago

The page is not loading some images, and loads others very slowly, which makes it difficult to see the phenomenon described in the post.

Might want to bookmark this and check it out later, after the HN traffic surge has died down.

thamer• 17 years ago

Cached version: http://nfg.2y.net.nyud.net:8090/games/ntsc/visual.shtm

rfurmani• 17 years ago

That makes a lot of sense. When we lived in the wild discerning different shades of green would have been crucial. It is also known that women can see more shades of red, probably in order to be able to more effectively gather edible berries. Now blue, who needs it? ;-)

zain• 17 years ago

Blueberries?

a-priori• 17 years ago

Blueberries are native only to North America, so our hominid ancestors would not have encountered them.

http://en.wikipedia.org/wiki/Blueberry

Avshalom3• 17 years ago

No, sorry that's just nonsense.

We're less sensitive to blue light because there is less blue light.

The black body spectrum of the sun peaks at green (which is why plants are green) and tails of into red. Blue light is not only a proportionately small section of the energy output but blue photons have more energy per particle and so there are even fewer photons necessary to make up that section.

That said while I've heard the women have better red resolution that doesn't necessarily mean they're more sensitive to the color number one. and number two explanations like "so they could gather berries" tend to come from a desire to see women barefoot and pregnant rather than solid science.

tesseract• 17 years ago

> The black body spectrum of the sun peaks at green (which is why plants are green)

Plants are green because they reflect green light, i.e. they do not absorb it and use it to facilitate photosynthesis. Chlorophylls tend to have absorption peaks in the red and blue, which are (in the grand scheme of the electromagnetic spectrum) near the sun's blackbody peak but do not coincide with it.

teuobk• 17 years ago

Although the black-body spectrum of the sun does peak at green, that's not why plants are green. Consider this: they appear green to us because they are not absorbing the green light.

So, then, why green? For that matter, why not black? Apparently, there's some speculation that early microbes were purple, which meant that they absorbed the green light. Chlorophyll might have appeared only in later microbes, so it would have needed to absorb what was left; namely, red and blue light.

philwelch3• 17 years ago

The evidence does seem to suggest that in hunting-gathering communities, women engaged in more gathering and men engaged in more hunting. Claims that otherwise reasonable hypotheses "come from a desire to see women barefoot and pregnant" tend to come from a desire to justify the unquestioned a priori belief that men and women are mostly identical rather than solid science.

indiv• 17 years ago

In more detail: http://en.wikipedia.org/wiki/Chroma_subsampling

lbrandy• 17 years ago

The inherent assumption on which chroma sub-sampling is founded is that luminance information is far more important than chromanance information. This is backed up by empirical evidence. From there, the effect follows directly from the equation for luminance:

Y = 0.2126 R + 0.7152 G + 0.0722 B

From here you can see that green is by far the most important color, followed by red, and then finally, way down there, blue.

ijnuhv• 17 years ago

Wow - green, the middle of the visual spectrum matters most. If only they had discovered this years ago there could have been some really good systems for sending color information by just doing intensity-red and intensity-blue. Or Y-U + Y-V = YUV if you prefer.

davepeck• 17 years ago

Interesting. Until reading this, I never stopped to wonder why OpenGL, for example, has GL_UNSIGNED_SHORT_5_6_5 (5 bits for red and blue; six for green) as opposed to some other allocation of 16 bits.

jobenjo• 17 years ago

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?

tel• 17 years ago

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

mdd• 17 years ago

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

Aron• 17 years ago

I would have started by reducing the bit depth of blue, rather than pixel width.

Daniel_Newby• 17 years ago

Ah, but the blue receptors in your eye are as accurate as the green ones, just less tightly packed. Spatial filtering therefore preserves more-visible color accuracy at the expense of less-visible blue location.

Not that that stops all those standards that use reduced bit depth for red and blue.

Your eyes suck at blue (neat image compression trick)

Related Stories