Loss in detail TIFF vs RAW???

[skoobey]

So I was wondering, how far do you guys push the raw before you are satisfied with the conversion?

I found that once it's "balanced", it is easier for me to add color/contrast/whatever adjustments as adjustment layers in photoshop, rather to go trough couple of conversions?

Maybe I do a separate conversion for the sky, or the shadows, or the hair, or the dress or something, but mostly it is to pull the detail, and then I do every little local thing in PS?

The advantage of this being that I can edit everything at any point, and not have to redo any healing, stamping, warping.

So, is this final contrast/color/brigthnes/curves/whatever in PS approach good? I don't think I'm loosing much if any detail compared to just having a number of raw conversions when there really isn't a need for them?

Does tiff loose detail faster than RAW when you push-pull? How much faster?

[Tony W]

Re: Loss in detail TIFF vs RAW???

Not really considered this in much detail before but overall I would want to accomplish as much as possible within the raw editor.

FWIW my initial thoughts on this and apologies for stating what may already be obvious to you.

Wherever possible I think it is always preferable to do as much as is reasonably possible in raw. The tools in newer versions of LR and ACR can accomplish quite a lot of what we used to have to do in PS layers.

Of course there are limitations to what can be accomplished using the raw tools and at some stage a conversion may have to be made to accommodate your image editing needs and of course a conversion will be made in most cases as your finished file will not be raw format.

Once you make the conversion and save to either Tiff or Jpeg you have effectively ‘baked your cake’ and have much less room for manoeuvre with image editing – still by that time you should be close enough that any additional editing should not have a detrimental effect on the final image.

A lot is made of the non destructive nature of editing raw files and while this holds true I do think at times this aspect is overplayed. It could be argued that any editing you undertake either raw or PS could be viewed as destructive (you are changing pixel values destroying the original values?).

With raw of course you can easily delete your editing steps and have the pristine original file as shot. But of course you can do the same using PS, your raw file will remain unchanged (other than those reversible editing steps made in the raw converter prior to opening in PS) and using layers gives you a method of back stepping should you go in the wrong direction.

If you are starting with a poorly lit/balanced shot (sometimes unavoidable other times you curse the photographer!) and having to make separate conversions then using raw in PS as a smart object and then using New Smart Object via copy can be very useful as you are now in the PS environment and able to make separate edits to the raw data. Of course these should be done prior to any other editing that is needed in PS.

So finally to get to your questions I do not think that a final fine tuning of colour/brightness etc is a bad approach and in fact may be needed after extensive working on an image, but wherever possible the global and local corrections (including sky, shadows, hair etc.) should be made on the raw data – this after all is the best data you have!

Does Tiff loose detail faster than raw well I guess that depends on the recipe you have used to arrive at your ‘baked’ Tiff. If you have really pushed the image too far and lost information then it is likely that you will not be able to recover loss of detail without stepping back to the point where you lost that detail and this may be as far back as the original raw.

TBH, it really does depend on you and keeping your eye on the details as you progress with your editing – sometimes easier said than done when concentrating on fine details to the exclusion of all else.

[klev]

Re: Loss in detail TIFF vs RAW???

Wherever possible I think it is always preferable to do as much as is reasonably possible in raw. The tools in newer versions of LR and ACR can accomplish quite a lot of what we used to have to do in PS layers.

The methods they used are designed to shoehorn in with existing users. The program is used in so many ways by so many people in however many languages, that it's difficult to break off into a more modernized paradigm. Even After Effects is better in that regard.

Also the ideal method would be if it just used LUTs and 16 bit half float exr output. You wouldn't have the issue of clipped reds or garbage values from weird profiles like prophoto.

[Tony W]

Re: Loss in detail TIFF vs RAW???

Also the ideal method would be if it just used LUTs and 16 bit half float exr output. You wouldn't have the issue of clipped reds or garbage values from weird profiles like prophoto.

Interesting comments but I have to say that I am not exactly sure what you mean about clipped reds or garbage values from ProPhoto as I have not really experienced this working on raw files - other than where I would anticipate this to happen.

Clipped reds apart from happening during capture (sensor related?) would be expected with a highly saturated red subject e.g. a flower after conversion to a smaller gamut space e.g. sRGB. Obviously the highly saturated red will not fit into the sRGB space so you will have to toss away some colour info.

Prophoto I see nothing weird other than being a very large gamut in comparison to others. AFAIK LR uses something similar to Prophoto the main difference being that Prophoto in PS has a gamma value of 1.8 and the version that is used in LR has a gamma value of 1.0 to more closely match a linear capture device. So again not sure where the 'garbage values' come into play.

[skoobey]

Re: Loss in detail TIFF vs RAW???

Okay, yes I understand what you're saying, and I even do things like split toning and contrast in RAW sometimes, as I have to batch process images for proofs etc... but....

For example, If I add contrast curve in the raw, then everything is already baked, and I want a "clean" image, and add contrast selectively where needed in PS. So, I don't want to raw convert for every one of those areas. Easier in PS.

Or if i want a low key image with plenty of clipped blacks, I want to make it clip in PS, rather then have two conversions, because I need a bit more detail in that arm or whatever. I just leave it as is, adjust WB, and in PS I add a clipping adjustment on top, which I can adjust locally.

I do things like exposure recovery, some "look defining" things in raw, but curves should definitely be left for PS in my opinion, split toning and contrast, and clipping curves. So, that's what I'm asking, I already asked some retochers "that are up the ladder" directly, and they all basically said: you need more detail, use raw, you need to remove detail, use PS, so I was just wondering if you guys/gals feel the same way.

It's more of a philosophical question, rather than case by case one I guess...

[Tony W]

Re: Loss in detail TIFF vs RAW???

You may find it easier and more intuitive in PS and in some cases just not feasible in ACR - IMO there is not necessarily a right way just the way you choose to reach your goal.

To pick the examples you quoted and using raw as an alternative to consider - no one right way just alternative!

...For example, If I add contrast curve in the raw, then everything is already baked, and I want a "clean" image, and add contrast selectively where needed in PS. So, I don't want to raw convert for every one of those areas. Easier in PS.

Virtually all the tools available to you in the raw editor are available to be applied selectively. Using the Adjustment brush, Targeted adjustment brush, Graduated filter and minus values can be painted and applied as required.

Or if i want a low key image with plenty of clipped blacks, I want to make it clip in PS, rather then have two conversions, because I need a bit more detail in that arm or whatever. I just leave it as is, adjust WB, and in PS I add a clipping adjustment on top, which I can adjust locally.

Why would you need two raw conversions in this example. Likely your arm detail could be just as easily recovered with a minus Adjustment brush (after you have clipped the blacks globally) and reside within the instructions of the raw file xmp. If you really must have two seperate images conversions then a virtual copy in raw is very effective in terms of convenience and does not use much in the way of resources

I do things like exposure recovery, some "look defining" things in raw, but curves should definitely be left for PS in my opinion, split toning and contrast, and clipping curves.

Everything that you can accomplish in PS curves can be done in raw starting in the Basic panel and progressing to the Tone Curve panel either directly or using the point curve tools to change individual RGB values - some may argue that the tools offer more flexibility than PS curves alone.

So, that's what I'm asking, I already asked some retochers "that are up the ladder" directly, and they all basically said: you need more detail, use raw, you need to remove detail, use PS, so I was just wondering if you guys/gals feel the same way.

That generalisation may have held true in the past but you can of course remove detail just as easily in raw as in PS. Learning a particular skill using certain tools in one application can sometimes blind us to the capabilities of any new tools presented in an application that we are perhaps not so familiar with and perhaps not having the time or just being a little lazy for want of a better word we stick to the old ways we have always used.

IMO editing in raw is not going to be a great deal of use for high end beauty retouching and for anything that requires precise pixel level work - at least I find it difficult and go to PS for this .

So I guess it could also be argued that you are going to end up in PS at some stage so you could just wait till you get there to apply your tried and tested methods.

Your original question was about loss of detail Tiff vs Raw was this question answered in any way with my take on this?

[klev]

Re: Loss in detail TIFF vs RAW???

Interesting comments but I have to say that I am not exactly sure what you mean about clipped reds or garbage values from ProPhoto as I have not really experienced this working on raw files - other than where I would anticipate this to happen.

Clipped reds apart from happening during capture (sensor related?) would be expected with a highly saturated red subject e.g. a flower after conversion to a smaller gamut space e.g. sRGB. Obviously the highly saturated red will not fit into the sRGB space so you will have to toss away some colour info.

My explanations are not always the best. Yes it relates to highly saturated objects and relative colorimetric conversions to something like Adobe RGB. Relative colorimetric clips per channel where the target value mapped through LAB or whatever reference space photoshop uses falls outside of the destination gamut. It would make for a much cleaner workflow if you could preserve all of that initially and make tweaks to preserve channel detail when preparing for output spaces. I prefer some of the LUT based workflows implemented in newer software, but that's a much broader topic. For a reference example check out opencolorio.

Prophoto I see nothing weird other than being a very large gamut in comparison to others. AFAIK LR uses something similar to Prophoto the main difference being that Prophoto in PS has a gamma value of 1.8 and the version that is used in LR has a gamma value of 1.0 to more closely match a linear capture device. So again not sure where the 'garbage values' come into play.

That is extremely interesting. I was using "garbage values" in a somewhat colloquial sense, not a programming sense. I meant that it can push colors into weird places that likely to clip in other spaces, yet I can't figure out anything that it closely represents. It doesn't work well as a universal reference color set where you can look up other colors between gamuts in reference to prophoto. It doesn't describe the behavior of specific hardware against measured reference values. My issue is that adjustments made within it against a layer with unmasked alpha can cause some extremely unusual hue shifts and buildups of saturation. Anyway there are better methods than using a gamma 1.8 space with an extremely broad spectrum.

.

For example, If I add contrast curve in the raw, then everything is already baked, and I want a "clean" image, and add contrast selectively where needed in PS. So, I don't want to raw convert for every one of those areas. Easier in PS.

What you're searching for is an unsupported workflow in photoshop. I'm not aware of any program that supports a 1.0--->1.0--->2.2 viewing space LUT kind of workflow. The raw files are always quantized relative to a somewhat restrictive gamut that is a subset of a reference color model where values are both stored and viewed in a non-linear space. I shouldn't get too far into this or I'll attract Adobe's number one fan to the thread (assuming he still reads this forum).

[skoobey]

Re: Loss in detail TIFF vs RAW???

So basically, do what you feel like.

Yes, I know I'm clipping it other color spaces once I turn it Prophoto or adobeRGB, but at the same time, I feel like I've really progressed with "being easy" on the image with doing less "final" steps in raw, and giving it some space to fine tune in PS.

[Tony W]

Re: Loss in detail TIFF vs RAW???

... Yes it relates to highly saturated objects and relative colorimetric conversions to something like Adobe RGB. Relative colorimetric clips per channel where the target value mapped through LAB or whatever reference space photoshop uses falls outside of the destination gamut. It would make for a much cleaner workflow if you could preserve all of that initially and make tweaks to preserve channel detail when preparing for output spaces. I prefer some of the LUT based workflows implemented in newer software, but that's a much broader topic. For a reference example check out opencolorio.

Now I see where you are coming from . Perhaps this is why some photographers advocate staying within sRGB for the whole workflows as it is a safe bet that the destination gamut will not be smaller than sRGB - still I think that misses much of what is achievable even with low cost inkjet printing which in some cases can exceed Adobe RGB gamut in certain areas. At this time I think it highly unlikely that we will see monitors capable of displaying the full visible gamut of something like Prophoto or printers capable of printing such a gamut - still I live in hope that things are improving and will get more affordable given time

That is extremely interesting. I was using "garbage values" in a somewhat colloquial sense, not a programming sense. I meant that it can push colors into weird places that likely to clip in other spaces, yet I can't figure out anything that it closely represents. It doesn't work well as a universal reference color set where you can look up other colors between gamuts in reference to prophoto. It doesn't describe the behavior of specific hardware against measured reference values. My issue is that adjustments made within it against a layer with unmasked alpha can cause some extremely unusual hue shifts and buildups of saturation. Anyway there are better methods than using a gamma 1.8 space with an extremely broad spectrum.

For me the choice of editing in Prophoto seems the logical step. First is the fact that it is generally a good principle starting with more data than we are actually going to need - we then have a choice of that which we have to discard for whatever reason. I think that this is postulated in the Nyquist theorem and would appear to hold true within our world of digital editing - more than that I cannot say as I do not claim to be an academic.
Secondly, AFAIK most if not all digital camera devices are capable of capturing data that when converted will probably exceed the gamut of Adobe RGB and therefore can only be fully contained within the larger gamut of Prophoto or similar. So I would prefer to make sure that all that data is maintained and available to work on both now and in the future.

[klev]

Re: Loss in detail TIFF vs RAW???

For me the choice of editing in Prophoto seems the logical step. First is the fact that it is generally a good principle starting with more data than we are actually going to need - we then have a choice of that which we have to discard for whatever reason. I think that this is postulated in the Nyquist theorem and would appear to hold true within our world of digital editing - more than that I cannot say as I do not claim to be an academic.
Secondly, AFAIK most if not all digital camera devices are capable of capturing data that when converted will probably exceed the gamut of Adobe RGB and therefore can only be fully contained within the larger gamut of Prophoto or similar. So I would prefer to make sure that all that data is maintained and available to work on both now and in the future.

Damn you for adding to my reading list! Anyway computer scientists who focus on computer graphics have done a huge amount of research in this area. As a result rendering processes have trended toward linear internal calculations over time. Post workflows in video have also trended toward linear pipelines, allowing for greater flexibility in finishing and mastering. The way they deal with the viewing issue is by LUT to LUT, so some kind of vector class that encapsulates a set of scalar values and maps them to a viewing set from within a reference set. It should be better for falloff like attenuations and things than trying to come up with math to be applied directly to an uneven distribution of values. I'm not sure what Adobe does. I always assumed it was some kind of homogeneous matrix transform operation when converting between spaces. I linked to the wiki for anyone (understandably) unfamiliar with the term.

Anyway I don't object to large spaces, but I don't always feel that photoshop's tools were designed with them in mind. Having read their blogs, they mention things like falloffs toward the shadows and highlights with burn and dodge tools. These things were designed for a different era. They wanted to provide some continuity with darkroom workflows that could be applied to scanned data. The amount of time spent on auto-selection tools directly relates to the fact that you're not working hand to image. It's somewhat disjunctive, so they try to algorithmize the process due to limitations and the sheer number of ways people interact with the program via mice, tablets of varying size, and trackpads.

I'm drifting off topic there, but I think the reason we're kind of stuck with these workflows really comes down to the age and ubiquity of photoshop.

[skoobey]

Re: Loss in detail TIFF vs RAW???

It is true, and I did find it be a steep curve to conquer, yet now that I understand what I am doing, I find it to be really "analogue", which I like. As someone who never dealt with film, and gone digital first, now I understand both.

Imagine, how tough it would be form someone like me to enjoy film processing, if Photoshop were a completely linear experience.

I think it will get more "to the point" with cc, but that might me unfortunate for people who use it in other fields (science etc.).

As for wast color spaces... well I do understand that when someone tells me: this is going to be printed on Epson bla bla, here's the color space you should use, but for everything else, I found Adobe RGB and LAV to be most predicable.

[klev]

Re: Loss in detail TIFF vs RAW???

Imagine, how tough it would be form someone like me to enjoy film processing, if Photoshop were a completely linear experience.

This has nothing to do with anything. We get hung up on qualities of film that were inherent to the medium. Trying to achieve the same look is a highly inductive process, as creative processes are, but the role of a piece of software is to impart control, nothing more. Tools are generally meant to be predictable. You would find that most photographers chose films and +1 +2 -1 -2 etc. processing based on past results. Beside that it would likely be easier to create a series of LUTs with a linear lookup assigned to sets of data. With ICC profiles the data bounds are drawn quite early, and large gamuts do not work due to the way things are quantized. Gamma encoding in general is just a way of dealing with a limited dynamics range. Your suggestion doesn't help pull more of a film look from your digital camera files. If anything it's more restrictive.

Regarding Epson, it's not really a guarantee that it will only be reproduced in one place. Predictable results that do not cause unintended temperature fluctuation within certain subsets of colors and subtle bugs due to uneven value distribution is a good thing. whenever the topic comes up though someone always misdirects to the topic of non-linear human vision and color perception variations due to age, color blindness, etc, yet it has minimal impact on making predictable calculations. I look at a lot of software and just see the many hacks that remain for legacy reasons, even if they have started to update certain algorithms. Here's an example. Brightness/contrast changed with CS5. Unsharp masking still uses an inverted gaussian distribution. I'm curious how something like a lanczos filter lookup would work on de-mosaiced raster data.