When to use an LUT?

What is an LUT?

LUT, or lookup table, is a sophisticated color correction technology that samples hundreds or even thousands of colors, loads them into a table, compares the measured values with the expected values for a specified gamut at each location in the color space, and then assigns an appropriate correction to each.

LUT's have been used for years in the professional world. One of the most important uses for the studios has been gamut matching. For example, the gamut of film is much wider than Rec. 709, which is used for Blu-ray and broadcast HD. Therefore, in the mastering process for home video products the original gamut would have to shrink to fit within the Rec. 709 boundaries. A LUT is used for this process. Another use would be post-production color correction. Sometimes a director wants to give a film a different look than what is originally captured on film. A custom LUT can be used to globally alter the appearance of the film prior to release.

LUTs have only recently been made available to the consumer world. For consumers the main use of an LUT is color correction as part of the video calibration process.

Why use an LUT?

The reason that LUTs are sometimes necessary for standard color correction is simply that a standard color management system (CMS), which operates only on the gamut boundary, may not actually reduce overall color error. In fact, it may even increase it! This is a little known fact, because the main consumer trade publications have pretty much ignored this aspect of color performance. The dirty little secret in the consumer video world is that a display that shows excellent color performance when measured by a standard CIE chart—even including additional information about color luminance—can actually perform quite poorly inside the color space, which a standard CIE chart that plots measured color only on the gamut boundary cannot show. Consider this passage from a CNET review of the Sharp XVZ-20000 DLP projector

The CMS, or Color Management System, is a double-edged sword. I found I could dial in the colors to near perfection, only to find color severely desaturated. Increasing the color control to get back the saturation resulted in strange-looking skin tones, so I reverted back to the factory settings in the CMS menu. This is a common problem with CMS systems.

The phenomenon the reviewer is describing is indeed common. The CMS achieved accurate colors on the gamut boundary, which is all he measured, but is is a different story inside the color space. For example, many reviewers have judged that the most recent (and last) generation of Panasonic plasmas have finally matched the performance of the 9th generation Pioneer Kuro plasmas that were discontinued in 2009. This is not quite right. The Panasonics have actually exceeded the Pioneer's performance. The reviewers are referring exclusively —almost obsessively—to contrast ratio, as though this is the only performance criterion one need consider. The Pioneer's color can measure nearly perfect when looking only on the gamut boundary, but this achievement falls apart when looking inside the gamut. The Panasonics, by way of contrast, can provide nearly reference color throughout the color space. The visible difference is not subtle.

The examples above show various ways to check the color performance of your display. Figure 1 is a standard CIE chart. This shows six colors: red, green blue, yellow, cyan, and magenta. The 8-bit color space used by commercial video includes over ten million colors! Simply looking at six of them is not enough to judge the true performance of the display. An easy way to do this is shown immediately below. Figure 2 shows a saturation sweep that samples multiple versions of the primary and secondary colors at different levels of saturation. This example shows twenty-four colors, or four versions of each color point. Figure Three shows eighteen colors provided by the ColorChecker, a tool that has been used for years in digital photography. These colors were chosen as example of realistic hues in the natural world. Although these 42 colors still represent a tiny fraction of the entire color space, it is a good bet that if these colors measure well, virtually any color in the gamut will be reasonably accurate as well.

Figure 1: The Gamut Boundary

 

Figure 2: A Look Inside the Gamut—Saturations
Figure 3: A Look Inside the Gamut—ColorChecker

Case Study: Pioneer PDP-5020FD Kuro Plasma

This 9th-generation Kuro plasma was universally praised at the time of its release. The only negative comments related to the lack of custom white, gamma, or CMS controls. To get these enhancements you need to purchase the more expensive Elite models.

What no one reported at the time, because no one writing for the trade magazines or posting on enthusiast forums cared to measure it, was that even with a full-featured CMS the Pioneer's color performance was poor.

I decided to test this by using a particularly powerful, though time-consuming, technique for testing the 5020's color. Using Excel's random number generator feature, I created three columns of 1000 random numbers within the video range of 16-235. Each row then represented a single RGB triplet inside the display's natural gamut (8-bit video has over 10,000,000 colors). Then I used a custom automated process we have developed to automatically generate test patterns according to those values, take measurements, and then record each xyY value. Using a very fast Klein K-10a meter, I was able to measure these 1,000 colors spread randomly throughout the gamut in less than 15 minutes. I then exported this data to a csv file, and imported into a data analysis spreadsheet developed for this purpose.

I tested the Pioneer plasma in five modes:

  • Uncalibrated (using the best presets)
  • Calibrated (using the Lumagen's grayscale, gamma, and standard CMS controls)
  • Calibrated at 75% of the HD gamut (using the Lumagen's grayscale, gamma, and standard CMS controls)
  • Calibrated using a Lumagen 125-point LUT
  • Calibrated using a Lumagen 729-point LUT

Here are the results.

The truly striking result from this test is that the uncalibrated state actually results in better overall color (lower average dE) than what's provided with a standard pri/sec color CMS calibration. Also, you might be tempted to blame the Lumagen for this, but it's not the problem. The problem arises from attempting to color calibrate a non-linear display at the gamut boundary only. Even when the colors at the gamut boundary are correct, inside the color space tells a very different story.

Calibrating at 75% of Rec. 709 (25% inside the gamut boundary) yields significantly better results. However, to get the full performance that this display is capable of you must use an LUT. As you can see, the 125 point LUT offers a massive improvement in accuracy and the 729 point builds on that producing what can be reasonably described as a perfect result. Since the average dE with the 729-point calibration is below 0.5, a larger number of points would offer no additional benefit.

It is also important to understand that the results otained by using the LUT do not merely produce good looing test charts. The real-world performance of the display is visibly enhanced, and not in a subtle way. The Pioneer was transformed by the 729 LUT. The combination of reference black levels AND reference color performance produced a stunning image. In fact, the LUT-enhanced image improved in ways I wouldn't have expected. In addition to the natural color, detail and intra-image contrast seemed improved as well. This was likely due to the improvement in gamma throughout the entire gamut (again, not just for white and color on the gamut boundary), but I can't be sure.

For displays like the Pioneer that lack linear color performance, I would consider the use of an LUT essential for video enthusiasts.

Is a LUT Always Necessary?

As useful as an LUT can be in taming colors errors inside the gamut, it is not always necessary. Some displays are engineered well enough that once you correct the colors on the gamut boundary the remainder of the color space measures excellent as well. This is simply a matter of good engineering. The late model Panasonic plasmas are a good example of this. Recent Samsung plasmas can be excellent as well, as is the great Sony 4K SXRD projector. However, you cannot count on this. The only way to know for sure is to measure inside the color space after using a standard CMS, or maybe only after selecting the display's best preset. For example, the Sony 4K SXRD has no CMS and doesn't need one. In the Reference picture mode, the color is nearly perfect all throughout the gamut.

Therefore, you are always better off getting a display that performs well to begin with, rather than having to correct it later. This is especially true with a LUT, which requires external hardware (an added expense) and depending on the number of colors sampled can take quite a while to run.

The Lumagen Radiance line of processors offer LUT color calibration of 125, 729, and 4913 colors, and ChromaPure supports each using its advanced auto-calibration process.