PhotoLab linear rendering

When using a generated .DCP with linear encoding and black point compensation disabled, Images rendered in PhotoLab do not have linear output.

Instead highlights have a significant aesthetically pleasing filmic rolloff similar to Adobe products, but is undesirable for technical use.

requesting an option to process images entirely neutrally.

Curious if you don’t mind sharing the reason why you need linear encoding?

1 Like

Hello. Just paste the linear curve into your .dcp profile. Profit! :smiley:

If the .dcp profile does not have a tone curve inside, then Dxo uses the adobe tone curve, as you correctly noted

I’m after the tone curve to be linearized for CG work, specifically photogrammetry. Jpeg output is still in sRGB

.dcp profile in use has a linear curve inside.

Expected output in Rawtherapee top, DXO has similar swatch samples except white which is significantly darker. Both applications are using the same .dcp

1 Like

Yes, I also noticed this feature. Dxo does not work correctly with .dcp profiles. Before the wide gamut update, the black point of the .dcp profile was not read correctly.

1 Like

If you do CG work, I would expect all applications you woudl use have built in tools for linearization, color conversion, both gamma and gamut tools and all else one might need. You can use free version of Fusion for example and you get all you need. 32bit float, linear, plus all the tools you might need for compositing or CGI work. Weather image comes from internet in sRGB or CGI application or RAW should matter little, since you have to unify them anyway. And program like Fusion allows for that. I don’t expect DXO PhotoLab which is primerally made for getting most out of RAW data to be more than that. It does its niche job well. You can save of course linear DNG, but I fail to see much use for it in CGI work. Other than HDRI maps, there is no need to do much more than do your conversion in compositing app itself. Like Fusion. Give it a try.

I was just playing around with ordinary compressed JPEGs you can find for free online to do a test CG scene for car paint style material I was making in Fusion and all was sRGB, compressed 8-bit source images, but you can linearize it all and conform it to all you need in Fusion with built in tools and if you work in 16-bit or 32-bit float, linear color space, native to Fusion and most other CGI apps, or use ACES, you should be fine. Unless you are doing something strange that I never done and you need something super specific, I think you are trying to use wrong tool for the CG job.

Its actually pretty incredible how far you can abuse an image in 32-bit float, linear space and get away with it. Even heavily compressed, sGRB, jpegs as source material.

…of course that is not the job DXO PhotoLab was ever made for and Fusion or other apps can’t do what PhotoLab can, so we use the tools we need to get it done.


What exactly do you mean by that?

Apologies I haven’t been clear.

I want a RAW processor to render images without any additional aesthetically pleasing additional contrast, S-curves or highlight rolloff that matches a .dcp profile.

I’m not trying to convert sRGB or RAW files to to linear gamma image output. That process occurs further down the production pipeline.

The linear request is regarding the Tone curve (Linear) and final image output accuracy. Photolab is adding a highlight rolloff and some black adjustment.

ACR tone curve RED, Linear tone curve Dotted white.

Lumariver correctly exposed image with a linear tone curve, and Automatic Black Subtraction turned off.

I need to process 10’s of thousands of RAW files to sRGB jpeg for processing for scientific use and be cost, storage and processing respective while I do so. There are command line tools, and I currently use RawTherapee in a parallel processing batch tool to do so.

  • RAW image to sRGB 8-bit jpeg (10-bit jpegXL would be nice though).
  • Regular support for new camera sensors and lenses.
  • Visual feedback
  • Easy for junior artists to use.
  • Exposure adjustment
  • White Balance adjustment
  • Correcting Vignetting
  • Correcting chromatic abberation
  • .dcp color profile support
  • Batch processing
  • Noise reduction

DXO Photolab does all these things, it just messes up the white values.

Those Jpegs you are using have all of the S-curve adjustments, blackpoint subtraction and compressed highlights I want to avoid.

Lumariver manual: .dcp black point subtraction

Adobe’s raw conversion products supports automatic black subtraction, which means that depending on subject more or less of the blacks will be subtracted to create blacker shadows and slightly more contrast. Among third-party raw converters it varies if automatic black subtraction is supported, and if it is it’s not necessarily the same as Adobe’s as their algorithm is proprietary and not part of the DNG specification.

All(?) Adobe’s own profiles have automatic black subtraction flag enabled, so if you are making a general-purpose profile for Adobe it’s generally wise to enable it. If you want 100% predictable result as for a reproduction profile, or if you will be using the profile in several different raw converters and want the same result you should disable it. Making a stronger shadow dip in the contrast curve gives a similar effect.

I see. What happens if you turn all all adjustments except for noise reduction and demosicing in DXO and only load up calibrated profile or DCP profile that you can either create yourself or use one that is provided by others. Than export DNG linear. Where does that problem in the chain happen?

Here is an image RAW file from Sony, I turned off everything and only used DCP profile from Adobe that should match the natural color profile in Camera.

Here is using generic rendering for the camera that ships with DXO.

And here is calibrated profile with all the adjustments for more corrected image.

Is the first example the thing you are looking for?

Sorry for going off-topic slightly.
From Lumariver Profile Designer :
Automatic black subtraction (DCP only)
Purpose: specify if the DNG profile should instruct the raw converter to perform automatic black subtraction (or not) before the profile is applied.

That’s probably about ‘DefaultBlackRender’ DNG tag (so the context is different), which is not part of a DCP profile. Perhaps I got confused because ‘profile’ was being used with two different meanings – DCP profile as described in DNG specs, and a higher level, Photoshop profile. On the demosaicked RGB (or monochrome) data you can still do some “black point correction” (e.g. if the light was very scattered) but that’s a different story.

Most cameras report black levels in their raw file metadata (mostly in a proprietary way, it may be global, per channel, calculated per row based on masked pixels, …), but some do black level subtraction inside the camera, like many current mobiles. Raw ‘black level’ reports the dark current in camera electronics – see e.g. US8698922B2 - Black level correction for imaging pixels - Google Patents . I think all photography software does ‘black level subtraction’ on raws by default. Otherwise photos from most cameras would look quite bad. For example, many Sony cameras with 14-bit raw encoding use black level of around 2000 (and there is very little data below this value in the raw files).