When it comes to b/w conversions, it's all subjective anarkhya; still, I see, thought extremely subtle, a difference. To eaches own. Pseudogrey's been around for quite a long time and discovered it a few years back. Even played with perl script that converted 16-bit grayscale files to 8-bit pseudogrey; it was pretty cool at that time but required Photoshop 7 (which I had at that time; lol) to convert the 16-bit grey to RAW format for the 8-bit RAW Pseudogrey. lol

Anyway, what I see is like the HDTV conversions of standard NSTC signals to something very crisp and clean.

Pat is now hosting this cool Script-fu at the Registry (official site now). He also made it better too.

http://registry.gimp.org/node/26515

Lyle, did you just eluded most of my questions ? :p

Anyway, as those were drifting topics, I will forgive you. (;

This pseudogrey technique is a very nice idea, and simple to implement.
I think I will add some of it in some G'MIC filters (B&W conversion probably).

Edit : Done, will appear in the upcoming 1.5.1.6 version (to be released very soon !)

Fantastic news David, but, do you have 1786 tones with your result? lol

I might be careful to label the conversion as not a true grayscale/b&w conversion so as to not confuse people. There is color there!



Yes, I just re-checked with the updated script, and the gradient layer does, in fact, report 1786 unique values. The final psuedogrey image comes pretty darn close, but is slightly less depending on how fully the values are represented across the image.

Yeah; after final edits, I usually do an auto-level and that more then helps to insure that the pseudogrey result gives you 1786 tones, but even that is not a 100% guarantee (even when I was using the sample colorize and pseudogrey target). Again, thanks a heap for your Script-fu Pat; at least 10 times faster then doing this manually. I just updated the tut again a few minutes ago that also links your Registry entry by the way.

To me, the pseudo-grey technique is basically a kind of simple dithering (but not in the spatial domain, but in the value domain). In G'MIC, I process my images with float-valued precisions, and the result of my B&W conversion filter is indeed float-valued. So, the pseudo-grey step can be seen as a quantification technique that quantize the float-value result in 1786 tones, instead of 256. Of course, there are colors at the end, but still the image looks B&W, so I don't think I should rename the filter after all

Of course, and I wish that GIMP would catch up enough that your G'MIC functions could retain full floating point precision back into normal GIMP workflows...

I can see the value in being able to map to 1786 values in lieu of 256 for sure - I just didn't know if it might give people the wrong impression on what to expect as output? Maybe someone would be expecting a fully grayscale image, and might be surprised at the slight color result, that is all.

Of course, I am all for it, personally. (btw - G'MIC is an incredible piece of software, and I have been using it heavily as a part of my normal workflow - so I just want to take 2 seconds to tell you - Thank you very much for the work you've done.)

Dear Patrick,
You are right, the resulting image is indeed in color. But the good thing is that if you *really* convert it back to a B&W image (e.g. by changing the image color mode in GIMP), you should exactly get the same result than the filter without the pseudo-grey dithering option activated. So, it just adds informations to make the image more good-looking, it doesn't completely change the image content. In this context, I was seeing it as a slightly improvement of the B&W filter (considering that the pseudo-grey dithering is disabled by default).

Also, thanks for your encouragements, you cannot imagine how it is beneficial for my mental health !

I almost understood what you are talking about David and that scares me. lol

I also understand the need to challenge the noggin; reason I took up retouching and digital art a decade ago.

Found out that samj had a beta version of 1.5.1.6 that she compiled yesterday so just tried the Pseudogrey option. Not actually getting 1786 tones though, but played with some of the sliders and got 1780. lol

Still, very cool that G'MIC can Pseudogrey too. Updated the tutorial thread links to include this as well David and thanks a heap.

So, I wanted to play around at the pixel level with this method to consider extending the encoding even further. I'm SUPER lazy, so I just went ahead and wrote it up as javascript.

Javascript PseudoGrey Converter

It currently does what I call "base" pseudogrey conversion - meaning that any RGB value can only shift by 1 to give 6 tones between integer tones of greyscale (1786 total tones).

It's using html5 canvas elements - so it will only work in modern browsers that support it.

Also - it DOES NOT upload your image anywhere. It just loads it up in your browser locally and does the encoding on your own machine.

Due to browser limitations, if your image is large (there is not limit on how big it can be - just the encoding time will increase linearly) you will not see a Right-Click -> Save As... option. In that case, Right-Click -> Copy Image - then paste it from clipboard directly into GIMP.

Cool stuff again, Pat; I'll add the javascript to my tut thread as a link. Since I don't browse using an html5 compatible browser, I'll have to let others share their experiences with it. lol

Is Color Cube Analysis an accurate count of Pseudogrey?

Another look at 1786 True Pseudogrey gradient.

Yet another cool way to count colors. Assuming the gradient that you created was straight black to white. Cool that if done right, it gives you the 1786 tones per the Pseudogrey specs. Might try the color cube analysis the next time I do a Pseudogrey odinbc.

Where can I get this plugin odin; thought I had this one, but color cube in GIMP (thought the name was different; lol) is not the same.

MV's Plugins the Wire Worm guy.
http://www.vicanek.de/

I used your gradient from here,
viewtopic.php?f=10&t=4245#p52399

Thank odinbc; the plugin works great in GIMP too which is cool (haven't tried wireworm yet but remember the thread where many here touted its coolness. lol). Did a comparison and you can see that the color band is wider then then straight desat version.