|
|
 |
 |
 |
 |
 |
The science behind color calibration on Cameras
Before getting into the fine details, it is always good scientific practice to set up a baseline. For our purposes the baseline revolves around understanding why the need for accurate test patterns exists. There are three elements that need to be "baselined", and they all involve the flow of how the image is originally perceived to how the final product is viewed on screen.
The Human Eye
Without getting into too much biological detail, recreating what we see from behind the lenc can only be achieved if we understand why we see what we see. The eye percieves images through photosensitivity - interpreted through rods and cones in the retina. The fovea is a spot within the retina that is densely packed with cones - all of which have a specitif red, green or blue protein separation, and thus help produce a visible color spectrum.
If all three cones are activated at the same time, they tell our brain we see white light. If some are firing at full intensity and some ore not, we see colors in between white and all other colors. Rods on the other hand, are very sensitive to light, but are unable to see color information. Cones have color sensitivity, but requite more light to function.
The way we measure how sensitive cones are to light energy is through wavelengths. Nanometers are distance markers that help us quantify what wavelengths of light make up our visible spectrum. The spectral band of colors gues from Ultra Violet (below 400nm) up to Infrared (above 700nm). Individuals percieve color information differently, but for the most part, the human visible spectrum ranges from 400nm (blue) to as high as about 675nm (red). You also can read more about hot the human eye percieves color and light information in this article about the color perception of the human eye.
The Camera
While the peak sensitivity of human vision is about Blue 440nm, Green 540nm and Red 580nm visually "true" Red for most people has a special peak of about 635nm. A camera and television set must reproduce the color at the same wavelength if we are to accurately recreate that color. This is a difficult task, given the characteristics of monitors and TV sets, also the fact that they are seldom set up properly. CCDs transmit information by converting light to voltage levels that vary with brightness. Ensuring that a digital image when recreated will be of a particular sprectral curvature bevomes a complicated science. In most cases cameras try to mimic he cones found in the fovea with three CCDs, one each for each of the Red, Green and Blue primary colors.
In taking a picture of a banana the yellow color will activate primarily the red and green sensors - the the time we see that same banana on televition, its color will incorporate the characteristics (i.e., color space) of the television set, which, in theory, should be similar to the "look" of the original banana.
Data from CCDs is converted to digital voltage levels whose characteristics can be adjusted using a camera's color matrix. CCDs interpret images based on specific color thresholds engineered into their algorithmic circuity. Optimizing these thresholds largely depends on how deep the camera's menu will go. Higher end cameras allow an enourmous amount of matrix setting control, but without accurate monitoring materials, it is neigh impossible to line up these colors with any from of consistency.
The waveform monitor
The waveform monitor is one of the two important tools that make up an effective camera shooter's toolkit. A waveform monitor displays a video signal in millivolts (mV) for HD and IRE (Institure of Radio Engineers) for NTSC. 700mV is equal to 100 IRE units.
Using a precision greyscale chart?  aligning a camera is usually quite simple. First, light the geyscale chart evenly, and then adjust the camera settings to produce equal spacing of the geyscale's step signals between 0mV and 700mV - this should result in an accurate greyscale alignment.
IMAGE of waveform monitor
When selecting a test pattern, dynamic range becomes very important. In the early days of television, the dynamic range of a camery was very limited, to about 20 or 25:1. Greyscale pattern had a similar low dynamic range. Unfortunately, such test charts are still available and being sold to align modern cameras with a dynamic range of 3000:1 or even higher. The progression rate between most 9 step charts and a modern 11 step test pattern is also different. The lightest step on a 9 step chart is 60% reflection where a similar 11 step chart has a reflection of 90% (on the first step). Obviously, different densitometric curves between test patterns will result in different image reproduction rom cameras. Aligning to an inaccurate or greyscale of limited dynamic range can result in poor quality images that are beyond redemption.
The need for an accurate Greyscale chart is essential to setting a precise gamma and exposure. The white chip on a chart is normally set to 700mV (100 IRE) and true black close to 0mV. Setting the black can be tricky. NTSC has a basic black level set up of 7,5 IRE above true black. It would seem logical to set the blacks at 0; however, absolute black is an impractival goal because all surfaces reflect some light, bowever little. Even the blacks found on a high quality testchart do reflect some little light. Only a true black cavity? on a black level might come very close to a true black.
The Vectorscope
The vectorscope is the other "must have" tool that should be part of every shooter's toolkit. Only after setting a camera's exposure levels and tracking can a shooter mote on to optimizing colors. Be sure you use a true color chart standardized by the ITU (International Telecommunications Union) for HD color IRU-R BT.709, (SMPTE 274M). The ITU 709 standard largely replaces a number of precious standards. While the NTSC colorimetry standard had an excellent wide color gamut (when TV sets were made to that tandard) they didn't sell, because the picture was too dim and had to be viewed in a darkened room. For this reason SMPTE C and Europe's EBU were born.
Make sure the test chart? has at least all 6 primary colors and for HD you should consider using an even finer adjustable color calibration chart with up to 28 colors.
IMAGE of vectorscope showing 6 aligned primary colors
Why not just use the primaries? With the advanced multi-matrix settings of modern cameras, it is possible to line up each primary color in their box, but reduce its working colorspace. The image below has been aligned to a 6 color pattern using a multi-atrix menu - as you can see; each primary is in its box. The following image is what a 28 color chart looks like with those settings - clearly the intermediary colors are completely skewed although the primaries are lined up.
IMAGE of vectorscope showing 28 colors aligned with skewed subprimaries
Why are those boxes so important?
When aligning a camera's matrix settings increasing or decreasing the sensitivity of a particular color will also change other colors. This often requires considerable patience in camera alignment. Always remember that the red and yellow are the most important as they largely define the skin tone. Here is an example of a camera operator error - if you line up red on the vector box to someone on set with a red sweater you will be altering the colorspace and gamut to reflect that red is your "true" red. This is examplified when going into the camera's matrix settings to adjust color sensitivity. Notice that in many instances a particular color cannot be selected on its own. Instead, alterations can be made to B-Y, R-B, etc. When these settings are increased or decreased, all of the other colors on the vectorscope shift as well - adjusting one color incorrectly will affect the entire colorspace and this should always be taken into consideration. Lining up to an incorrect color is worse then lining up to no colors at all - much like the common error of white balancing to a sheet of white printer paper.
Most camera manufacturers preset their matrixes on the warm side and you will notice that many cameras worth 10kEuro and above will have only the red and yellow are the only colors that fall into the vectorscopes boxes when you take the camera straight "out of the box".
Camera matching becomes impossible without an accurate test pattern. No two cameras are identical making it unreliable to upload the settings from one camera to another. The only way to match cameras effectively is to use the same test chart pattern and lighting condition. Set exposure levels, white balance, align the colors to the vectorscope and then match the second camera to a freeze frame of that pattern.
More often then not, the perfectly aligned vectorscope image is not the ideal setting based on the characteristics of the scene. However, by first aligning to a test chart pattern, it becomes an accurate baseline with the widest colorspace as a reference.
Color Correction notes
When setting up a camera, common practice is to do so a 2000 lux, with both light sources at 45 degrees. This however does not necessary mimic teal life shooting scenarios. You can use a color correction chart or test pattern as a color correction tool. Simply record just a few seconds of such a test chart or pattern at the head and the tail of the scene, the lighting characteristics of that scene are reproduced to tape, and editors can use this data to color correct or match scenes in post production. Secondary colors are very useful in post production. They are used to see changes to other colors when one color is changed dramatically.
|
|
 |
|
 |
|
|
|
|
