Can't D&B technique be greately simplified with...

[_jason_]

...with a trick?

I am watching the famous Dodge&Burn tutorial video from http://nienna1990.deviantart.com/ now.

It's amazing. And impessive.
Especially the hours of work, all the nearly pixel-based micro dodging and burning needs... Well, the results are impressive so far...

...but the idea spontanuously comes to me the is: can't that be extremely simplified with doing some logical/automated tricks?

Why not simply generate some certain greyscale highpass images based on the picture contrast and use it as a masked overlay with Lighten/Darken Mode to dodge or burn with a layer mode? This way all the hours of manually dodging and burning could be traced down to some easy steps on wide parts of the picture?

I mean, if it were possible to capture the areas of darker and lighter micro areas into a greyscale image, then it theoretically should be possible to use that grayscale image as a dodging/burning mask...

Has anyone tried that yet? I do not know whether that really works but i'll surely make some tests next days. If that works, it could be a greatly improveme workflow and save alot of precious time.

It's an really interesting theme!

[mistermonday]

Essentially dodging and burning is the reduction (or increase) of contrast. Doing at the extremely localized level (a few pixels at a time) produces the best results as you have seen but takes many hours for most images. Yes, there are many techniques out there which are applied to an entire image or parts of an image. Many involve blurring, filtering, and advanced blending. Some are very effective and realistic looking but no one has come up with an automated way that looks as good as the "hard labor" pixel level D&B. That being stated, many of us continue to develop new techniques to speed up the process without sacrificing quality. I have not yet found another solution which is acceptable for high end retouching.
Regards, Murray

[0lBaldy]

you might try Luminosity masks on a curves layer to adjust specific Light, Dark, and Mid tones... Luminosity masks are explained here: Tony Kuyper Photography (luminosity masks tutorial)

Amica has made and provided an action that puts all of the luminosity masks into the channels pallet and can be downloaded from here: RetouchPRO>Blogs>amica999

To use the mask: Control-Click on the masks then back in the Layers palette add a new Curves Adjustment layer - this will load the selection in the Adjustment Layer mask. Now you can adjust your individual Light, Dark, and Mid tones. Paint the mask black on the parts you do not want lightened/darkened

Have fun and when you get your technique perfected make sure to post it here!

[_jason_]

Originally Posted by mistermonday

Essentially dodging and burning is the reduction (or increase) of contrast...

Well exactly. But basicaly you make the darker areas (pixels) lighter and vice cersa the lighter pixels darker to match an nearly equal tone. And all that manually over hours but with a slightly randomized imperfection due to human errors, which is part of the system.

It's in the century of computer algorithms somewhat puristic and cumbersome I think.

I believe there is the possibility to automate such behaviour or at least develop a tool or technique which supports us doing just that: making the darker micro areas lighter and the lighter micro areas darker in a much faster manner than doing it quasi per pixel manually. Possibly this could work similar to the healing tool or such...

But the principle schould be very clear. There must be a logical mechanism which can support us with that D&B. I even believe this could be improved drastical with some smart calculations/algorithms including that human error factor.

And yes, it probably has something to do with highpassing or bandpassing and blending modes...

[_jason_]

Originally Posted by 0lBaldy

you might try Luminosity masks on a curves layer to adjust specific Light, ...

I think it hasn't so much to do with masks, rather with blending modes.

I am currently experimenting with highpassing the image and using certain blending modes. There are same really nice and promising combinations found already ...

But I still need more time to understand the theory behind.

It definately hasn't something to do with bluring or healing or similar techniques. It's all about creating/calculating a specialiced overlay image for certain blending modes like dodge and burn to automatically equalize the dark and bright micro areas as on wrinkles and blemishes and certain kind of unwanted shadows.

[Flashtones]

Jason, it sounds like you are going down a path many have gone down before (see degrunge, split frequencies, high pass, inverted high pass guassian blur, apply image, Imagenomic Portraiture, clone tool at low opacity, airbrushing, etc.)

All of these are effective to varying degrees and can diminish the amount of D&B required (and possibly eliminate it on lower end work) but they are ultimately global solutions to local problems. Or, when used locally they are local solutions to micro problems. IOW, blunt instruments when exactitude is required.

If all that was required was that darks got lightened and lights got darkened the problem would have been solved a long time ago. The problem is that it has to be done selectively and judgmentally to differing degrees across the surface.

[Cuervo79]

The thing is that D&B is applied with the vision of the retoucher and as of yet it can't be done via an automation process. Look at the healing and clone tools, there isn't a filter or automation that can take away scars or pimples automatically and with a realistic look, that same principle applies to D&B since it's not done with a defined method. Nor will the problems be the same with every photo. Since I ignore how the math works in photoshop I have no idea if there can be a way to make a filter or an automation that can do that work.

There are at times shortcuts you can do (for example copying the red channel and adding it to a new layer with the luminosity blending option leading to less contrast on the skin) but if you're looking for the "realistic" look the shortcuts don't always work.

[_jason_]

Ah well. I've seen already.
The technique I tried still needs alot of manual adjustment.

But it basicly works (in principle) and all traces down to generating a kind of lightness analysis and 2 resulting masks or overlays, which then can be used to apply some sort of automated/calculated D&B on selected areas.

Don't get me wrong, but I still believe this could dramatically speed up the entire process of using D&B.

I'll do some more research on that...
(Saying "No, it won't work" is the enemy of any invention.)

[Quantum3]

Degrunge is the fast way to D&B, but be careful with this technique.

[TimeActor]

Hi quantum3 i´ve seen your homepage and your work looks realy good. Is it possible to talk about your "degrunge" technic?
I´ve never heard of that.

[Quantum3]

Originally Posted by TimeActor

Hi quantum3 i´ve seen your homepage and your work looks realy good. Is it possible to talk about your "degrunge" technic?
I´ve never heard of that.

Yes (btw, it's not my technique).

Here is the tutorial: http://retouchpro.com/tutorials/?m=show&id=213 The part of "Quick" is the one you need. Then read this other article (a bit more harder, but easy at the end and more easy to understand than the not "Quick" part of the first tut): http://www.modelmayhem.com/po.php?thread_id=439098

[mistermonday]

The human eye is quite sensitive & perceptive to to very small differences in luminosity. Typically when you D&B in a micro localized area (for example a pore or section of a skin line and their surrounding / adjacent areas) , the difference in luminosity of 2/100 or 3/100 (I use LAB color L channel as a reference) makes all the difference to our perception whether or not we have produced a pleasing result for that specific tiny area. The problem with skin, is that it is not uniform at all. One pore or wrinkle line may need to be dodged higher 15/100 in luminosity, and another pore which is 10 pixels away might need to be dodged 6/100 higher. Most techniques using blurring, masks, degrunge, etc which work over "relatively larger" areas do not have precision to affect a particular small number of pixels without applying the same changes to adjacent pixels - and this is very noticeable. For some types of retouching this is acceptable but for high end beauty, it is most often not acceptable.
But I agree with Jason that the technology exists to help automate and speed up the process. I think Adobe could help at the PS code level by writing some tools specially targeted to speed up D&B.
For example, a brush setting in the Brush Tool option bar that would allow the user to set a threshold that would stop the flow when the pixels being dodged reached a luminosity which is = x/100 lower than the adjacent pixels. Ideally two settings, one for luminosity threshold, the other for radius of surrounding area. This would allow us to dodge a pore without having to brush over it 8 times at low Flow. It would save time by preventing us from over-dodging / over burning and having to switch to black brush to correct it. Another tool Adobe could give us would be the ability to generate masks which are derived from pixel(s) which differ in luminosity from adjacent pixel(s) by a certain value, the value being definable by the user. This would allow us to isolate all the dodge or burn targets that require luminance increase of "X". We could then use one setting of brush OP and Flow and do all those targets very rapidly.
A third tool Adobe could give us would be an on screen info window similar to the eyedroper info in the Info window. This would provide us with a better visual indication in our working area of the actual luminosity of the target and its surrounding area. Right now when we work on the layer mask we only get to see a luminance value of the mask we are working on and for D&B purposes that has little to no value.
I guess it may be time for a discussion with some Adobe developers.
Regards, Murray

[Flashtones]

MisterMonday, I can't wait for you to do a video/tutorial/RetouchProLive event, as you have so much knowledge to share.

Couple of questions:

Originally Posted by mistermonday

For example, a brush setting in the Brush Tool option bar that would allow the user to set a threshold that would stop the flow when the pixels being dodged reached a luminosity which is = x/100 lower than the adjacent pixels. Ideally two settings, one for luminosity threshold, the other for radius of surrounding area. This would allow us to dodge a pore without having to brush over it 8 times at low Flow. It would save time by preventing us from over-dodging / over burning and having to switch to black brush to correct it.

How is this better/different than user adjustable curves to set threshold, and pen pressure to set opacity/flow/size on the fly?

Another tool Adobe could give us would be the ability to generate masks which are derived from pixel(s) which differ in luminosity from adjacent pixel(s) by a certain value, the value being definable by the user. This would allow us to isolate all the dodge or burn targets that require luminance increase of "X". We could then use one setting of brush OP and Flow and do all those targets very rapidly.

How is this better/different than user created luminance masks derived from channels/selective color/calculations, etc?

I'm not challenging your ideas, just trying to wrap my head around them. Is it that the current means are mostly visually based and you feel that by taking a mathematical approach less visualization would need to be applied "on the fly"?

I'm wondering if some of this math isn't what's being used under the hood of a program like Imagenomic Portraiture, which does a pretty good global degrunge. But it's still global.

[gamedonechanged]

btw if you are looking at the degrundge technique maybe it's better to look at the newer version of that in the high pass sucks thread over on MM.

[mistermonday]

Flashtones, thanks, they are on my todo list, which at the moment is time challenged and in overflow mode. I am working on it.

Couple of questions:
Quote:
Originally Posted by mistermonday
For example, a brush setting in the Brush Tool option bar that would allow the user to set a threshold that would stop the flow when the pixels being dodged reached a luminosity which is = x/100 lower than the adjacent pixels. Ideally two settings, one for luminosity threshold, the other for radius of surrounding area. This would allow us to dodge a pore without having to brush over it 8 times at low Flow. It would save time by preventing us from over-dodging / over burning and having to switch to black brush to correct it.


How is this better/different than user adjustable curves to set threshold, and pen pressure to set opacity/flow/size on the fly?

Good question. In fact I have experimented with multiple threshold curves (see attachments) 1 and 2. There is no reason to use the conventional curved pinned to 0 and 255 with a big arc (above centre for dodge; below center for burn). Add a fixed number of points to the 0 point and the same number to the 255 point. For example if you add 10 points, this adjustment curve contrains every pixel in the image to a maximum luminosity increase to 10 pts. So you can start with 5 pts and add several curves with limits a 5, 10, 15. You can add a curve on top of the stack limiting to 1 or 2 points so the curve can act as a final fine tuning curve. Given you have a a range of opacity of 0-100 and a flow of 0-100 on each curve, you have a limitless range of very fine control. There are a great many variations to limit threshold curves including solid color layers set to variuos blend modes in RGB or in LAB color mode as nested smart objects within an RGB file. However, depending on how fancy you get, in the end there may not be sufficient time savings for the extra complexity. While you can set limits, you eye will not be able to easily tell which spots need to be dodged 6 points versus 3 etc. Without some intelligence behind the scenes, the process will be slow for us humans.
Now consider that you choose a spot to be dodged (like a pore). The pore is fairly dark with a luminosity of 51. The surounding skin is pretty uniform but its luminosity is 63. Ideally you want the pore to be somewhere between 59 and 61. Your brush OP is set to 15%; flow set to 15%. With that large a gap in luminosity to close, you either need to change the OP/Flow settings or use the lower settings and brush many more times over the pore. Your brain can do these calculations and assessments on the fly BUT at a very slow pace. Instead we are forced to close the gap visually as we watc our contrast layer.
Now imaging an additional option in the brush tool palette that allows you to set a "relative threshold". So you set your OP much higher, your flow at some medium rate. The relative threshold is set to stop the paint flowing when the luminosity of the pore / pixels being dodge hits a limit of 2 or 3 (59-61) bel0w the surounding skin (63). You can very quickly brush over a dodge target without 8 strokes and without worrying about going over. Same applies to burn. There is no reason I can see for PS to be able to do this. Right now we have to do it by sight, and trial and error. Lots of experience helps us in establishing how we set OP, Flow, and how much we brush. The contrast layer helps us with visualization, but I think we struggle much more than we should have too.

Quote:
Another tool Adobe could give us would be the ability to generate masks which are derived from pixel(s) which differ in luminosity from adjacent pixel(s) by a certain value, the value being definable by the user. This would allow us to isolate all the dodge or burn targets that require luminance increase of "X". We could then use one setting of brush OP and Flow and do all those targets very rapidly.

How is this better/different than user created luminance masks derived from channels/selective color/calculations, etc?

Masking is one of the disciplines I spend zillions of hours on throughout my time in Photoshop. I have found some interesting and different methods of making masks including fine detail masks, pore masks, etc. I experiment on challenging images such as the ones posted by RP member Cuervo70 showing difficult skin retouch challenges (thanks to Cuervo79 for permission to display this image -attachment 3). Nothing seems to produce a mask that will permit appropriate isolation and transparency to properly doge/burn one area without to some extent adversely affecting another area in the image - usually blurring it or the equivalent of blurring it.
Now imagine if you were able to select / isolate all areas of an image where the pixels within some radius were separated from adjacent pixels which differred in luminosity value by = X/100 (or x/255). Then you could apply a dodge or burn or other change to an entire group of pixels across the image that were separated by a luminosity difference independent of absolute luminosity value . Effectively you would be dodging or burning in arrays where targets are equivalent in the amount by which they had to be dodged or burned. This would speed up the process exponentially. For most images, if you set one sampler in the info palette to LAB and just move the eyedropper it around the image and look at the L value only, you will see that the difference between the darkest spots you need to dodge and the lighter values to which you need to dodge them, typically all fall in a range of 0-20. Moreover within that range, there is a distribution of smaller ranges where most of the targets fall.

OK I will stop rambling now.
Regards, Murray

[TimeActor]

Originally Posted by Quantum3

Yes (btw, it's not my technique).

Here is the tutorial: http://retouchpro.com/tutorials/?m=show&id=213 The part of "Quick" is the one you need. Then read this other article (a bit more harder, but easy at the end and more easy to understand than the not "Quick" part of the first tut): http://www.modelmayhem.com/po.php?thread_id=439098

Thank you.
i read this and i´m knew this technic - but not the english name ;-)
I´m don´t using the hp filter.

regards
Martin

[gamedonechanged]

Just need to read the first post. All the info is there.

[Flashtones]

Originally Posted by mistermonday

However, depending on how fancy you get, in the end there may not be sufficient time savings for the extra complexity. While you can set limits, you eye will not be able to easily tell which spots need to be dodged 6 points versus 3 etc. Without some intelligence behind the scenes, the process will be slow for us humans.

This was my concern in trying to set the variety of limits that would be needed on any given subject - that pre-thinking it and setting it up would be more cumbersome than traditional methods.

Now consider that you choose a spot to be dodged (like a pore). The pore is fairly dark with a luminosity of 51. The surounding skin is pretty uniform but its luminosity is 63. Ideally you want the pore to be somewhere between 59 and 61. Your brush OP is set to 15%; flow set to 15%. With that large a gap in luminosity to close, you either need to change the OP/Flow settings or use the lower settings and brush many more times over the pore. Your brain can do these calculations and assessments on the fly BUT at a very slow pace.

Isn't this in large part ameliorated by setting your brush to a higher opacity (lets say 100%) and lower flow (lets say 1-3%). That way your top threshold is limited by the layer itself and you control in between levels on the fly via pressure/duration=flow.

Now imaging an additional option in the brush tool palette that allows you to set a "relative threshold". So you set your OP much higher, your flow at some medium rate. The relative threshold is set to stop the paint flowing when the luminosity of the pore / pixels being dodge hits a limit of 2 or 3 (59-61) bel0w the surounding skin (63). You can very quickly brush over a dodge target without 8 strokes and without worrying about going over.

This sounds to me not too dissimilar from making a curve that raises lum 51 to 63, and setting brush opacity to ~90%.

...imagine if you were able to select / isolate all areas of an image where the pixels within some radius were separated from adjacent pixels which differred in luminosity value by = X/100 (or x/255). Then you could apply a dodge or burn or other change to an entire group of pixels across the image that were separated by a luminosity difference independent of absolute luminosity value . Effectively you would be dodging or burning in arrays where targets are equivalent in the amount by which they had to be dodged or burned. This would speed up the process exponentially.

That does sound interesting.

Anyway, You're more versed in the traditional methods than me, so if you think these tools would help I believe you. I'll be looking to see what techniques you come up with in the meanwhile.

[_jason_]

Originally Posted by Quantum3

Degrunge is the fast way to D&B, but be careful with this technique.

Interesting!

There are some minor similarities of what I want to achieve. But my approch has nothing to do with bluring, like the degrunge stuff.

It can be seen more like a highly specialized kind of "removing dirt and scratches" with the principles of dodge and burn (or lighten and darken) based on several analysis masks. But with more fine control over the entire processing.

The same technique also could be used to achieve complete contradiction: To emphasize all varieties of contrast differences. I.e. To make a skin looking just more structured (as long there is any structure information) whithout sharpen the picture in general.

And yes, the technique has to be used with care too and can't be applied on allover the picture and still has alot of manual/visual adjustments based on experiences and "trial & error".

I am still experimenting with it. Do not know whether it will be worth to publish a tutorial or not yet...

[_jason_]

Originally Posted by mistermonday

...But I agree with Jason that the technology exists to help automate and speed up the process. I think Adobe could help at the PS code level by writing some tools specially targeted to speed up D&B...

Murray, fantastic! Thats just the thing i thought also.
The math behind that even won't be that difficult at all.

I mean, if the D&B purists want to stick to nearly pixel based dodging and burning for hours, they still may continue to do so...

[_jason_]

Originally Posted by mistermonday

Masking is one of the disciplines I spend zillions of hours on throughout my time in Photoshop. I have found some interesting and different methods of making masks including fine detail masks, pore masks, etc...

Very interesting!
I am currenly also heavily experimenting/refining the mask generation for my idea. Because this is in fact the key of that all!

I am curretly trying to excessively use the highpass filter and shifting the values of such generated images to the area of masking. (That's just simple byte math *smile*)

However, I am currently not sure what to do with that color information inside those "bump maps" generated from the highpass filter ...

I also experimented many hours with the "Selective Color" Tool to generate masks. But that went into wrong direction for several reasons and generally was much too complex to master manually with precision...

[_jason_]

Finally I just want to point out, that such "D&B simplifying" won't make all "patchwork" superfluous. It's merely a nice challenge to discover something with new possibilities and techniques with a slightly different approach to good old and prooved things. Why not using the power of the CPU as much as even possible to automate always repeating tasks to give the freedom to concentrate on the workflow and thus speedup things drastically? For the retoucher time is money, right?

[mistermonday]

Flashtones,

Isn't this in large part ameliorated by setting your brush to a higher opacity (lets say 100%) and lower flow (lets say 1-3%). That way your top threshold is limited by the layer itself and you control in between levels on the fly via pressure/duration=flow.

OK, so you run into a pretty dark pore (L value=20 lower than the surrounding good skin color) and your OP=100. With the flow at 3% you will need to run the brush over the pore or wrinkle about 20 times before the target is dodged up to where you need it to be. This is fine but when you have a large number of these as well as a ton of dark lines, wrinkles, divets, and tiny depth anomalies, and all of these targets are at different luminosity levels, a single flow value will either not be convenient and / or the D&B process will take too many hours. This is where you need to assist the process with some automation, even if that automation is at the pixel level.

This sounds to me not too dissimilar from making a curve that raises lum 51 to 63, and setting brush opacity to ~90%.

There is no problem making a curve that raises every 51 value to 63 BUT what do you do with all the oixels that were at 51 that you did not want modified AND what to you do with the other 500 - 100o targets that need to go from 47 to 52; 68-73; etc etc. Playing with the curves is not manageable in my opinion.

Regards, Murray

R

[Cuervo79]

Originally Posted by _jason_

I mean, if the D&B purists want to stick to nearly pixel based dodging and burning for hours, they still may continue to do so...

I doubt that actual retouchers would shun a quicker way to do D&B. This isn't about being purist or not. It is about if someone can actually find a way to shorten the time of work and make it look realistic.

[_jason_]

Right, That's last but not least a somewhat philosophic question.
Porsche also has some serial production. But they still do to cars which are nearly made per hand in nearly every detail... *smile*.

[_jason_]

@mistermonday

Murray, you posted a really interesting Mask with one of you posts above.
I think this mask has too much contrast to be any useful.

In my opinion there is not enougth information (capured from the greytones in the mask) which could sucessfully be used to gain dodge or burn information for the correction. I think it is simply too hard (it's nearly black&white).

Generating such masks is indeed very challenging, it seems...
...and it, of course, depends on the usage uf such mask afterwards...

[_jason_]

@mistermonday

Murray, can I please use the picture (that eye from where you made your mask) to demonstrate the mask generation fom high pass filtering?

This would be very nice.

[mistermonday]

Jason, actually what I posted was not intended to be a conventional mask. That one is solid black and white and was an experiment in dodging where the areas which are not being dodged are protected and another tool is used to do the dodging - not a brush. Also that mask should be inverted.
As for the image, it belongs to member Cuervo79 here at RP. He has been quite generous and I think he will not mind you using it. However, always ask permission so could you please send him a PM.
Thanks and regards, Murray

[_jason_]

Thank you. I did send him a PM.

[holgaman]

Jason,
In response to your original question -The question you're dealing with is very intriguing, but involves only the technical aspect of dodging and burning. That basically involves using the tools correctly, or as you suggest, teaching a computer to use the tools correctly. It would be wonderful if that was all that was involved. A professional level retouch requires many artistic choices (based on personal preferences, which are as numerous as there are retouch artists), including, but not limited to, the dodging and burning process. You just can't teach a computer do that. To make the problem even more complex, every face is different with it's own set of problems and solutions. Then, multiplying that complexity to the power of 10, you have to consider what the end use of the photo is - is it a model comp for an agency, a generic editorial publication, a fashion magazine cover, advertising, personal? Anyone can learn to dodge and burn (or maybe even program a computer to), but it's the learning where to and where not to D&B from one face to the next that separates the good work from the bad. As with photography in general, the photographer is expected to be familiar with his or her tools and know how to use them (a fully automated camera is not going to make anyone into a great photographer). It's how that knowledge and expertise with the tools is applied to achieve a certain artistic vision that makes one photographer more desirable over another for any given project.