Research | Colour Management

Making my own D-Cinelike>Lin LUT

August 1st, 2018


Based on all of my prior research in to both the perceived and visible contrast ratios of both D-Log and D-Cienmalike gamma within the DJI OSMO, I have confirmed that venthough D-Log is better for grading, its display transform is not truly logarithmic and thus cannot achieve a 0 – 255, limit in 8 bit space.

This of course, has the negative impact of not being truly scalable to Linear, Scene Referred colour space using Sobotka and Chambers (2018) colour transform methodologies. To this end the following video documents my decisions throughout the LUT building process, and sheds light on my approach to the capture of light/exposure information; analysing and recording 8 bit RGB values; and, the normalizing and Linearization of this same data to a 1DOSMO_to_LIN LUT (Look Up Table)

  • Validity of Research 95%
  • Complexity of Content 100%
This is a photo of the very basic mounting rig I had to create to angle both the 50% grey card at 45 degrees to the lens, and to keep the OSMO relatively static in spite of its lack of tripod mount. It proved to very effective in the end.

Practice as Research


Even though Paul Chambers had recently proved that a mid range DSLR could capture lighting data that could be converted using Troy Sobotka’s Math, the OSMO had a lot more doubt hanging over its head. Despite its availability, low price and smoothness of movement, it has a multitude of issues that may infringe on its ability to be converted as per mid range SLR models.

For example, the basic OSMO model has no aperture control for the camera so only Shutter speed can be used to stop up or down. Manual controls can fix white balance and ISO, but the Zebras seem far from accurate following my previous tests into D-Log and C-Cinelike profiles; there is no way of mounting the camera to a tripod; and, the lens is so wide that it has noticeable barrel distortion and edge to edge luma shift issues.

To this end, in the video you will see that I not only have to create some “custom” rigging, but also apply some guest work to EV stop increments above -2 +2 stops, as the OSMO readout stops before the exposure does. Additionally, I have to trust the Histogram for exposure readings; and, lastly have to put the lens cap on the camera to ‘zero’ the exposure at black; even in D-Cinelike.

this all said, despite all of these issues, I continued with Chambers and Sobotka’s methodology and it seemed to work none the less. This is as much a celebration of the math as it is my efforts to create accurate and consistent readings that match exactly the manufacturers specification sheets for dynamic range, despite not being able to read or control – accurately – stops above or below 2.0 either side of 0.

Core findings


The baseline OSMO is very awkward to use in a controlled or static environment: no tripod; not aperture; inaccurate readouts (Zebras, Histogram, EV); and, Lens Vignetting.

D-Cinelike still cannot achieve 0

D-Cinelike, despite the histogram readouts to contrary, cannot achieve pure black without capping the lens.

Shutter Speed control gives approx. 0.3/0.4 EV control per increment

Luckily, it seems that despite the fact that the EV display on camera could not display all of the exposure range I recorded, by counting in my head the number of Shutter speed increments, I was able to count back or forwards the limits of EV stop range below and above 0 EV. The math proves that this worked.

Sobotka and Chambers methodology is forgivingly powerful

Even Though my guess work was informed and seemingly accurate in the end, it also seems that Sobotka and Chambers workflow for calculating the math – once captured and measured – is very powerful indeed; converting display referred colour to a virtually linear curve.

The Curious Engine

My Professional Portfolio
Go There Now

Recent Posts

Here are some recent posts that I have written as part of the Digital Media Practice 3 project – ‘FUSE’