Nefertiti's Family

This entire discourse on this topic has led me to a question for Tom and the forum.


Can a set of consistent procedures be developed for the treatment of all images? I am thinking peer review in the case of some arguments and would like to see some type of standardization for the handling of these images by the various photo imaging software programs used.

Mark Vitrone

rd writes: "This is the same as previous posts, with one exception. I ran "Histogram Equalize" on it."

Based on the definition Tom quotes in the above post and on my previous understanding, these new enhancements Rich has introduced in this post; "noise reduction," and now "Histogram Equalize," seem to me to be valid enhancment procedures and should be in the same category as "contrast and brightness" adjustments. They don't add any new data into the raw image; they just make the existing data more discernable to the human eye. As my son Tom just told me, this new method makes the light shades lighter and the dark shades darker at the same time.

The "Woman" in the above example has been improved markedly, so has Rinny. But I still can't tell if "Piccard" has a big hook nose, or a "Dick Tracy" nose like I drew in my key.

Rich and I are working on a forever-work-in-progress paper on this subject, and I am inclined to revise it yet again to incorperate Histogram Equalize into it. What do you say Rich?

Neil

<blockquote id="quote"><font size="2" face="Verdana, Arial, Helvetica" id="quote">quote:<hr height="1" noshade id="quote"><i>Originally posted by tvanflandern</i>
yet Photoshop's description makes it sound like a natural thing to do for all the images. <hr height="1" noshade id="quote"></blockquote id="quote"></font id="quote">

(this message has been edited by me, to reflect a better understanding of the way Histogram Equalize appears to work - 4/15/2006)

Tom, I'm glad you posted that definition, because Corel's Paint Shop Pro doesn't spell it out that clearly, although you can see what it's doing to the histogram.

I agree that this seems like a "natural thing" to do, and falls under the category of "massaging" the data, and not adding or subracting data. At least, I think. It doesn't add any "count" to any of the grayscale bins (with the exception of the zero bin), it just spreads them out. It changes the output. For instance, suppose in the original image there were 2100 pixels with grayscale 181. After the Histogram Equalize operation, there is still a bin with 2100 pixels in it, but now it represents the grayscale 221. And so on. Here's what it looks like before and after:

{Image deleted temporarily} Before_Hist.jpg

{Image deleted temporarily} Family_AfterHist.jpg

If you compare the histogram in the "after" to the histogram in the before, you can see how there are now spaces between the grayscales that have content, as opposed to being all packed together, and there appears to have been some slight amount of averaging.
It's really sort of like a contrast adjustment, but like Neil's son said, it's moving the lower bins lower, the higher bins higher, and spreading the midtone bins. Also, as opposed to contrast , it appears to be operating at the bin level instead of at the pixel level, and contrast adjustment just moves the higher values higher (not sure what the threshold is).

To me the questions this raises are:

Why should this make it more lifelike?
Is it a self-fulling type of thing? In other words "men" created this notion of a histogram, and then wrote scripts that spread them out across the full grayscale. Did they do it because they knew it would make it look more lifelike? Or did they discover that doing that made it look more lifelike?
Does it make it look more lifelike because it is lifelike? Or because spreading the grayscales accross the entire spectrum on a "sort of lifelike image" makes it look more lifelike?

I really have no idea, but I might try to ask some developers who might know.

rd

<blockquote id="quote"><font size="2" face="Verdana, Arial, Helvetica" id="quote">quote:<hr height="1" noshade id="quote"><i>Originally posted by MarkVitrone</i>
<br />Can a set of consistent procedures be developed for the treatment of all images? <hr height="1" noshade id="quote"></blockquote id="quote"></font id="quote">

Mark, yes I can certainly see why you would start to think along those lines.

Personally, I don't think I know enough to answer that yet. For instance, what if the image was already congested in the low and high grayscales, would doing Histogram Equalize have a beneficial effect? I don't know.

Also, I know there are many other things that can be done, but I don't remember enough about them, yet. I'm just hacking away, trying to remember some of this stuff. The last time I worked on anything like this was in the '87 to '89 timeframe, and I haven't thought about it this much since. For instance, there's a "custom filter" that opens the door to all kinds of manipulations on adjacent pixels. I recognize it, but it will take some reading to figure out.

At this point though, I think we're safe with smoothing (i.e., averaging of adjacent pixel groups), noise reduction (not exactly sure what Paint Shop's "one step noise removal" algorithm is, and I don't know if they would tell me exactly), and any of the choices that come under the Contrast and Brightness menus, which include these Histogram Adjustments, and I would say that anything under the "Effects" menu is out of bounds for anything but photography.

rd

<i>rderosa writes</i>:<blockquote id="quote"><font size="2" face="Verdana, Arial, Helvetica" id="quote">quote:<hr height="1" noshade id="quote">Why should this make it more lifelike?
Is it a self-fulling type of thing? In other words "men" created this notion of a histogram, and then wrote scripts that spread them out across the full grayscale. Did they do it because they knew it would make it look more lifelike? Or did they discover that doing that made it look more lifelike?
Does it make it look more lifelike because it is lifelike? Or because spreading the grayscales accross the entire spectrum on a "sort of lifelike image" makes it look more lifelike?
<hr height="1" noshade id="quote"></blockquote id="quote"></font id="quote">
I don't think it's that complicated. I think this is just a more sophisicated way of increasing contrast between the light and dark parts of an image (though I'm not sure what happens technically in the old style contrast adjustment). If you notice, the Man in the example above is not helped much with the histogram adjustment, but as I said, the Woman is helped considerably. Why? Because she was already outlined with relatively dark lines, and she already had dark hair. This new process made it much more so, and hence easier to see. I'd say that is a good thing. But does it make them more "lifelike"? I don't know. They're just easier to see. The Man wasn't helped as much by this process because he never had much of a dark outline, but you can seem to see his "temple," "cheekbone," and even his "eyebrow" better now. I'd say that is an improvement.

Another interesting outcome; the imperfections and blotches in the image are brought out into sharper relief now too.

Neil

<blockquote id="quote"><font size="2" face="Verdana, Arial, Helvetica" id="quote">quote:<hr height="1" noshade id="quote"><i>Originally posted by neilderosa</i>
<br />I don't think it's that complicated. I think this is just a more sophisicated way of increasing contrast between the light and dark parts of an image (though I'm not sure what happens technically in the old style contrast adjustment). <hr height="1" noshade id="quote"></blockquote id="quote"></font id="quote">
I have trouble telling if there's a "greater truth" associated with any of this manipulating of the data, but that's me.

Here's the formula for "contrast": (Lmax - Lmin) / (Lmax + Lmin)

I think what happens is the signal is increased in such a way as to make the parts of the image with higher reflectance (luminance) more reflective, so that bright pixels are brighter still, without affecting the parts of the image with lower luminance. Not sure about the technical details of "signal increased" though.

Another major difference between this Histogram Equalize and regular contrast adjustment is the fact that the midtones can be lowered significantly.
rd