What is linear DNG and how to use it in your postprocessing workflow?

by Wolf Hauser

Linear DNG is a file format for partially developed RAW images: it allows performing different parts of the RAW conversion in different photo editing applications This allows combining their strengths and opens new creative possibilities.

Recap the basics: RAW conversion, linearity and DNG files

First, what’s a RAW image? To understand that, we first must note that the image sensor in your camera does not “see” the image as it is displayed on the screen or saved in the JPEG file. A lot of sophisticated digital image processing is required to transform the signals that the sensor measures at each photosite into a beautiful color image. The unprocessed data is called “RAW data” and the processing is called “RAW conversion”. Shooting RAW images means saving the unprocessed data, directly out of the sensor, either in addition to or in place of the processed image.

In some way, RAW images can be compared to the negatives of analog photography: they contain all the information that was captured when taking the photo, but more processing is needed to see the final picture.

Why should you shoot in RAW images?

What’s the point of saving data that still needs to be processed, rather than ready-to-view JPEG files? There are two reasons.

1. The first is image quality. As mentioned above, the RAW conversion is quite complex. Doing it outside the camera allows using more sophisticated algorithms, which yield better images.
2.The second reason is that it allows you to manually control certain aspects of the RAW conversion, such as exposure, color, and contrast.

Once again, a comparison with analog photography might serve as an illustration: JPEG files are a bit like Polaroid. You have them instantly—that’s a huge advantage. But their image quality cannot match that of negative film. Moreover, with negatives, you can adjust exposure and color to your liking while developing the prints in the darkroom.

Now you might say that it is possible to adjust the color and exposure of JPEG images. That’s true, but, like film negatives, RAW files contain much more information than JPEGs: more dynamic range, more colors, and more gray levels to preserve smooth tone transitions.

RAW data is linear

Now, what does that mean? Mathematically, a linear system has the property that scaling its input by a certain factor is equivalent to scaling its output by the same factor, i.e., f(αx) = α f(x) for all α. Think of a photographic lens: if you double the intensity of your studio lighting while keeping the aperture unchanged, the amount of light that hits the image sensor in your camera will also exactly double. That means the lens is a linear system. But the pixel values in the JPEG files recorded by your camera do not simply double in that case. Dark areas are more impacted than bright areas. Some colors may change. This means that the digital camera is not a linear system.

When you adjust exposure in postprocessing, you want to obtain exactly the result you would have obtained by changing your studio lighting. And for that, you must make the adjustment in the linear domain.

As we already have observed, the lens is linear. The sensor is also linear—at least if no saturation occurs—and thus the data in the RAW file is also a linear function of the amount of light in the scene. Even the first steps in RAW conversion are linear. But at one point during the processing, nonlinear steps take place. In particular, the color rendering and the gamma correction are non-linear functions. You want your exposure adjustment to be applied before these non-linear functions and JPEG files don’t allow you to do that. With RAW files, however, where the entire RAW conversion is applied in postprocessing, adjusting the exposure in a linear way is possible. The same applies to white balance, highlight recovery, and many more.

Finally what is DNG?

The acronym stands for Digital Negative and designates a RAW image file format designed by Adobe. Their goal was to have an open standard for storing RAW data, as an alternative to the proprietary formats used by camera manufacturers. When you convert a proprietary RAW file (ARW, NEF, CR3, …) into DNG, this only changes how the pixel values are stored, but the values themselves are unchanged.

The DNG format also provides a standardized way for storing metadata, called XMP (Extensible Metadata Platform). These metadata include such things as shooting parameters, copyright, and star ratings, but they also contain settings for Adobe’s RAW conversion algorithms.

Now that we have recalled the basics, we can introduce Linear DNG: While (normal) DNG files store unprocessed RAW data, Linear DNG files store intermediate results of RAW conversion, with some processing steps applied—but only linear ones. Therefore, they provide the same flexibility as RAW files for adjusting color and contrast.

Why would you want to save intermediate results of RAW conversion into files? Because it provides great flexibility in your workflow. It allows you to freely combine the strengths of all your favorite tools.

Objective and subjective processing steps

RAW conversion consists of many steps. At DxO, we classify them in “restoration” and “rendering”. Restoration stands for the algorithms that serve to restore the perfect image from imperfect RAW data. This obviously includes demosaicking and denoising, but also corrections of optical flaws such as lens shading, geometric distortions, chromatic aberrations, and lack of sharpness in the corners of the picture. Rendering, on the other hand, stands for the algorithms that produce color and contrast. This includes basics such as white balance, exposure, selective tones, and unsharp mask, but also more sophisticated algorithms such as HSL, micro-contrast, or DxO ClearView. It also includes effects such as Nik ColorEfex, Nik SilverEfex, or DxO FilmPack.

Restoration is generally considered to be an objective task, in the sense that there exists a ground truth and the goal is to come as close as possible to this ground truth. For example, in demosaicking, we want to restore the missing colors as if the sensor had observed them. In denoising, we want to restore the pixel values as if the picture had been taken with better lighting and at lower ISO. Computers tend to outperform humans in these tasks.

Rendering, on the other hand, is fundamentally subjective. There is no such thing as a perfect exposure or perfect white balance. Trade-offs must be made between highlights and shadows, and the goal is not necessarily to faithfully reproduce the original color, but rather to express a certain atmosphere or emotion. Even if computers become better at guessing what rendering humans might appreciate, many photographers want to take control over their rendering.

You don’t recognize famous photographers by their way to denoise their images—you recognize them by their very individual way to treat color and contrast.

That’s why 90% of the adjustment sliders in a typical RAW conversion application are connected to rendering and only 10% to restoration, even if the application might spend 90% of the compute time on restoration tasks.

Spitting up the RAW conversion

Just like the definition of good restoration is objective and that of good rendering is subjective, electing the best tool for applying restoration is rather objective while selecting the best tool for applying rendering is very subjective.

For restoration, journalists and amateurs around the world have used test scenes and evaluation targets, they made objective comparisons and confirmed that our optical corrections are the most accurate and that our DeepPRIME technology can remove more noise and at the same time preserve more detail than any other RAW converter on the market. Regarding restoration, DxO should be an obvious choice for all photographers seeking to maximize the quality of their pictures.

But for rendering, the choice is, well, more subjective. Certain ways of adjusting color and contrast appear intuitive to certain people. It is also a matter of habit. Your individual style might be influenced by the tools you are familiar with. Changing the RAW converter would require learning new recipes or even developing a whole new style. We acknowledge that switching from Adobe Lightroom or Phase One Capture One to DxO PhotoLab might be prohibitive for some—just like people don’t switch from Canon to Nikon or vice versa because they are used to finding certain buttons in certain positions.

Wouldn’t it be nice if you could combine the best of both worlds? Apply DxO’s leading image restoration algorithms and at the same time keep your habits and your favorite tools for applying the rendering? This is where Linear DNG, DxO PhotoLab, and the all-new DxO PureRAW come into play.

In DxO PhotoLab, in addition to introducing DeepPRIME, we have reworked our entire RAW conversion procedure, reordering the processing steps in such a way that all restoration algorithms are applied in the linear domain. We now propose to export this intermediate result, with all restoration steps applied, into a linear DNG. This file can then be opened in any other RAW converter supporting Linear DNG in order to apply the rendering. For best interoperability with Adobe, we even use the XMP metadata to tell their RAW conversion algorithms that they should no longer apply denoising or lens corrections.

This is different from exporting JPEG or even 16-bit TIFF, where non-linear processing steps are applied, and part of the information is lost. The pixel values in the Linear DNG are still in the linear domain, they are still in the native color space of the sensor, like in a true RAW file. When you open them in, for instance, Adobe Lightroom, and make your adjustments there, the color and contrast are indistinguishable from what you would have obtained by opening the original RAW file in Lightroom. But the image quality is objectively better thanks to DxO’s restoration algorithms.

To make our superior image restoration even more accessible, we have crafted an entirely new application: DxO PureRAW. It includes the same restoration algorithms as DxO PhotoLab, packaged in an intuitive user interface, and is intended as a preprocessing step for those who want to stick with their familiar tools for adjusting color and contrast.

DxO PureRAW is like an upgrade to your camera: it improves the objective quality of your pictures. It does not change the way you make your adjustments. It lets you keep your habits and your style.

DxO PhotoLab

The best photo development software. It's that simple.

$ 219