debayer test

[Nils Ruinet]

Just looking at the pics in Mac's Preview at 100% (1:1) magnification (witout touching anything else) on my 17" Macbook Pro screen, based on the noise characteristics and sharpness, I'd say :
Cyan = Full
Yellow = half
Magenta = Quarter

But I have to admit the differences are smaller than I would have expected.
I guess the difference in noise level becomes more apparent when played back at 24 fps.

 

[Kobe Van Steenberghe]

Magenta 1/4
Yellow 1/2
Cyan full

 

[Frank Cueto]

How about the actual facts? its been 2-3 days!

 

[Andrae Palmer]

Ok so we've had 35 people respond to this test:

Twelve thought it was:
Yellow - Full Res. Premium
Cyan - Half Res. Good
Magenta - Quarter Res. Good


One person thought it was:
Yellow - Full Res. Premium
Cyan - Half Res. Good
Magenta - Quarter Res. Good

One person thought it was:
Yellow - Full Res. Premium
Magenta - Half Res. Good
Cyan - Quarter Res. Good

Twenty one thought it was:
Cyan - Full Res. Premium
Yellow - Half Res. Good
Magenta - Quarter Res. Good

----------------------------------------------------------------

The True Answer is:
Cyan - Full Res. Premium
Yellow - Half Res. Good
Magenta - Quarter Res. Good

 

[luigivaltulini]

Thanks Andrae for testing


Luigi

 

[Andrae Palmer]

Before this test it was difficult for me to see the aliasing in the image. After a few posted about the aliasing in the bokeh area of the image... I could clearly see it on the "Yellow". I think as colorist we have to be keenly aware of how debayering affects the image. A few also noticed the difference in brightness between the images.

Although many had a problem telling the difference and I believe even this test some were influenced by the answers of others. A true test would be one where the taker didn't see the previous answers. In any case there is a "noticeable" difference in quality to the trained eye. You really have to make a decision over the sharpness or noise/aliasing.

 

[luigivaltulini]

certainly agree with you.

 

[Chuck Z]

Can you make poll only threads? That way, people can vote without any outside influence. After voting is over, you can start a discussion thread.

 

[Andrae Palmer]

Yes a poll with no comments would be best. I just don't know if the polling feature on here is sufficient. Is it possible to lock a thread until a poll is finished?

 

[Matt W.]

This really shouldn't be so surprising to anyone. Take a high-res image from a dSLR shot at 100ISO or so. Scale it down to web size using bicubic and then using nearest neighbor. Nearest neighbor looks a lot sharper.

This does kind of put the insane sharpness of medium format digital backs in perspective, though. The resolution isn't entirely "real." Long live 4x5!