True 3D LUT Calibration
LightSpace CMS works with true 3D colour management, both for profiling and final calibration, based around an unmatched 3D colour mathematics engine for uncompromised final calibration.
Some alternative calibration systems that have evolved from home use applications have attempted to adopt 3D colour management based workflows, but lack the advanced colour mathematics required for true and accurate operation. This results in a 'guess-work' based approach to calibration that inevitably produces inferior final results.
There is also no need to profile with a 3D cube to then be able to generate true 3D LUT for calibration. If you have a perfect display (ideally with near perfect RGB Separation and good RGB Balance) you can actually profile just the grey scale and single R, G, B patches, and then generate a perfect 3D LUT that would control not just the grey scale, but gamut as well. Very, very few calibration system understand this fact.
But, different displays do have different requirements for profiling, and one approach will not work across the board. For displays with a good level of RGB Separation (RGB decoupling) and reasonable RGB Balance (white point) a Quick Profile approach will work, while for displays with poor RGB Separation and poor RGB Balance full 3D cube profiling will be required. Additionally, the number of points used within the 3D cube profiling data set is key to getting a final accurate calibration.
Alternative systems that use a guess-work approach to calibration will struggle to generate the necessary accuracy, as the number of points measured are too few, and too spaced apart, and combine profiling with calibration which is not an ideal approach as it locks the profiling to the expected end colour space (gamma and gamut). For further info on this see later.
If the profile has the correct number of points for the display's needs, and that can mean using a minimum of a 17 point cube (17x17x17 cube), a 21 point cube (21x21x21) or a Hybrid 1D + 3D profile for a display with poor RGB Separation, RGB Balance and/or low gamut, a final calibration LUT can be generated that has a smaller number of points, but that retains the inherent accuracy of the higher resolution of the profile.
When profiling with a small number of points it is virtually impossible to generate a calibration LUT that will be anywhere near accurate enough.
An example of this is the use of 125 point 3D cube calibration performed by some alternative systems. What 125 point calibration actually means is a 3D cube that is 5x5x5 points. For combined profiling and calibration this is no where near accurate enough.
However, a 5x5x5 calibration LUT can be accurate for final use, when combined with a higher resolution 1D LUT for grey scale management, if the initial 3D cube based profiling was done at a much higher resolution than the final LUT resolution - which is why a minimum of a 17^3 profile is really needed for accurate profiling.
Note: This is not the same as using fewer points initially, and interpolating the final results - that will never give as accurate end results.
The smaller Lumagen Radiance LUT boxes are an example of this, where a 21 step 1D LUT is use for grey scale management, in combination with a 5 point 3D LUT for colour management, which can produce good results when the initial profiling is performed with a higher resolution cube.
The above images show a 5x5x5 cube, as used to perform 3D cube profiling by many alternative calibration systems, compared to a 17x17x17 cube, as used by LightSpace CMS for profiling. The difference in profiling accuracy is obvious to see. With LightSpace's ability to use 21x21x21 (or greater) cubes the difference is yet more obvious.
The following images show the difference this makes in a real-world situation.
The above LUT for a real display shows the calibration accuracy in 5x5x5 and 17x17x17 data. The complex nature of the calibration is due to the display having a very non-linear RGB response, which is common for many home TV displays. From the above comparison it is easy to see how the 5x5x5 cube is unable to hold the necessary level of accuracy.
The following images show this inaccuracy in more detail - the first image shows a 5x5x5 LUT expanded to a full 17 point LUT, compared to the real 17x17x17 LUT shown in the second image.
The difference is obvious... the 5x5x5 LUT just can't hold the required data accuracy.