Bayer Pattern

[HurtinMinorKey]

So why does everyone use it? The line i always hear is that the human eye is more sensative to green than red or blue. But is that true for everyone? What is the actual magnitude, and just how well documented is it? Are there any notable exceptions(in terms of cameras)?

[andy lee]

[url="http://www.eoshd.com/comments/topic/983-john-galt-the-truth-about-2k-4k-and-the-future-of-pixels-everyone-should-read-this/"]http://www.eoshd.com/comments/topic/983-john-galt-the-truth-about-2k-4k-and-the-future-of-pixels-everyone-should-read-this/[/url]

have a read of this thread I started last month

[HurtinMinorKey]

[quote name='andy lee' timestamp='1345656952' post='16228']
[url="http://www.eoshd.com/comments/topic/983-john-galt-the-truth-about-2k-4k-and-the-future-of-pixels-everyone-should-read-this/"]http://www.eoshd.com...ould-read-this/[/url]

have a read of this thread I started last month
[/quote]

I've read it many times. I want to know where the science of human sensitiity to different colors of light originates.

[pietz]

ok let me try to help you as good as i can or as far as my knowledge goes. your answer in one word, would be: "[b]Evolution[/b]".

if you have a 3-sensor camera (1 sensor each for red, green and blue) you essentially get 3 different color streams that you have to put back together. if you mix these colors 33,3% each, you would notice the final image looking very odd. to get a correct image you need to mix the colors like this: [b]29,9% Red, 58,7% Green and 11,4% Blue[/b]. Here we can find the reason why there are twice as many green pixels on a Bayer sensor then red and blue. Because the green color has to be boosted almost twice as much to get a natural looking picture for the human eye.

Now imagine a picture that is mixed 3x33,3%. The reds would be much stronger and the blues even more than the reds! this wasnt very practical for the first human beings 100.000 years ago. the bright blue sky blinded the shit out of them! the color that was most important back in the days was green! because they went hunting in the woods. they didnt give a shit for how bright the sky was, they wanted to get the most detail in nature. obviously this isnt something that changed from 1 day to the other, it took thousands and thousands of years for any creature to adapt to its surrounding. for humans that meant improving what their life depended on: hunting and not getting killed. i know this sounds a lot like what your parents tell you when they dont know the answer either, but its the reason i learned in filmschool and the one that makes the most sense to me as well.

same thing goes for audio. the human ear is more sensitive for certain frequencies. especially around 12kHz and everything lower than 100Hz. 12kHz applies to rustling leaves and the lows to footsteps of large animals on the ground.

i hope this helps.

[HurtinMinorKey]

^Thanks. Your information makes me ask two key questions:
1. So wavelength amplitude implies sensativity?
2. If the ratio you state is true [b] (58,7% Green,2[/b][b]9,9% Red,[/b][b] and 11,4% Blu) [/b]and a determining factor, then why not instead use a bayer-like pattern with a more representative rato like a 4:2:1 GRB ratio? Perhaps tiling on a 2D sensor grid like that is pain in the ass?

[pietz]

first of all. dont just take my word for it :)

[url="http://en.wikipedia.org/wiki/YUV"]http://en.wikipedia.org/wiki/YUV[/url]

on the wiki link youll find a section "Conversion from RGB" it describes how a YUV signal is created from the 3 RGB channel. Y is the luminance channel, the one that holds information for the brightness of an image. if you look at the calculation you can find the exact percentages that i stated above, to create the luma channel. thats where the numbers come from.

1. well the human vision adjusted to what it needed to see. thats why green is right in the middle of the human visible color spectrum and red and blue are more to the lower and upper limits. so yes, we can see different wavelengths of light (=different colors) differently good.

2. i know that this guy talks really bad about the bayer pattern but you have to understand what we achieved through this. we dont need 3 seperate sensors anymore, but just 1. OBVIOUSLY, through this reduction we have some quality loss, BUT because of the bayer pattern this quality loss is REALLY low.

[url="http://scien.stanford.edu/pages/labsite/2007/psych221/projects/07/demosaicing/bayer_cfa.JPG"]http://scien.stanfor...g/bayer_cfa.JPG[/url]

take the first 2 by 2 pixels of the linked picture. the color of the pixel relates to the color information the pixel can store. out of these 2x2 (B-G-G-R) single color pixels we can create 1 fully custom colored pixel, the result we want for our picture. we create 1 finished pixel out of 4 R, G and B pixels. thats means we need 4 times as many R, G and B pixels as the resolution we want to achieve, right? WRONG!

see, to get the next custom colored pixel we dont have to go 2 pixels to the right, but only 1! we can re-use the 2 pixels in the second column and combine them with those in the third to get the next fully colored pixel. every intersection of the black lines in the picture stands for 1 finished pixel that can be created from the pixels surrounding it.

that means if we want a picture with a resolution of 800x600Px on a bayer pattern, we need a sensor with a resolution of 801x601Px. thats pretty amazing!

i hope you were able to follow me with this.