We've seen that some LUT using systems are better than others at interpolating the missing data, and it has to be said some systems use larger 3D LUTs than the 17 point ones that are the industry standard - although I'm not convinced this is necessary at all. A 17 point 3D LUT with good interpolation is impossible to distinguish from a larger LUT.

But, the making of LUTs does tend to be shrouded in a lot of unnecessary secrecy.

The simple fact is a LUT is a way to make one display medium match that of another, as outlined in the following diagram:

• Scanner: calibrated to Cineon/DPX 10bit LOG format
  (Although as we have seen, this is not a major requirement...)
• Display Device: calibrated to manufacturer (or alternative known) specifications to allow LUT to work correctly
• LUT: calibrated to accurately display a 'print' or other target colour space image from a Cineon/DPX image on a 'calibrated' display device
• Film Recorder: calibrated to accept Cineon/DPX specification images for accurate output
• Film Lab: calibrated to Kodak specifications and accurate to AIM

If these requirements are met the workflow will be as expected, and the image viewed through the LUT will match the final film projected image.

LUT Generation
The basic need when building a LUT is to develop a conversion matrix that takes the image attributes of one display and maps them to a different display...

So, in simple terms, the approach taken is to use a selection of reference images and use them to profile the first display, and then use the same images to profile the second display, and generate a conversion matrix based on the difference between the two display profiles.

This is the basis for all LUT making.

Systems that attempt to automate the LUT building process tend to have a range of colour and Grey reference patches, from 0 to 100% colour in steps, taking in the colour space defined within a 3D cube.

The above 5 point cube (if the flats are equally spaced - which is often not the case!) would take a total of 125 flats to map. Often, 1000 or more flats are actually used.

Each flat is passed through each workflow (DI system, routing, interface boxes, digital display - projector/monitor) and the displayed colour value is measured with a colour analyser, which can be measured off-screen, or via film density for film print.

After measurement, there will be two lists of colour values - one being the target or aim list, the other being the DI or grading display values to be matched to the target values. The LUT is therefore the matrix used to match the DI display to the target or aim display values.

Obviously, the more colour flats used the better the final calibration is likely to be, and LightSpace has many options...

However, the reality is that there are no automatic measuring systems available that can easily and accurately see small colour and brightness variations, especially within the dark range of the test images. And this means the generated LUTs are often not as accurate as you might think. This is true of any automated LUT process, including the Hubble and LightSpace system sold by Light Illusion - and is why we started our LUT management development with the necessary manual tools for accurate LUT generation and adjustment by eye!

The human eye is very good at seeing image differences, and will quickly spot inaccurate LUTs, and so can be used to improve inaccurate automatically generated LUTs...

Next Page - LUT Building By Eye