Contrast masking

[Don]

I just discovered digital contrast masking which I use to extend the
dynamic range of the scanner by scanning twice - once for highlights
and once for lowlights - and then combining the best of both images.

The instructions on how to do this say I should convert the bottom
(highlight) image into a layer mask, add some Gaussian blur to avoid
potential artifacts, and then apply it to the top (lowlight) image.

Wouldn't a better result be achieved to create a layer mask from
*both* images, average it out and then apply that to the top image
without bothering with Gaussian blur?

Yeah, yeah, I know... Try it out and have a look... ;o) But I'm
learning and would like to know the theoretical justification for one
or the other.

Don.

 

[Raphael Bustin]

I just discovered digital contrast masking which I use to extend the
dynamic range of the scanner by scanning twice - once for highlights
and once for lowlights - and then combining the best of both images.

The instructions on how to do this say I should convert the bottom
(highlight) image into a layer mask, add some Gaussian blur to avoid
potential artifacts, and then apply it to the top (lowlight) image.

Wouldn't a better result be achieved to create a layer mask from
*both* images, average it out and then apply that to the top image
without bothering with Gaussian blur?

Yeah, yeah, I know... Try it out and have a look... ;o) But I'm
learning and would like to know the theoretical justification for one
or the other.

Mask making is an art. There is no "proper"
way to make a mask for this purpose or any
other -- it depends in large part on the image,
your skill, your tools, and the desired effect.

Often when masks are used to effect large
differences in gamma or contrast, there will
be artifacts along the edge between the two
regions. Those artifacts will make the image
look unreal or surreal.

That's why blur is often used to soften the
edges of the masks.

For contrast masking, I often use a simple
gradient mask where the gradient extends
over a significant portion of the frame.


rafe b.
http://www.terrapinphoto.com

 

[Don]

Often when masks are used to effect large
differences in gamma or contrast, there will
be artifacts along the edge between the two
regions. Those artifacts will make the image
look unreal or surreal.

That's why blur is often used to soften the
edges of the masks.

Indeed, that's the reason given in various web pages I found on the
subject.

However, what I was getting at is that, combining masks from both
images would create an "in-between" mask which would be equally "off"
in regards to both images. So, not only in Gaussian blur not
necessary, but such a mask would be much more accurate (being based on
both images rather than only one).

I was just wondering if this train of thought is on the right track...

For contrast masking, I often use a simple
gradient mask where the gradient extends
over a significant portion of the frame.

Does that mean you don't convert one of the images to grayscale but
simply "paint" a large swathe of gradient in the areas of interest,
right?

Don.

 

[Raphael Bustin]

Indeed, that's the reason given in various web pages I found on the
subject.

However, what I was getting at is that, combining masks from both
images would create an "in-between" mask which would be equally "off"
in regards to both images. So, not only in Gaussian blur not
necessary, but such a mask would be much more accurate (being based on
both images rather than only one).

I was just wondering if this train of thought is on the right track...


Does that mean you don't convert one of the images to grayscale but
simply "paint" a large swathe of gradient in the areas of interest,
right?

Exactly. The Photoshop equivalent of a graduated
neutral-density filter, if you will. The advantage of course
is... no edge artifacts, ever.


rafe b.
http://www.terrapinphoto.com