How to quantify small medium and large in Topaz Detail 3?

[earlybird]

Hi,
 I have been enjoying working with Topaz Detail 3, and generally just make adjustments by eye while viewing the monitor at a scaling percentage (42%) where 300 pixels of data appears as an actual inch on my monitor.

 I have seen comments about the frequency that humans seem to respond to most readily, but I have not been familiar enough with the subject for the ideas to sink in.

 I there a way to quantify the ideas, where perhaps you have a planned image size, a p.p.i. density, and an anticipated viewing distance in a way that the terms "small", "medium" and "large" are more meaningful and or specific?

 I am afraid I don't really know how to pose the question, so I am hoping that someone who does, can rephrase the question if that is helpful.

 Thank you. 

[Bart_van_der_Wolf]

Hi,
 I have been enjoying working with Topaz Detail 3, and generally just make adjustments by eye while viewing the monitor at a scaling percentage (42%) where 300 pixels of data appears as an actual inch on my monitor.

Hi,

Yes, if the final output size is known, it can help to zoom out to the appropriate level to get a better feeling of how the final result will look when scaled to output size.

Topaz Detail itself only cares about the total file size in pixels. Medium details in a large file are larger than Medium Details in a small file, but the same details are targeted by "Detail". The affected details are proportional to image size (not  physical size), which makes it easier to create a single preset that will work on all file sizes. But for a better sense of what the end-result will look like one should zoom to the approximate output size for a preview, just like in printing and output sharpening. The best zoom ratio is 'display_PPI / output_PPI' , but on screen zoom often uses poor resampling quality, so what you see may not be what you get, hence approximate.

I have seen comments about the frequency that humans seem to respond to most readily, but I have not been familiar enough with the subject for the ideas to sink in.

That has to do with the Contrast Sensitivity of the Human Visual System (see followup post). But it is often misinterpreted that those medium resolution regions (say 4-6 cycles/degree), where our eyes are most sensitive to, are the only ones that need to be pushed for a better looking image. Instead, we are given an opportunity to (also) boost those other spatial frequencies that our eyes are less sensitive to, the higher spatial frequencies, the smaller details at 60 cycles/degree. Those spatial frequencies, or detail sizes, is exactly what Topaz Detail offers control over, and then some (e.g. based on color contrast, or by deconvolution sharpening).

I there a way to quantify the ideas, where perhaps you have a planned image size, a p.p.i. density, and an anticipated viewing distance in a way that the terms "small", "medium" and "large" are more meaningful and or specific?

Besides to above mentioned zoom ratio to approximate the effect at output size, there is a trick to better understand what it is that you are altering with "Detail". That is useful, because "medium detail" in a macro shot will target totally different structures than in a wide angle Landscape shot. So I recommend the following trick to better understand what you are changing. The amount of change is a matter of taste/creativity.

TIP (the effect is easiest to see/learn at 100% zoom, the final amounts can be tweaked afterwards at the simulated preview zoom ratio):
1. Start with reducing the Small / Medium / Large Detail controls to their -1.0 minimum value (see attached).
2a. Now only increase the Small Detail slider towards the default of 0.0 and try to find values above or below that neutral position that achieve a boost (or a reduction) of the detail that it considers to be "Small". You can quickly reset the slider to neutral by double clicking the "Small details" text in the Detail controls panel, and start varying from there.
2b. If a boost is used, and that generates noise amplification in smooth skies or other smooth regions, then a small negative "Small Details Boost" will usually solve that.
3. Now reset the "Small Details" control to its -1.0 minimum, and do the same exercise as above with only the "Medium Details" slider. You can also keep the "Small Details" controls at their previously found positions, but that will make it harder to see which spatial detail is specifically targeted with the "Medium Details" control. With a bit of experience you will not need to reset controls, but in the beginning it may help.
4. Repeat for the "Large Details" control.
5. Now set all controls at the same time to the values you've previously found (remember them or take a note), and tweak if the total gets too much (or to little) improvement, specifically at the approx. correct zoom ratio.

Do note that these settings will be applied to "Overall" detail, or "Shadow" detail, or "Highlight" detail separately, or cumulatively to all 3 regions!  The three option boxes that can alternatively be active are independent of each other, but the control settings are all applied, if set to anything else than default neutral positions. That makes it possible to e.g. apply overall detail settings, but specifically boost the highlights changes further, or reverse the other settings on them. So the three Overall/Shadow/Highlight option boxes represent 3 sets of control settings that will all be applied simultaneously.

The respective "Boost" versions of the 3 detail size controls, affect the very low contrast detail of the specific size that is targeted, which allows to modify the effect of the regular contrast controls which thus need to be set first. That's how it allows to reduce noise amplification with a negative boost amount if the Small Details control has been changed from Neutral. It's common to use a bit of negative "Small Detail Boost if regular "Small Detail" has been increased.

And that's only for the spatial frequency detail. The Tone Panel offers further choices to amplify or reduce detail differences based on their tonality. All these controls try to avoid unwanted Saturation changes, so one can concentrate on the specific areas of interest without having to worry about side effects.

Cheers,
Bart

[Peter McLennan]

Excellent, Bart! Thanks for that. Topaz should hire you.

[Bart_van_der_Wolf]

Excellent, Bart! Thanks for that. Topaz should hire you.

Thanks Peter, I'm open to offers ...

Cheers,
Bart

[earlybird]

Hi Bart,
 Thank you very much. This info is very helpful.

[Bart_van_der_Wolf]

That has to do with the Contrast Sensitivity of the Human Visual System (see followup post). But it is often misinterpreted that those medium resolution regions (say 4-6 cycles/degree), where our eyes are most sensitive to, are the only ones that need to be pushed for a better looking image. Instead, we are given an opportunity to (also) boost those other spatial frequencies that our eyes are less sensitive to, the higher spatial frequencies, the smaller details at 60 cycles/degree. Those spatial frequencies, or detail sizes, is exactly what Topaz Detail offers control over, and then some (e.g. based on color contrast, or by deconvolution sharpening).

For those not fully up to speed on this Contrast Sensitivity Function (CSF) stuff, this link will give some pointers with a fairly good summary of the issues for viewing, and even though it's from a Telescope/Astronomy point of view, it's equally relevant (if not more) for our regular image output situation. Output is limited in contrast, compared to our eye's and display's capabilities, so the CSF even more important when we produce printed output.

When we shoot an image, there are already things like Veiling Glare that will reduce the original scene's contrast on the sensor, and squeezing a high contrast image content into a limited range contrast output (display, or worse a print), will require proper tonemapping (e.g. a good implementation of Clarity) or else the image will look flat and lifeless.

But then we still need to manage the Human Visual System's (HVS) contrast sensitivity which drops when detail gets smaller and smaller until the contrast that our eye sees becomes so low that we cannot detect any detail. If we were to only boost the spatial frequencies (or angular detail if we also account for viewing distance) where the HVS does best, then the image would only gain a bit of 'pop', nicer contrast for those size details we can already see the best anyway.

But when we would boost the contrast of those small details where the HVS can barely see any detail anymore, around 60 cycles/degree (or an average 20/20 vision limit, or something like 290 PPI at reading distance), then the image will look sharper overall, because we can see more small detail despite our eye's loss of contrast there. Not only will we see e.g. more/smaller tree branches against the sky, but we can also start to see more surface structures like e.g. veins in leafs, despite the loss of contrast of our Visual system's limitations.

Our HVS also loses contrast sensitivity at the lower end of resolution. And because that determines the overall impression of contrast, we can manipulate that in our favor. Also remember that medium to low spatial frequency detail will become higher resolution detail when viewed from a distance, so if we boost those spatial frequencies then the output will remain looking sharp from a distance as well. The peak of our eye's contrast sensitivity also changes with age (very low a birth, increasing with age, and then falling again as we get older) and it varies with the level of illumination at which we view an image. So if we know that illumination levels will be poor, we need to compensate a bit more.

Finally, we can also reduce contrast if we want to de-emphasize certain detail sizes, e.g. because the image gets too 'busy' and might distract from the main subject / composition. Topaz Detail allows to apply a mask if we want to only apply local enhancements, or an overall opacity if we want to subtly reduce the resulting effects a bit. It also allows to "Apply" a given stage of adjustments, and then keep on building more effects on top of the earlier ones while still in the plugin. It also offers Snapshots, e.g. to compare between different approaches while in the plugin environment, or to capture different stages of progress should one want to backtrack just a few steps at once, or it allows to Undo per latest adjustment steps.

Cheers,
Bart

P.S. Jim Kasson's blog has a nice post on the subject as well, and a follow up post that describes the differences between Luminance and Chrominance Contrast sensitivity. The latter cannot be addressed with Topaz Detail directly, but may be dealt with by indirectly targeting local Luminance as a substitute, or by using the Tone adjustments that allow to change relative color contrast, or by boosting saturation locally. But Luminance detail is most important for Human visual resolution.

[Alan Smallbone]

Thanks Bart that is an excellent write up with a lot of very useful information. The helps me a lot.

Alan

[Peter McLennan]

Great stuff again, Bart.  Thanks.

For more on this important and fascinating topic:

https://www.amazon.com/Vision-Art-Biology-Margaret-Livingstone/dp/0810995549/ref=sr_1_1?ie=UTF8&qid=1472311906&sr=8-1&keywords=vision+and+biology

An excellent book on the human vision system and how we see.

[Bart_van_der_Wolf]

You're welcome guys.

I like to share my findings, no need for everybody to re-invent the wheel over and over again, and it helps if we can 'hit the ground running'.

That's not to say that the wheel cannot be improved by applying new insights and techniques. But for detail enhancements in images, the combination of Clarity for contrast, and (spatial frequency based) Detail enhancement are the most effective methods. A slick user interface helps, but it's the (educated) user that makes the real difference.

Human vision is a wonderful thing with many interesting evolutionary optimizations, which we can exploit if we know a bit about them.

Cheers,
Bart

[Jack Hogan]

But for detail enhancements in images, the combination of Clarity for contrast ...

Hi Bart,

Do you mean Topaz Clarity?  If so could you share your workflow for it?  If not, what algorithms do you prefer?

Jack

[Bart_van_der_Wolf]

Hi Bart,

Do you mean Topaz Clarity?  If so could you share your workflow for it?  If not, what algorithms do you prefer?

Hi Jack,

I think that Topaz Clarity offers a very good implementation of local contrast control, although there are possibly other variations of Clarity that I have not tested and that are also very good. So I'm open to suggestions, and in the mean time I use Topaz Clarity because it's the best I have sofar found.

What makes it so good, is the level of control it offers with real-time preview, its adjustments are halo-free, it hardly affects saturation when contrast is changed (so no unpredictable side effects).

It's much harder to give specific tips for Clarity, because its effects depend heavily on available contrast in the image content, so different images can require different settings. Also, cameras with a large dynamic range can render images flatter if not tonemapped more. Also scene lighting will change overall contrast, so different lighting requires different settings. And there is no right or wrong, it's very much a matter of taste. But in general one can say that because our capture process reduces contrast, a boost is usually warranted.

Topaz Clarity addresses specific ranges of contrast, throughout the tonescale, so it is not a midtone contrast tool.

"Micro Contrast" targets very small local contrast differences, like in cloud structures that are mostly smooth but 3-dimensional, or low contrast micro-detail near the resolution limits. "Low Contrast" targets slightly larger contrast jumps. Both Micro and Low are good at penetrating haze, or increasing it by reducing contrast. "Medium Contrast" targets yet larger contrast jumps, and "Large Contrast" targets the largest contrast jumps, and has more an effect of setting the overall image contrast.

To control the potential clipping of shadows and highlights, there are "Tone Level" controls. The black point is pretty much anchored, but the Black Level allows to lighten up the shadow regions, and the "White Level" allows to avoid featureless highlight areas that became too light. The "Midtones" control allows to generally brighten or darken the image midtone regions. But because that will also change the contrast and thus the effect of the "Dynamics controls", you may need to iterate a bit between "Dynamics" and "Tone Level" controls.

To get into a certain ballpark setting, depending on what I want to achieve with certain subjects, I created a few presets that I tend to use for the initial step, and then I can adjust the individual controls to improve the look.

For example, if I want to beef up the image contrast into something more 3D looking, I use a preset that I named "Pop 30", if that's too heavy handed then I use a "Pop 20" version with similar but reduced settings, or more subtle one called "Pop 10", or just reduce opacity. If the overall look is fine, but e.g. micro details are too strong, I just reduce that slider, or if the overall contrast gets too high, I'll reduce the "Large Contrast" slider (and re-adjust the Tone Levels).

One does not have to use the same setting for the whole image, the powerful masking allows to specifically target certain regions, e.g. clouds. One can also pre-record a few Cloud presets if one shoots a lot of those, which would probably produce a range of Dynamics settings that are highest at Micro or Low contrast, and lower for Medium contrast and lower yet for Large contrast, and a reduced White level that adjusts for the brightening on whites to avoid clipping.

Cheers,
Bart

[brandon]

Thanks Bart
Really helpful advice.
I found with Topaz Clarity that I struggled with it pushing my shadows to dark, thinking it more like a midtone contrast boost which as you note it isn’t (in part). I could boost the dark tones it seemed to defeat the purpose and look a lot. Now I’ve just been processing via CaptureOne using linear raw conversion and by that approach even the default Clarity controls seemed much better suited to get a full dynamic range AND the nice punch and 3D look you refer to. Are you using it this way (from a low contrast raw conversion)?