From: Dan Margulis
Date: Wed, Dec 22, 1999, 5:20 PM
RE: Blurring the AB channels

Folks,

The use of the AB channels in LAB continues to cause confusion. Here's a recent question I got; my reply is in a separate message.

Regards,
Dan Margulis

************

I have a few questions about some techniques I've read about and tried, but wanted to get some feedback on others experiences. One technique that seems to be used for images from digital cameras is the soften a, b, channel (gaussian blur 3 or less points or Dust & Scratches filter) and sharpen the L channel. This is said to reduce some of the rainbow artifacts common to digital cameras. Is this technique at all useful for normal scanned images? From what I can tell, it seems to reduce some noise, but I can't tell if it effects sharpness or not? A similar technique in RGB mode is to select the channel with the most noise (generally blue) and copy this to a new layer,  blur this channel and combine this with the original layer  using the "color" mode. Are there any advantages to one technique over the other for noise/grain reduction?

The second techniques involves increasing the contrast of the a, b, channel by moving the sliders from 5 to 15 for both shadows and highlights (depending on the original). The effect is some extra "punch" or increased color saturation. I'm not sure what the benifits are to using this method over the Hue/Saturation filter, or for that matter how they differ.
Comments?

Any input you have would be appreciated!


________________


From: Dan Margulis

Date: Wed, Dec 22, 1999, 6:33 PM

RE: Re: Blurring the AB channels


>>I have a few questions about some techniques I've read about and tried, but  wanted to get some feedback on others experiences. One technique that seems to be used for images from digital cameras is the soften a, b, channel (gaussian blur 3 or less points or dust & Scratches filter) and sharpen the L channel. This is said to reduce some of the rainbow artifacts common to digital cameras. Is this technique at all useful for normal scanned images? From what I can tell, it seems to reduce some noise, but I can't tell if it effects sharpness or not?>>


The technique works for any noisy image. Gray noise is always preferable to colored noise.


The A and B channels do not contain contrast information and thus do not affect sharpness. The only drawback of blurring them is that areas of sharp contrast in colors may get a little fuzzy. If you have small colored objects you have to watch out for this problem.


>>A similar technique in RGB mode is to select the channel with the most noise (generally blue) and copy this to a new layer,  blur this channel and combine this with the original layer  using the "color" mode. Are there any advantages to one technique over the other for noise/grain reduction?>>


The method you describe of blurring the Blue in color mode is also

effective. There are some mild technical arguments in favor of doing it in LAB (the B channel of LAB actually has some impact on red and green, as well as blue; plus normally the A channel has less noise than the B, so you can customize the blur with two different settings.)


I would think, however, that for all practical purposes the results would be the same.


>>The second techniques involves increasing the contrast of the a, b,

channel by moving the sliders from 5 to 15 for both shadows and highlights

(depending on the original). The effect is some extra "punch" or increased

color saturation. I'm not sure what the benifits are to using this method

over the Hue/Saturation filter, or for that matter how they differ.

Comments?>>


The A-B move is much superior and will yield more natural-looking results. Hue/Saturation works based on what color is dominant; A-B works on whether the magenta-green and yellow-blue are in balance. If you decide that an image is lacking in blue, and steepen the B curve, blues will indeed get more intense. However, an object that is wine-red will also move toward purple (more blue) which will seem more natural in context.


If you did this instead by increasing saturation in Blues, nothing would

happen to the wine-red color because it is not a  blue. And if you did it

by increasing saturation generally, the wine-red would move toward

fire-engine red, just the opposite of what we would like.


Similarly A-B moves can create valuable variation in colors that can't be addressed with Hue/Saturation. For example, imagine a tropical ocean. The water, on the average, is mostly cyan. However, the deeper parts will be more blue, and the shallower parts where the sunlight hits the water will be more green, even though both will still technically be of a cyan color. A move in the B channel will exaggerate the difference between the two. In Hue/Saturation this won't work; the command can't move some cyans in the blue direction and others in the green.


Dan Margulis



Adobe Photoshop training classes are taught in the US by Sterling Ledet & Associates, Inc.