My New Approach To RAW Processing

Raw processing is an ever evolving cycle of changes.
We have a multitude of new raw developing software being launched and new Ai technologies introduced to better handle noise reduction and sharpening.

I have tried most of them at some point, but always end up back where I started - with LR and PS.
The Adobe offerings are solid and well developed and I personally prefer the output I get with this software.
Others may have other preferences.

This thread will show how I approach my developing at the moment.
Perhaps there will be something useful in here that others may want to try for themselves.

Today I use LR for the main editing, but round-trip the raw file to DxO Photolab 5 for its raw developing capabilities.
There is something DxO has figured out with regards to the de-mosaicing of the raw file that results in a sharper file with better colour and less artefacts.
As far as I understand, DxO uses lens profiles in combination with the de-mosaicing to calculate the pixel data.
The software downloads and applies these profiles when you open a raw file in the software.
Others may chime in with a better explaination.

Whatever the tech, I like the outcome.

DxO offers two types of software that offers this process - DxO PureRaw 2 and Photolab 5.

DxO PureRaw 2 is a barebones demosaicing software with no user adjustable input.
You load the raw file and it saves out a DNG file with its magic applied.

Photolab 5 offer the same tools that PureRaw 2 uses (and much more), but give the user some control of how they should be applied.
You can adjust the amount of global lens sharpening and choose what noise reduction algorithm to apply etc.

I have created a preset in Photolab 5 with my chosen settings.

In adition to the DxO processing I also apply a linear camera profile as I feel I get a nicer end result with this.
Many users will have their own preferences for camera profiles - especially Fujifilm users as there are many creative profiles on offer.

These days I shoot with the Canon R5 and I find the camera profiles for the Canon pretty awful to be honest.
Very contrasty and some strange colour casts to my eyes.
This will be very personal and what I like may not be what others are looking for.
I come from medium format (Fujifilm GFX100) and have been used to the characteristics of the wider dynamic range and 16bit files and the Fujifilm profiles.
The process I have outlined below will apply equally well to other raw files from other brands.

Changing cameras can be a headache if you have a certain look you have come to like.
New sensor designs often change the colour science and ‘look’ and it can be hard to tweek a new camera to fit your old preferences.
I am very much going through this process with the Canon R5 at the moment.

So for now I have settled on the following process.

1. I will load all the raw files into LR and cull the shoot.
2. The selected images will then be sent to DxO Photolab 5 where I apply my preset.
3. The raw files are then converted to DNG with the Photolab adjustments applied and returned to LR

The round trip takes around 30 seconds per image.

To round trip the raw file from LR to PL and back - do the following:

Select the raw file in LR - the adjustments you may have made in LR are ignored.
Goto File - Plug-in Extras - Transfer to Photolab 5 (same process for DxO PureRAW)

In Photolab 5 I use the following settings (I have saved a preset with the settings):

Vignetting set to 100 - removes the darkened corners, if present.

- DeepPRIME noise reduction ON
- Lens Sharpnes is ON and set to +0.69 (this can be adjusted to taste)
- Chromatic Aberration ON
- Unsharp Mask OFF
-
No other adjustments are applied.

I have saved this as a preset named DxO PureRAW

Then export the file back to LR with the LR Export button.

It will prompt you to choose a setting - I use this:
Click Export

Once the file is back in LR - switch OFF sharpening and Noise Reduction.
Also dissable Profile Corrections as that has already been applied by DxO
Again, I have a preset saved in LR with my preferred settings.

From here you simply edit the raw file as normal.
Choose a camera profile.
Adjust to taste.

- Important -
You must make sure that the DxO DNG isnt’t overexposed.
LR will not be able to recover as effectively as it normally does.
If the raw file is too ‘hot’, bring the ‘Exposure Compensation’ down in Photolab before exporting to LR.
Then simply add the lost exposure back in LR.

Let’s look at what is achieved by this process.
In the following examples the normal LR file is on the left and the DxO processed file on the right.

Remember that NO sharpening and NR is applied to the DxO files in LR.
The LR de-mosaiced file has normal Unsharp Mask and Color NR applied.

I will use some simple holiday images in these examples.

Example 1:

100% crops
Notice the finer detail in the rocks and plants - almost as if an anti-aliasing filter has been removed.

Example 2:

DxO also does a good job of removing aliasing artefacts and bring out the finer details in fabric.
The DxO processed files are more colour accurate and neutral to my eyes.
Notice the magenta/purple cast in the regular file on the left (Canon Neutral profile):
200% crop:

Another example where DxO has removed artefacts and brought out finer detail.
200% crop

Finer detail in foliage

Now let’s look at the noise reduction from DxO.
The DeepPRIME algorithm is very good and combined with the DxO lens sharpening it does a great job of reducing noise whilst retaining fine details.

ISO 1600 example

Notice how well DxO has managed to retain detail in the hair and fabric whilst reducing the noise.

Another example at ISO 1600 (I haven’t shot any higher yet).

Now, there is one downside to this process.
The DNG files that are created by DxO are really big.

A standard Canon R5 CR3 file is around 45-60mb.
The DxO DNG files vary between 130mb and 180mb.

That is a big increase in storage requirements.
Whether that matters is down to induvidual circumstances.
Personally I want the best files I can get so I will accept the memory hit.

In summary I find the DxO proccessed files to have better colour, more detail and lower noise.
Combined with the linear camera profile I feel I get a better result than with LR alone.

I am interested to hear what others do.
Please feel free to discuss here.

Mads,

This is almost exactly my workflow too!

A couple of points:

1. I set LrC to save metadata to XMP files and apply a preset on import that sets the Camera Profile (I use Color Fidelity), applies Lens Corrections, a slight contrast boost curve, a slight Vignette, and Auto Settings to provide me with a reasonable starting point.

2. For selected images (based on composition and sharpness etc.) I then set exposure and whites and blacks to taste / avoid clipping. I may also tweak other settings as part of my selection process - a kind of “can this image be made to work” test.

3. On return from DxO PhotoLab, the DNG files in LrC have no noise reduction (except Color) or sharpening applied. Also Lens Corrections are zeroed. This appears to be done by the DxO plug-in, as before the round trip they were set in LrC. The other settings from the original RAW file in LrC are also applied to the returning DNG automatically.

Many thanks for going through all the effort. My workflow involves DxO Photolab as well (though I do my initial culling using Canon's DPP).

IMHO it's DxO's Deep Prime NR that truly enables use of Canon's crop of slow(er) RF lenses (like the 100-500 +/- 1.4x) which I truly adore.

R2

R2D2 wrote:

Many thanks for going through all the effort. My workflow involves DxO Photolab as well (though I do my initial culling using Canon's DPP).

IMHO it's DxO's Deep Prime NR that truly enables use of Canon's crop of slow(er) RF lenses (like the 100-500 +/- 1.4x) which I truly adore.

R2

Yes, there is some 'magic sauce' that DxO is able to apply to the raw files.
As you say, it really expands the capabilities of slower lenses or shooting in dark conditions.

I don't do my editing in DxO beyond what I have already highlighted as I personally prefer the working environment of LR.
I like the masking tools in LR and the HSL sliders work better for my needs than the colour wheel in Photolab.

LR is better at recovering highlights in my experience.

It is great that we now have so much choice and the tools are improving at a fast pace at the moment.

cfieldgate wrote:

Mads,

This is almost exactly my workflow too!

A couple of points:

1. I set LrC to save metadata to XMP files and apply a preset on import that sets the Camera Profile (I use Color Fidelity), applies Lens Corrections, a slight contrast boost curve, a slight Vignette, and Auto Settings to provide me with a reasonable starting point.

2. For selected images (based on composition and sharpness etc.) I then set exposure and whites and blacks to taste / avoid clipping. I may also tweak other settings as part of my selection process - a kind of “can this image be made to work” test.

3. On return from DxO PhotoLab, the DNG files in LrC have no noise reduction (except Color) or sharpening applied. Also Lens Corrections are zeroed. This appears to be done by the DxO plug-in, as before the round trip they were set in LrC. The other settings from the original RAW file in LrC are also applied to the returning DNG automatically.

So it sounds like you do more of your processing in Photolab than I do.

I have tried many raw developers over the years, but always get drawn back to LR and PS.
For my commercial work I need PS and when shooting in the studio I need tethering.

No, not at all 😀.

I have a preset much like yours in DxO PhotoLab and just use it just for demosaicing, lens corrections and denoising with Deep Prime. All my editing before and after the trip to DxO PhotoLab is done in LrC. Although, I will occasionally use Photoshop layers and the TK8 plug-in for masks etc. for any final polishing if needed.

cfieldgate wrote:

No, not at all 😀.

I have a preset much like yours in DxO PhotoLab and just use it just for demosaicing, lens corrections and denoising with Deep Prime. All my editing before and after the trip to DxO PhotoLab is done in LrC. Although, I will occasionally use Photoshop layers and the TK8 plug-in for masks etc. for any final polishing if needed.

Very similar indeed.
I too really like the TK8 panel in PS.

I fully agree with everything you have written. I was a big LR user starting with LR 1.0 and upgraded to every version since 1.0. After I bought my R5 I was not happy with the LR output. I tried several different RAW programs and finally settled on DXO PL 4, now 5. I probably do more in PL than you do, I enjoy some of their tools.

Thank you for your post, great information on using DXO.

One major difference is the processing power required for DXO PL. LR pretty much worked fine with my R5 RAW Images on my 6 year old XPS I-7 machine. DXO was unusable until I doubled my memory to 32GB.  Still the exporting was a slow. I finally upgraded to a new XPS with the new I-7 12700 32 GB of memory and a 8GB GPU. Now the images complete in just a few seconds.

Kokopelli_Rocks wrote:

I fully agree with everything you have written. I was a big LR user starting with LR 1.0 and upgraded to every version since 1.0. After I bought my R5 I was not happy with the LR output. I tried several different RAW programs and finally settled on DXO PL 4, now 5. I probably do more in PL than you do, I enjoy some of their tools.

Thank you for your post, great information on using DXO.

One major difference is the processing power required for DXO PL. LR pretty much worked fine with my R5 RAW Images on my 6 year old XPS I-7 machine. DXO was unusable until I doubled my memory to 32GB. Still the exporting was a slow. I finally upgraded to a new XPS with the new I-7 12700 32 GB of memory and a 8GB GPU. Now the images complete in just a few seconds.

Glad you found it useful.

For the record, I use a 2021 MacBook Pro M1Pro with 16GB ram and 10 core CPU and 16 core GPU.
This is the midrange model with fairly basic spec.
It has always felt fast and responsive with all my production software.

This thread also exists over on the Medium Format forum as that is were I used to post.

There is some interesting posts there where some with more technical insights are offering their insights.

https://www.dpreview.com/forums/thread/4668157

DanInSoCal wrote:

Thanks for the detailed review and workflow.

I know a lot of people like DXO. I am not one of them, unfortunately. I tried the PureRAW plugin and I liked some things:

- Excellent noise reduction
- Excellent reduction in color casts and lens artifacts/distortion

However, this was completely outweighed by the negatives:

- Slow to the point of unusability for large numbers of files (20+ sec/image)
- Over-saturation (you can see this in your examples, e.g. in the orange)
- Huge DNG intermediate files
- Over-sharpening, to the point of "adding" details that were not even there

I found too PureRAW not to my liking for some of these reasons - answer was Photolab 5 where can fine tune and turn off sharpness if not wanted or turn all down.....

+1

After I bought my R5 last year I continued to use LR (LR user since version 1). I was not happy with the results I was getting with LR and decided to try different RAW tools. After several months of testing I settled on DXO PL4. I agree with you the product is resource intensive. I was using a 6+ year old XPS with a I-7 6700 with 16GB. I doubled my memory to 32 and the software ran better, but the processing of image was slooooow. The results were good so I continued to use DXO and upgraded to version 5.

Last wee I was in Costco and saw a sale on a Dell XPS 8950, I-7 12700, 32 GB, and GeForce 3060 with 8GB, Killer™ WiFi 6 (2x2/160) Gig+ and Bluetooth® 5.2. Best of all the machine was $400 off. I am very impressed with the new Alder Lake CPUs the first Intel CPUs using 10nm process.

So how does the system perform with DXO - all I can say is amazing. Takes a few seconds to process an R5 RAW image (not cRAW). So fast, that I don't even have time to start on my next image before the first is complete. The new system is great with DXO PL5, Topaz and Luminar Neo.

I know my solution is not possible for everyone, but I am thrilled with the new system. The upgrade cost less than an R7.

Kokopelli_Rocks wrote:

After I bought my R5 last year I continued to use LR (LR user since version 1). I was not happy with the results I was getting with LR and decided to try different RAW tools. After several months of testing I settled on DXO PL4. I agree with you the product is resource intensive. I was using a 6+ year old XPS with a I-7 6700 with 16GB. I doubled my memory to 32 and the software ran better, but the processing of image was slooooow. The results were good so I continued to use DXO and upgraded to version 5.

Last wee I was in Costco and saw a sale on a Dell XPS 8950, I-7 12700, 32 GB, and GeForce 3060 with 8GB, Killer™ WiFi 6 (2x2/160) Gig+ and Bluetooth® 5.2. Best of all the machine was $400 off. I am very impressed with the new Alder Lake CPUs the first Intel CPUs using 10nm process.

So how does the system perform with DXO - all I can say is amazing. Takes a few seconds to process an R5 RAW image (not cRAW). So fast, that I don't even have time to start on my next image before the first is complete. The new system is great with DXO PL5, Topaz and Luminar Neo.

I know my solution is not possible for everyone, but I am thrilled with the new system. The upgrade cost less than an R7.

Glad to hear this result since I am in the process of a similar computer upgrade.
Ryzen 9, 32GB, 1TB SSD, RTX3080, 2TB HDD. Have not had a chance to try it out yet.
Now my computer costs more than my camera system

Question: Is there any setup ritual required to get PL5 using the graphics card?

Bert