Fovea Pro

[Doug Nelson]

By Doug on Sunday, April 08, 2001 - 11:24 am:

Does anyone have any experience with a Photoshop Plugin called "Fovea Pro" or it's little brother artfully named "the Image Processing Toolkit"?

I'd be interested if any of its features are adaptable to restoring old photos.

http://members.aol.com/FoveaPro/ for details

By Chris Russ on Monday, April 23, 2001 - 02:14 pm:


Wow. You caught me. I am the author of both and have actually used FoveaPro to do retouching. No, this was not its intended use, but there are a number of plug-ins in the suite that are well suited for retouching and other things. (My favorite is an action that I call "Out-Spot-Out" that does the following to a selection:
1: IP*Render->Fill Region (with "from Edges" selected)
2: Expand the selection 1 pixel in radius
3: A small Guassian blur ~1 pixels radius (sometimes more)
4: Add Noise, Gaussian, typically ~6, although for some sharper images I use less)
This does an excellent job of removing dirt from a uniform background quickly.)

But, the current price is somewhat high so we should talk about a subset that would be well suited to y'all.

we have an oneline tutorial at:

http://www.ReindeerGraphics.com/tutorial

-Chris
Reindeer Games, Inc.

[jeaniesa]

Moving the discussion on the Image Processing Toolkit and Fovea Pro to this thread (from Critiques->image to examine)...

Doug, you said that you have a copy of the software? Have you tried using it to remove the "white dots" created in the scanning process of textured color photos? The webpage that I pointed memphishooter to deals with B&W, but I haven't found anything on that site that deals with color images. (Then again, I haven't spent hours pondering it either.)

I can tell just from that website that there won't be a simple user interface to the SW, but if it will do a good job at removing the white dots, I'd pay $250 and figure out how to use it! I was thinking of sending one of my scanned images with the white dots to Russ/Chris and asking them to show me what's possible. Don't want to purchase the SW if it won't work. But if they (or you?) can show me it will, I'd buy it tomorrow (well, almost that fast. ;-)

Thanks, Jeanie

[Doug Nelson]

Ok, here's what I've discovered so far about 'pattern removal' with Fovea Pro (which is virtually identical to the IPT except it can handle 16bit images):

The manual shows some dramatic examples which I would say, if replicatable, would justify the purchase price for any working professional.

However...

I have not been able to replicate any of these effects in any sort of attractive manner. I do not feel this is a limitation of the software, but more of the opacity of the instructions and interface.

There are no preview fields and slider bars. When you activate the filter you get a dialog box asking for numbers. Sometimes as many as 5 numbers with frigtening-sounding names like 'coefficient'.

Now, the software CAN and will provide you with these numbers, but they're functions of other filters, and you're expected to know what you want in very specific terms.

In one example, removal of halftone dots, you first use one filter to calculate a pattern which represents the mathematical repetitions (patterns, as opposed to more varying elements such as picture contents). You then make a mask based on this (basically it looks like domino dots) and then apply it back to the original using some of the aforementioned numbers (which you knew to get via other filters).

I have a Pentium4 1.4ghz machine and this still takes awhile

I must repeat (especially since Chris reads this forum) that I am in no way criticising this software. It's obviously brilliant, more brilliant than I just as obviously. It's just not aimed at non-image processing professionals.

Chris mentioned the possibility of marketing a subset for less scientific purposes. I have sent him my own recommendations (which were pretty much along the lines of 'simplify the interface').

[thomasgeorge]

Doug, That type of interface is fairly common in the medical/Astro image processing software fields. Some, like Iris and a few others will leave a person scratching their head. They are powerful programs but that power comes with a price---a rather steep learning curve and most assume a knowledge of math that perhaps most folks outside the research areas of those fields may not have. They are Excellent programs and if the designers and engineers which "make 'em work" were to simplify the interfaces I think even at $$$ range they would have no problem selling them. If anyone out there is reading this--- The Richardson/Lucy and Max Ent. deconvolution algorithms would be of great interest to lots of Image Processing folks who are not involved in Astronomical image processing. Thanks, Tom

[Doug Nelson]

Deconvolution is how I first ran across Fovea Pro.

For those who don't know (which is all of us), deconvolution is basically taking a blurred image, running some horrendous math on it, and generating a sharp(er) image (before you go running around and doing handstands, it won't work with a lot of blurred photos)(but it does work with some).

A couple of years ago I ran across a deconvolution plugin for the Linux program called "The Gimp" (a nice image editor, check it out if you use Linux). The author had some dramatic examples on his website. I exchanged a few emails with him about the possibility of porting it over as a Photoshop plugin.

He wasn't interested, and I forgot about it until a few months ago. I did considerable websearching trying to find the information again (I failed) and ran across Fovea Pro instead.

Fovea Pro does offer several deconvolution options, but I've yet to get even one to work on a 'normal' photo.

[thomasgeorge]

Doug, From what I understand there has been some success applying the deconvolution algorithms to sharpening scans of non astro subjects. Unfortunately with the exception of a program called Maxim DL( pretty popular plus user friendly) and perhaps a couple more( I am not sure) the interface to run the deconvolution routines are complicated, and the one I tried required a defocused/blurred point of light to focus its"attention" on.Nothing about it was user friendly. Running the algorithms themselves require experimentation as to how many iterations to run as there is no "cook-book" recipe--it is a matter of experience and intuition developed by "playing" with them. They are also memory intensive and one should have a fairly large amount of RAM and a fairly fast processor as when these routines were first introduced it could sometimes take up to 10 hours to complete a series of them. However with most computers now the time has been cut to much more reaonable levels. I think Maxim DL is avaliable for download in a demo version, however the price is almost $300 for the regular flavor and the one with CCD camera control module is over that amount by a fair amount. It features several other tools which might be of interest also. Tom

[Doug Nelson]

Got a link for Maxim DL and the 'couple more'?

[thomasgeorge]

Lets see, if memory serves me correctly, Mira is one, Qmips is another, and for the rest I will do some digging but the others all had deficencies as regards the interface friendliness or were rendered somewhat less than useful by the introduction of the two aforementioned ones due to advances in the algorithms. Of all of the ones mentioned, Maxim DL seems to me to be perhaps the best of the lot for user friendliness. As I recall Mira required you add "modules" to the basic program in order to have the deconvolution functions, but I may be mistaken. Tom

[thomasgeorge]

Doug, You might want to checkout Megafix and another little number named Astroart. Tom

[thomasgeorge]

Try this link for starters-you can get many places from there. www.astrovid.com - Tom

[jeaniesa]

Thanks for the info on Fovea Pro, Doug. I got an inkling as to what the interface would be like by reading the online manual for the halftone dots solution. What I didn't realize was how many "experiments" I would have to go through to get the results that they do. Perhaps I could finally find a use for that math minor I got 15 years ago. (Yeah, right - I remember pulling out my diffy-Q notebook 5 years ago and looking at it like it was written in Greek - which I guess it was to a certain extent.)

Anyway, since I'm still intensely curious as to whether or not it's POSSIBLE to get the results I'm looking for by removing the grid of white dots created by scanning textured (color) photos, I might just send a sample to the authors of FP and see if they will spend the time to show me what it can do. Then, I can decide if it's worth the immense amount of time it would take me to figure the whole thing out.

Jeanie