Color information in images?

Well, I've stumbled across something on accident... and I'm sure there's
a method that takes advantage of this; however, to recreate my steps...
do the following (assuming you have Adobe Photoshop):

1] Open a colored image.
2] Select and copy the colored image, and paste it into a new window
within Photoshop (so now you have two images open).
3] Convert one of the images to greyscale, then convert back to RGB
mode.  This will result in one of the images in 256 greyscale (but you
need to put the mode back to RGB afterwards, even though the color won't
reappear).
4] On the color image... resize to about 1/10 the size.  For example, if
you have a 100x100 images... resize to 10x10.
5] Resize the color image again... back to the original 100x100 (or your
dimensions) size.  You'll notice it's REALLY blurred.  That's alright!
6] Select and copy the greyscaled (in RGB) image and paste it over the
blurred colored image.
7] Right-click on the greyscaled layer and select "Blending Options" to
open the "Layer Style" window.
8] Right on this page, you'll see a box labeled "General Blending" at
the upper right section.  Take the "Opacity" slider and adjust it to
about 75%.  Now, click "Okay".
9] Flatten the image by going to the Layer menu option in the Adobe
Photoshop menu.
10] Press Ctrl+U to bring up the "Hue/Saturation" box.  Now, set the
saturation to 75 (or whatever you adjusted the "Opacity" slider value to).

If you did this right, then you should have an image that appears
similar (lossy) to the original colored-image (except that you recreated
it using only a greyscaled version, and a SEVERELY reduced colored
version).

So, with this in mind... is it accurate to assume that the greyscale
acts as a luminance map to each pixel, and since the human eye is more
susceptible to this, you can remove alot of the colored information
(which is what we did by resizing the colored layer to 1/10 the size and
resizing it back)?

This is pretty much how jpeg , or colour tv works. The eye
is a lot less sensitive to detail in chrominance than in luminence.

"Eric D. Brown" <nospam@nowhere.com> wrote in message
news:10n3178n0b5uid7@corp.supernews.com...
> Well, I've stumbled across something on accident... and I'm sure there's
> a method that takes advantage of this; however, to recreate my steps...
> do the following (assuming you have Adobe Photoshop):
>
> 1] Open a colored image.
> 2] Select and copy the colored image, and paste it into a new window
> within Photoshop (so now you have two images open).
> 3] Convert one of the images to greyscale, then convert back to RGB
> mode.  This will result in one of the images in 256 greyscale (but you
> need to put the mode back to RGB afterwards, even though the color won't
> reappear).
> 4] On the color image... resize to about 1/10 the size.  For example, if
> you have a 100x100 images... resize to 10x10.
> 5] Resize the color image again... back to the original 100x100 (or your
> dimensions) size.  You'll notice it's REALLY blurred.  That's alright!
> 6] Select and copy the greyscaled (in RGB) image and paste it over the
> blurred colored image.
> 7] Right-click on the greyscaled layer and select "Blending Options" to
> open the "Layer Style" window.
> 8] Right on this page, you'll see a box labeled "General Blending" at
> the upper right section.  Take the "Opacity" slider and adjust it to
> about 75%.  Now, click "Okay".
> 9] Flatten the image by going to the Layer menu option in the Adobe
> Photoshop menu.
> 10] Press Ctrl+U to bring up the "Hue/Saturation" box.  Now, set the
> saturation to 75 (or whatever you adjusted the "Opacity" slider value to).
>
> If you did this right, then you should have an image that appears
> similar (lossy) to the original colored-image (except that you recreated
> it using only a greyscaled version, and a SEVERELY reduced colored
> version).
>
> So, with this in mind... is it accurate to assume that the greyscale
> acts as a luminance map to each pixel, and since the human eye is more
> susceptible to this, you can remove alot of the colored information
> (which is what we did by resizing the colored layer to 1/10 the size and
> resizing it back)?

Hi,

> Well, I've stumbled across something on accident... and I'm sure there's
> a method that takes advantage of this; however, to recreate my steps...
> do the following (assuming you have Adobe Photoshop):

/* snip */

> So, with this in mind... is it accurate to assume that the greyscale
> acts as a luminance map to each pixel, and since the human eye is more
> susceptible to this, you can remove alot of the colored information
> (which is what we did by resizing the colored layer to 1/10 the size and
> resizing it back)?

This is not at all an surprise. In fact, color subsampling is a frequent
lossy image compression technique. Consider color TV as the most prominent
example.

So long,
Thomas