RawTherapee Sharpening Question

[ppmax2]

I agree with some of your points above; I'm going to take another stab at the C1 render later tonight. On comparing these again, I think I cranked the USM a bit too much in C1...looking back over my settings did I really use 168/1.1? This will bring the contrast and halos down for sure. The highlights on the bottles, and the labels on the bottles are way too cranked.

I'll also post my files with the shadow noise...finding the sweet spots for NR in RT is still eluding me.


thx--
PP

[ppmax2]

This one is better I think--USM now at 120 and .8.

http://ppmax.duckdns.org/public.php?service=files&t=004aa88c81f01843fed0716af1ca420c

Ill post my files later tonight.

PP

[Fine_Art]

That is quite a bit better, the crayons are starting to look like wax, the thread is maybe polyester now instead of nylon. Do yourself a favor and ditch the USM in favor of a deconvolution package. The yellow cloth still looks like mustard, the orange like ochre, the pink a bit purple. Maybe we need them to describe the real color. If you are comfortable with C1 use it. I think YOUR result in RT was better. For example, look at the metal of these side by side brushes. In C1 the brush is well rendered. In RT the metal looks more like aluminum with subtle colors. The C1 is just shades of grey. Look at the bristles. Much finer in RT. Maybe all you need to bring out the extra is deconvolution with C1?

[Fine_Art]

Since nobody has thrown in a lightroom conversion here is Imaging Resource ACR conversion. Are they pretty good with the program? Probably. Are there better on this board? Without question.

http://www.imaging-resource.com/camera-reviews/sony/a7r/FULLRES/AA7RhSLI00100_ACR.HTM

[ppmax2]

So just for fairness here's one from Aperture:
http://ppmax.duckdns.org/public.php?service=files&t=6ee8ebae1babdf52dbfe6e9fc826648e

I sharpened a bit, added some detail, and pulled back the black point. Curiously, the colors in the checker looks closest to RT. The yellows in LR look too punchy and a bit green to me...I'm not a fan.

That is quite a bit better, the crayons are starting to look like wax, the thread is maybe polyester now instead of nylon. Do yourself a favor and ditch the USM in favor of a deconvolution package....If you are comfortable with C1 use it....Look at the bristles. Much finer in RT. Maybe all you need to bring out the extra is deconvolution with C1?

The first try in C1 was way overdone--my bad. It was late at night!

The yellow cloth still looks like mustard, the orange like ochre, the pink a bit purple. Maybe we need them to describe the real color....I think YOUR result in RT was better. For example, look at the metal of these side by side brushes. In C1 the brush is well rendered. In RT the metal looks more like aluminum with subtle colors. The C1 is just shades of grey.  

I thought you said you thought the color in C1 was better?

Each of these tools have their strengths and weaknesses and I think it's good to have a working knowledge of all them so that you can bring out the best in your photos. I feel like I have a better grasp of RT now...my workflow is shaping up something like this:
* before doing anything, set the exposure to neutral!
* set camera profile and lens profile
* set WB
* get rid of noise
* adjust exposures and curves
* adjust color (still struggling with this in RT...but will get the hang of it)
* sharpen

My only gripe so far with RT: would it be so hard to park the histogram somewhere where you don't have to constantly open/close a panel to see it (with your image maximized to fit)?

Fine_Art, if you are interested and up for it, I'd like to see what you can do with these two files in RT. Blow me away!

Edit: posted the wrong file:
http://ppmax.duckdns.org/public.php?service=files&t=c3c11413debde530a4c24b66c8222256

These are the right ones:
http://ppmax.duckdns.org/public.php?service=files&t=2457caaea1b22a63622dca139d6638b8
http://ppmax.duckdns.org/public.php?service=files&t=af778de4fb2e78531e4d4058faf6061b



thx
PP

[Fine_Art]

I thought the colorchecker in C1 looked better, the cloth colors look better in RT. The imaging resource ACR conversion has good color. The whole file looks a bit soft relative to RT. I have not checked out yours yet.

Thanks, I will be happy to check out your files late today. I have a long day of work ahead so I may not post anything until tomorrow. I'm sure others will have a go as well.

[Fine_Art]

I just got home and cleaned up its 10pm so I wont start editing your files. The Aperture looks pretty good. I have never seen output from that program before, thanks for posting it. The yellow cloth still looks like mustard, the pink is a bit red. The detail is decent, nothing looks plastic that shouldn't. That is a good program ( I dont use mac).

[ppmax2]

No worries, take a load off. Thanks for giving those files a shot.

Aperture is a good app that doesn't get a lot of respect. Considering it hasn't seen a major update in a long time it holds up really well, IMO. It shines with well lit (nominal/"good") images and is good at getting you where you want to go quickly. I believe it has the best interface for adding local adjustments, and is well regarded as having great library mgmt and org tools. I also think it does great for coping with blown highlights. It gets a lot right, and it does a lot well.

On the down side it starts to be really limiting when trying to recover details in underexposed and noisy low light zones. I'll post a render of one of the cr2's I posted and you'll see that while color is good other tools like C1 and RT have so much more to offer to deal with these situations.

Of course, it's also missing things like lens corrections, gradient masks, and the one thing I really like in LR--the ability to adjust curves right in the histogram. I also wish it offered local exposure control.

On the other hand, no other tool I've used enables you to stack multiple adjustments of the same type...e.g. N number of curve adjustments can be added, some local, some global...and any adjustment can be enabled/disabled selectively. C1 and LR can do this to a degree, but their implementations are "half empty" rather than "half full."

Oh well, now that Aperture has been EOL'd, I'll keep using it until I'm forced to move on. I hate the thought of losing all the "value" I've added to my library (all the metadata that enables me to access all these images quickly and efficiently). Faces, Places, all the keywords, flags, tags, ratings...plus all the nondestructive adjustments...they can take it from my cold dead hands!

Pp

[Fine_Art]

Ok, Ive looked at your telescope image, very cool BTW, I have made a few adjustments in RT which may not be available in other programs. First is lab, There is some interesting color in the earth so I brought that up with a curve that saturates the dark tones. Another thing is line noise, this would enhance it which is already showing up. I set the line noise preprocessing at 15. The NR is set strong in luminance 65 with the detail set at 96 or 98. You can feel free to use strong NR while telling it dont mess with my detail. I set the red and blue NR about 25.

Given you have a nice image, I am not going to put a full size on the net, I will attach a few screen shots along with the RT processing parameter file. You can play with it from there. I will take another look tomorrow.

[Isaac]

* before doing anything, set the exposure to neutral!

There's a "Preferences" button in the top-right, and the "Image Processing" tab allows you to set default parameters to Neutral (or some other profile).

[ppmax2]

All I can say is WOW. I like this:
http://ppmax.duckdns.org/public.php?service=files&t=3bdf0d0c2c764e08706369d3e8c9cb51

For comparison, here is the same shot with C1 and Aperture. Keep in mind that Aperture and C1 both have the benefit of local adjustments, and I think there is better detail in the highlights of the clouds in these pics...but I feel I could get there with RT and some out-of-app local tweaks in Pshop or Pixelmator.
C1:
https://flic.kr/p/ofeMCx

Aperture (slightly cooler and pinker WB and highlights):
https://flic.kr/p/mNCPzM

Fine_Art, you have showed me the light. I used your file as a starter to create the JPG above. The detail, noise control, and color are all really good.

The turning points for me were:
Your initial recommendation to ditch USM and try RL deconvolution (not available in C1 or Aperture or LR);
Recommendations for RL settings (make a big difference);
Your sample PP3 file that used Lab curves for Luminance, Chromaticity by Hue, and Chromaticity by Luminance;
Some of your NR settings, particularly the Line Noise Filter tucked away in the Preprocessing tab;

I'm really impressed with how RT can tweak color in so many different ways...which take some time to learn. This time I did all my color adjustments and Luminance tweaks in the Lab space, rather than "running for cover" in the familiar RGB and HSV curves. I love Lab 

If tools were musical instruments, RT strikes me as a violin: it takes long time to learn and play it right. Whereas other tools can get you closer quicker (IMO) RT is can really produce when stroked right

Off to try more....

PP

[Fine_Art]

I tried it again tonight after work. I mostly tuned the NR settings.
I aways start from the position of I want all my detail. Then I do red, then blue, then luminance. Without a lot of hunting I think it removes the noise at small cost.

[Fine_Art]

I'm having a look at your second image now. That is a very good test! It is significantly underexposed (based on ETTR) with a fine, low contrast line texture in the building. It will be interesting to find a balance of removing the Canon line noise while keeping the lines close to the pixel level. I think I will need a couple hrs to play with it. I am going with the clouds were not as blue as the default RT opening of the document so I will shift the temp to about 5000 to make it more grey. Let me know if that is not more realistic from your memory of the shot.

[Fine_Art]

This screenshot shows the lines of the building with NR parameters. I moved the luminance up in Lab.

The parameters file is also attached.

[ppmax2]

Those last two renders you made were really good. I think you absolutely nailed the noise/detail balance on the last shot with the fine vertical details in rear of the building. Nice work man.

For reference, here is a test from one of my first try's with C1:
https://flic.kr/p/nXLooZ

It's clear from your shot that you can pull detail from shadows in RT! I really wish it had some local adjustment capabilities...it would be sweet to bump the specular highlights a bit to give the image a little zing. Is there a published roadmap for RT anywhere?

Here's a version I've been working on in RT--I copied your NR settings and built from there. I like a dark/moody version of this:
http://ppmax.duckdns.org/public.php?service=files&t=8429111e848feb30a089752a4ee029e6

After using this tool for many hours at night I think I'm finally starting to get it. What I don't understand is why there appears to be so much saturated pink/gold when first rendering a shot using the "neutral" profile. Most of the shots I've worked on have required substantial desaturation to get things looking right and balanced. Does this happen to you?

Also, how do you copy specific settings from one shot to the next?

I like the Lab controls but it certainly is different vs. RGB/HSV curves. I really like the different curve controls you can use.

I really appreciate you taking some cycles to work these images--you've taught me a lot and shown me some really great capabilities with the tool. Thanks!

PP

[Fine_Art]

To copy settings from one image to another, or many, you can right click on the thumbnail - copy, the select the images to copy to - paste. Or Ctrl C then Ctrl V just like an office document.

I have not had problems with saturation unless I use the camera profile then use "Use DCPs tone curve" that can get very saturated. If so, just turn unclick it.

Glad to help, I am surprised nobody else did a conversion. It is like they are protecting their secret sauce. 

[ppmax2]

Hello again Fine_Art and other RT fans...Im back with a few new questions about RT:

What is the preferred method to reduce red channel highlight values? Imagine a hot sunset with reds that get clipped or nearly clipped in the histogram. Would you use a Lab curve adjustment to bring down Luminance according to Hue, or use an RGB curve to bring down the high end?

Considering you can apply a tone curve (2) in the Exposure section, or use a Lab Luminance curve, is there any pro/con to applying these in one place vs. the other?

I've started experimenting with the CIECAM controls--is there any documentation for the Lightness (J), Brightness (Q), and Contrast (J), Contrast (Q) controls do?

Thanks again...still discovering the marvels of this tool. I haven't touched C1 in a week

PP

[Fine_Art]

Good question, I would like to know what other people do as well.

Personally I bracket looking out for clipped channels. I think once the channel clips you cannot get real, accurate data of the scene. You can only fill from nearby data in various ways. I have ran into this with red flowers, where there seems to be no detail, just red. I tend to do a linear conversion in another program to look at where the data has contrast in the very stretched out linear. That usually gives me an idea of where to modify the image. I also downsize to 50% knowing the real single red or blue channel data is sparse.

For exposure vs lab luminance someone else will have to help, I have not gone deep into it. Ciecam, I must admit I know nothing about it.

[Tim Lookingbill]

Hello again Fine_Art and other RT fans...Im back with a few new questions about RT:

What is the preferred method to reduce red channel highlight values? Imagine a hot sunset with reds that get clipped or nearly clipped in the histogram. Would you use a Lab curve adjustment to bring down Luminance according to Hue, or use an RGB curve to bring down the high end?

Considering you can apply a tone curve (2) in the Exposure section, or use a Lab Luminance curve, is there any pro/con to applying these in one place vs. the other?

I've started experimenting with the CIECAM controls--is there any documentation for the Lightness (J), Brightness (Q), and Contrast (J), Contrast (Q) controls do?

Thanks again...still discovering the marvels of this tool. I haven't touched C1 in a week

PP

Just lurking about here, don't use RT, but to help in answering your question you might look into changing profiles along with applying curves and HSL adjusts going by what you said in your previous post...

After using this tool for many hours at night I think I'm finally starting to get it. What I don't understand is why there appears to be so much saturated pink/gold when first rendering a shot using the "neutral" profile. Most of the shots I've worked on have required substantial desaturation to get things looking right and balanced. Does this happen to you?

I've got some shots of sunlit orange Pomegranate flowers that always overwhelm my camera's sensor/processing of jpegs. Shooting Raw and processing in ACR I have to resort to finding the right combination of white balance, HSL adjusts, a very large pulled down luminance curve shaped like a shepherd's hook along with using Adobe Standard profile instead of dual and single illuminant custom DNG profiles I made. I have to do the same with sunsets as in your case.  

[Bart_van_der_Wolf]

Hello again Fine_Art and other RT fans...Im back with a few new questions about RT:

What is the preferred method to reduce red channel highlight values? Imagine a hot sunset with reds that get clipped or nearly clipped in the histogram.

Hi Paul,

First establish the cause of the (near) clipping. Is the Raw data clipped (look at the Raw data histogram)? If not, then the White balancing may be the cause of e.g. Red channel boost.

That can be addressed by reducing the Raw white points value (Raw tab), which does a linear gamma reduction of the white point before White balance can boost it again, thus avoiding huge clipping. It is only the beginning of course, because you now need to review your color and exposure tabs adjustments again.

Cheers,
Bart