| Forums | Register | Polls | Search | Statistics |
You must be logged in to post content on this forum.
Tips and Tricks Light Illusion Forums / Tips and Tricks /  

Inaccurate Grey Scale Results?

Author Steve

#1 | Posted: 5 Nov 2021 10:51 
There has been some user discussion suggesting that ColourSpace can cause colour errors in the resulting grey scale of a generated 3D LUT.
The reality is that ColourSpace cannot introduce any colour error in the grey scale (or elsewhere within the volumetric LUT) by its own actions/processing.

It is simply mathematically impossible.
The only way that an incorrect grey scale can be generated is if the profile data is incorrect/invalid.

Such invalid/inaccurate measurements are obviously possible if the probe in use is not providing accurate readings throughout the volumetric display range, or the display itself is unstable, or if the TPG used is itself inaccurate (as with the inbuilt TPGs in some displays).

It must be remembered that there is no such thing as a 'grey scale' as a separate entity compared to all other 'colours'. All points are equally valid in all display calibration. Consequently, if there are invalid readings anywhere within the measured colour volume - not just the measured grey scale - the resulting calculated 3D LUT will contain inaccuracies.

Due to the understandable desire to assess grey scale as a separate entity to the full volumetric colour, such errors can be more easily visible, compared to the rest of the volumetric colours.
But the errors are due to the original measurement errors, not the ColourSpace colour engine, or generated LUT.

To explain further, with an example:
With low-light readings, patch measurements surrounding the measured grey scale are even more critical, as there are few colours measured around the grey axis.
For example, the readings immediately above black – the next level up – are just 3 readings surrounding the grey scale value, with a default 21^3 profile.
If those 3 patches have invalid/inaccurate readings, they will skew the grey scale result at that level, as well as the actual volumetric colours.
However, the skewed values are totally correct for the (incorrect) measured values.
So if the measured volumetric values are incorrect (irrespective of the grey patch measurement), the resulting grey patch value will also be incorrect.

For information, if the grey scale is accurate to the target colour space pre-profiling, the impact of inaccurate surrounding volumetric patch readings/measurements will be reduced.
However, if the pre-profiling grey scale is not accurate to the target colour space, the result WILL BE PERFECT if the associated volumetric colour patches are measured accurately.
It is ONLY when the profile measurements are inaccurate that the result will also be inaccurate. And that is BOTH the actual grey scale measurement, as well as the surrounding volumetric measurement. Either will cause inaccurate results.

Anyone can obviously test all this very easily by generating synthetic profile data, and assessing the results.
The LUT generated will always be a perfect result, based on the input profile data.
With the Colour Engine within ColourSpace it is impossible for it to be anything else.
We run such tests repeatedly when we make any changes to the ColourSpace colour engine, and every result is always as perfect as it should be.
There are no issues within the ColourSpace colour engine.

This therefore means that when a calibration result appears to show any colour inaccuracies, in the grey scale, or elsewhere, the underlying profile data used to generate LUT will be the cause of such errors (assuming the application of the LUT is not at fault...).

Note: There can be issue with some tristimulus probes due to the filters not accurately matching the CIE observer, causing them to read colour errors in low-light readings vs. highlight. And that means a FCMM probe match will not correct such errors, as the 4 colours measured are bright.
Potentially, ColourSpace's unique MPVM could help with this

For any user that thinks they have inaccurate LUT results, it will be worth testing different LUT Generation options, from the same profile data.
For example, does Peak Chroma show different grey scale results to Hybrid, or Peak Luma, etc?
(Hybrid mode is based on Peak Chroma, but does two things that can potentially help with the accuracy of the LUT's grey scale, and low-light performance, by isolating those areas of the profile, and processing them independently from the rest of the profile's volumetric data. This maximises the accuracy of the grey scale, and helps remove low-light errors introduced by inaccurate probe readings.)

It is also possible to extract only the 'Grey Ramp RGB' patch data from the profile, and seeing if that generates a significantly different grey scale result.
(As a Grey Ramp RGB profile has no 'volumetric' data surrounding the measured grey data, that would help isolate if the volumetric measurements are the issue.)
(The pop-up Point Info box can be used to locate the Grey Scale, and three RGB, patches.)

It may also be worth trying different user generated patch sets, to see if reducing, or increasing, the near black 'colour' patches helps overcome the measurement issues.
(Depending on the measurement issues, either more or less associated colour patches could potentially help overcome the underlying measurement error issues.)

For example, after performing a standard volumetric profile, Save the profile, and then use Manual Measurement to add/delete measurements, and then re-save the profile with a different name.

Note: If using 'user generated' patch sets it is easily possible that the patch set is causing measurement issues, by having more patches within areas that the probe/display combination has issues with.

It should also be realised that if the display's underlying grey scale - pre -calibration - is inaccurate, adding addition points in to the 'grey' axis (RGB matching triplet values) can cause more issues, as the actual target grey axis is not being measured by those patches. This can exasperate the issue when using 'user patch sets', if the probe is not measuring some colours accurately.

As a final point to remember, all LUTs should be verified using the 'Active LUT' capability within ColourSpace, and then re-verified with the LUT uploaded into the display/LUT box.
Unfortunately, many displays/LUT boxes distort LUTs when applied within their own signal processing system.
Such issues can be seen in the Calibration Issues page of the website.

The bottom line is, with ColourSpace ONLY inaccurate measurement data can generated LUTs that show inaccuracies in the calibration result.
(Or issues can be caused by the LUT application not correctly implementing the LUT, as shown by verifying via Active LUT vs. the Uploaded LUT.)
Using the above information it should be possible to locate where/why measurements inaccuracies/errors are being generated.

Steve Shaw
Mob Boss at Light Illusion

Author mrtickleuk
#2 | Posted: 5 Nov 2021 13:24 
Due to the understandable desire to assess grey scale as a separate entity to the full volumetric colour, such errors can be more easily visible, compared to the rest of the volumetric colours.

This makes sense and it raises an interesting question (well, interesting to me!). The "RGB Balance" graph by default is the RGB Balance of the grey scale. But what if there were also equivalent "RGB Balance" graphs for the Red-scale, the Green-scale, the Blue-scale, the Yellow-scale, the Magenta-scale and the Cyan-scale? I'd be very interested to see what they would all look like, and how flat (or not) they would be in comparison. Also, it may illustrate more obviously the point about there being nothing "special" about greyscale (even though it's understandable to want to measure it separately).

Author Steve

#3 | Posted: 5 Nov 2021 13:32 
With the 3D Graphs in ColourSpace, you can actually already see that data.
And if you use the Tangents you can see the error, be it Luma, or Chroma - both in the 3D CIE ands Cube graphs.
(And you can click any points to see the actual point data.)

You can also use the Secondary Execution function to export the measurements into something like an Excel .csv file for external plotting.
(There is an example Batch file program included in the Customer Downloads - the 'Integration Protocols & Examples' download.)

Steve Shaw
Mob Boss at Light Illusion

Author Steve

#4 | Posted: 11 Nov 2021 12:09 
It should also be pointed out that there are numerous options/tools included with ColourSpace to help understand why there are issues within any generated LUT, as well as ways to manage such issues.

  • Augment Data, to enable enhanced Grey Scale profiling & LUT Generation
  • The ability to swap LUT components
  • LUT Adjustment tools, such as the Axis Blend filter, which will remove any colour alteration in the grey axis
  • Alternate LUT Generation processing, such as Hybrid vs. the default Peak Chroma
  • And Focused Patch Sets

The website contains information/instructions on all the above.

And one very good 'test' is to generate a LUT using Peak Chroma, and then a second LUT using Hybrid, as a profile that has no invalid/incorrect measurements will produce an identical LUT...

Steve Shaw
Mob Boss at Light Illusion

You must be logged in to post content on this forum.
Tips and Tricks Light Illusion Forums / Tips and Tricks /
 Inaccurate Grey Scale Results?


Online now: Guests - 1
Members - 0
Max. ever online: 192 [11 Jan 2023 08:39]
Guests - 192 / Members - 0