REQUEST: Collaboration for testing a new effect for substituting color in images

[GSResearch]

Hi all, I'm new to the forum and I'm a freelance researcher in image editing.
I'm italian, so sorry in advance for my english .
I'm also a little crazy , I left my job as employee 'cause I believe in my researches and I want to make them my only job.
I developed a new effect for substituting colors in images. I would like to test the goodness of this new approach useful for retouching photos ... so I ask you a little collaboration, if you want to help me, you should retouch the image that I attached in this way: change the color of the pen fromred to white.
It's important to me to know how much time you spent in retouching the image and what applications you used for achieving the result, so please post the altered image together with the requested info. In a few days I'll post the image elaborated with my algorithm.
Really thanks to all, in advance, for the collaboration.

GS

 

[IamSam]

Hi GSResearch and welcome to PSG.

Please forgive my suspicious nature but something is off about this request. Is your algorithm associated with Photoshop?

 

[MrToM]

3 minutes 11 seconds.
(3 minutes and 1 second was for masking, I did it manually).
Nothing but Photoshop.



Regards.
MrTom.

 

[GSResearch]

Hi IamSam, thanks for the welcome! I answer:
No, I'd like it was so... I explain better: I would like to propose this algorithm to Adobe or other companies but, for now, I just want to test its goodness and compare it with images elaborated mainly with Photoshop by expert users; I simply think that Adobe Photoshop is the best application for elaborating images and so the best way to get a valid comparison. I hope there is nothing wrong in my request.


Cordially, GS

 

[IamSam]

Ok, I understand a little better now. No, there's nothing wrong with the request. Good luck!

 

[GSResearch]

Thanks so much MrTom! The thumbnail is so small though... is it possible for you to post a larger image? No problem for the watermark ! Is it a problem for you to explain the tools you used? Although I'm not so expert with photoshop I tried to use it to achieve what I asked in my request but in my tries (Photoshop CS6) I didn't get a good result with:
1) "Replace Color" in "Adjustments" of "Image" menu
2) Changing lightness and saturation after selecting only the pen
3) Masking the pen to a new layer and doing the right operations for preserving shadows and shades
... but your elaboration seems to be better! However The image you posted is too small and I can't compare it with the image produced by my algorithm. Just to tell something more on my algorithm I got a optimal result without the need of masking the pen though.


GS

 

[Tom Mann]

To select the reds, and then desaturate and brighten the resulting grays, I used three methods:

a) One of my oldest and most valuable plugins, Asiva Select. Below is a screen shot from the user interface for this plugin. Click on it to see it at higher resolution. Once your image was loaded in PS, it took me well under 30 seconds to open Asiva Select, develop the settings you see, and produce the selection.



It then took me at most another 15 seconds to use that selection to mask a levels adjustment layer and a vibrance/sat adjustment layer, resulting in the 2nd attached image. As you can see from the animated GIF below, Asiva select included some areas that shouldn't have been selected (eg, white line under the chin, white thin horizontal bar in the left foreground). I could have easily cleaned these up, but, for the purpose of this exercise, speed, I didn't bother.


b) Nik's Viveza - This was probably the fastest of these three methods, probably only around 30 seconds, but the results were not as good as either of the other methods.



c) Finally, PS's native color range tool - The corresponding time for this was more like 2 or 3 minutes. Since this is a native tool, and presumably everyone knows how to use it, I didn't post a screen shot of its user interface.


Finally Click on the in-line preview of the image iteself, below, for an animated GIF comparison of the three methods.

HTH,

T

 

[MrToM]

I've updated my original post with the image.

In Photoshop:

1. Duplicate original layer.
2. Open 'Channel Mixer'.
3. Set Red value of RG+B channels to 100%, the others to 0%.
4. Mask out the bits you want to keep as original.

10 seconds to change the color, 3 minutes to mask.

Regards.
MrTom.

 

[GSResearch]

Thank you very much MrTom, you explained in a very detailed manner the way you achieved your image and it was exactly what I needed. Your third method is the better than the others you have shown, infact from what I saw in Photoshop the 'channel mixer tool' in conjuction with the 'color range tool' is the best way to alterate a particular color in a image, far better than the 'Replace Color tool'. You achieved a result very similar to the output of my algorithm. My algorithm has some advantages,the same ones that the 'Replace Color tool' has but without its drawbacks (its results aren't always good):


1) It's simple and intuitive
2) You can operate on the whole image with an unwanted alteration of colors limited or totally absent
3) You simply set the source and the destination color
4) You can calibrate how much preserving shadows and shades versus a shifting of color shades towards the destination color


Well, finally, I attach two images obtained with my algorithm: the first one preserves shadows and shades, the second one is a little shifted towards white. I achieved these results in max 30 seconds but only for calibrate in the best way the parameters.

GS

 

[GSResearch]

Oh I apologize, I made a little confusion between the posts of Mr. Tom Mann and MrTom (Similar name!), so... first of all thank you very much both of you...well anyway in my previous post I said what I think despite the confusion. I presume that the third method used by Mr. Tom Mann was the same used by MrTom just with a little difference in the way the mask was selected. Sorry and thanks a lot again for your support.


GS

 

[Tom Mann]

No problem whatsoever, GSR. Our two names are very close.

Your algorithm sounds like it has a lot of potential. If the algorithm has been turned into an actual PS plugin, is there any sort of pre-release copy we could try / beta-test for you? Depending on its capabilities and how well it works, this is something that I might have use for.

Tom M

 

[admin]

Impressive result @GSResearch - from this image it looks like your tool has potential.

Are you planning to share or distribute this as a plugin?

I'm sure you could find some good beta testers here to provide feedback.

 

[GSResearch]

Mr. Tom Mann and Mr. Admin thanks a lot for your encouragement. I explain you well the actual situation: I'm testing the algorithm on a standalone platform I developed for testing and improvements purpose. I would like to develop a plugin for Photoshop and I thought about it as a second step in my to do list but now, after your words, I could try to do the plugin for Photoshop cs6 so as to give it to you for beta-testing.
I attach another image I ask you to change (blue to black) if you want ..thanks in advance




GS

 

[admin]

It would be interesting to see as a Photoshop plugin if you can develop it that way.

Here's my go at the second image. The image is not the best quality, there are a lot of artifacts already in the image. This makes it difficult to get a clean looking result.

This took about 2 minutes.

 

[GSResearch]

Again... thank you so much Mr Admin for your support. Well I don't want to judge your work I'm sure you are making it the best way possible with Photoshop, anyway your black is a little greysh.. one of the advantages of my algorithm is the ability to shift colors toward the destination color in the best way possible, tell me if I'm right. As I said another advantage is preserving mainly the color shades and the shadows of the image so that the result doesn't look altered. Well.. I attach three image with various grades of black, few seconds were necessary to get the result and I didn't care too much of achieving the best result.
Please feel free to express your opinions, good or bad are helpful for me!


GS

 

[admin]

your black is a little greysh

Yes you are correct. To me it looks more realistic this way, like the original it looks like real fabric with some parts faded from washing or similar. It was a choice I made in the process, could have easily made them more saturated if desired.

Your images appear to have too much saturation in the black, to me it doesn't look realistic. Also, the color of the eyes are changing in your images. This happened to me as well, so most of my 2 minutes was spent on masking out the eyes, face, and hand. In my image, only the fabric color is changed and nothing else.

Of course you could mask parts of your images as well to achieve the same result.

 

[MrToM]

To be honest your method is too clinical.

Admins' 'black' may be 'grey' as you put it but in my eyes it is more true to life.
His results show a knowledge of not only how the real world is perceived to the human eye but also how to replicate that in a digital image.

The preservation of detail is of most importance when altering any digital image. You have to agree, Admin has quite obviously done just that with a result which would keep any client happy.

Much of the work relating to image manipulation comes from the heart, from the trained eye of the artist to the tip of the brush....if it doesn't 'look' right they have the years of experience to fall back on and correct it almost immediately, no matter how fine a difference it may be, something you cannot define by x=y*z.

I can't really see much of difference in the 3 images you supplied but that may be down to monitor calibration etc etc. One thing that does stand out though is that they all now lack any detail in the areas in question.

It does sound an interesting process but if its too clinical it'll never get the results of a skilled artist.

All the above is of course based on the information supplied at the time.

Regards.
MrTom.

 

[GSResearch]

Thanks Mr. Admin and MrTom for your analysis, I agree your elaboration is more realistic (Mr. Admin) but the purpose of my request was really substitute the the source color with the destination one trying to preserve the image quality at the best, I know, a full saturated black is not the best choice but I made that choice because of a greater difficulty , I apologize for my lack of clarity. MrTom, I need to be clinical, the purpose of my request is the analysis of an algorithm, it's really like x=y*z, I'm making a tool ... the artist will use it in the best way. Ok ...so... just to go on and to be practical... I made some elaboration using Mr. Admin's image as Reference...and Mr. Admin, it's obvious, I worked on the whole image ...I don't care about the alteration of the blue in the eyes. Just calibrating better the parameters I obtained a result similar to Mr. Admin's image, not perfectly the same because without masking is not possible to darken the brightest pixels globally. In a second image I try to preserve more the light reflections present in some points of the cloth or as you prefer the the points faded from washing, with a black a little darker. I attached two images for comparison too (on the right Mr. Admin image). The differences are minimal I know but try to say what image best matches the original from the point of view of shades and shadows. Well .. I dont want to judge your work Mr.Admin I repeat I want only show what my algorithm is able to do... and if you think that I'm working without masking I think It's not so bad.
Thank you so much for your support...really


GS

 

[Tom Mann]

To take the comments of Admin and MrTom to their logical conclusion, unless the software that you eventually create (based on your mathematical algorithm) provides the end user with easy and quick adjustments to highly detailed aspects of both the starting and target pixels, I doubt that your eventual software (ie, algorithm PLUS user interface) will be generally accepted, at least by more experienced Photoshop users.

As a specific example, Admin's techniques gave the cloth in question less contrast and made it be a bit brighter than the results of your algorithm. So, Admin's results were described as "gray". In addition, it looks like his method desaturated more of the shadow area near each strand of cloth, whereas your method left some of the nearby shadow areas slightly blue. Of course, as one works on other projects, there will be some images on which I would definitely prefer Admin's look, and other situations where I would prefer the look that your algorithm produced.

However, speaking for myself (and probably many other long-time users of PS), IMHO, unless your future software product provides a very easy way to be adjusted to produce both possibilities and everything in-between, I will likely continue to use the very flexible, elemental tools built in to PS for the entire job, rather than start the job with your software and then have to finalize the result using native tools such as a curves layer or a bit of unsharp masking to increase the local contrast. I would likely take this approach not only because it is (hypothetically) more general and flexible than yours, but also because I would rather darken and add contrast to Admin's results than attempt to lighten and reduce the local contrast in your darker result.

Just my $0.02,

Tom M

 

[GSResearch]

Mr. Tom Mann I agree with you and infact I have to say that the UI for my algorithm is very easy to use: you have just to select source and destination colors and move generally max three sliders for calibrating the tollerance. It would be a replacement for the "Replace color" tool that usually doesn't produce very good results, especially when the destination color is white or black. I have to repeat that the only intent of this algorithm is to substitute a color with another, trying to preserve at the best the look of the image ( contrast, shadows, shades.. ) and not to improve the quality of the original image. I think the advantages of using it can be deduced from a calculus based on time spent, effort and skill requested.... from these points of view I can say:
1) generally, few seconds are required to obtain a good result
2) the effort is low because you have just to calibrate parameters through some sliders and select source and destination colors
3) no particular skill is required


Well, no more requests, I only show some result produced in a few seconds and always without masking (the original is the red version).
Thanks all