Color Matching Film to Video
The visual color fidelity that is attainable through the translation of video data to film has been a source of concern for video based film makers since this process emerged. It has resulted in service suppliers inventing catch phrases like “Color Maintenance Program” without clearly explaining what such a program does. These phrases are always applied with the caveat that the visual results on film will be “as accurate as technically possible given the differences in the color space of video vs. film.” I would like to examine the differences in the color space of video vs. film in a more detailed manner and present the latest technological breakthrough that can give you true visual fidelity between video and film.
Let’s step back and talk about the universe of color before we can get to our little galaxy called video to film. I define Color Space as the entire universe of color – every color that exists. This provides the context to discuss the two areas that we are really interested in namely video and film. Now, let’s define all of the colors that a device or process can generate as its gamut. Generally different devices/processes have different gamuts. This means that they may display the same data differently and colors that are possible on one device may not possible on another. Colors that can not be reproduced by a device are said to be “out of gamut”.
We are interested in the color that an average human perceives when viewing two different devices. When the same data is presented on those two devices do the colors look the same or different? It is also important to consider how the comparison is made – is one of the representations recalled from memory while the other is directly viewed or are they viewing both media simultaneously in a controlled viewing environment?
In comparing video data with it’s translation to film we are comparing two very different mediums. With video we are viewing yuv data on the phosphor of a broadcast monitor. With film we are viewing an image created by passing light from a xenon lamp through an answer print that was printed from a digitally created negative (subject to film recorder calibration). Each media has specific requirements that should be met to ensure proper viewing.
It has been well documented that video and film have different color gamuts. The following three questions can be used to define all the possible outcomes in the video to film translation process.
1. Are
there colors that translate from video to film providing an accurate visual
match?
2. Are there colors in video that are also possible on film but are not accurately
translated?
3. Are there colors that are possible on video that are not possible on film?
The answer to all three questions is yes. Before we can improve the current state of the video to film transfer we need to understand where the discrepancies in the transfer process come from. Broadcast video and digital film each have a set of standards that allow for a consistent image to be viewed in different venues. It is not practical to expect the broadcast world to change standards in order to accommodate greater color fidelity in video to film transfers. Therefore, I will assume that the color that is displayed on a calibrated broadcast monitor is the standard that we are trying to visually reproduce on a one light answer print. The only option that we have is to look at the digital film recording technology that we use (both hardware and software) to find improvements.
First, it is helpful to understand how film recorders “control” color in order to understand some of the limitations in the conventional video to film translation. The first and only step in conventional film recorder color control is calibration. Calibration is based on recording a series of grey patches (typically 21 steps) that vary from black to white onto negative motion picture film stock. The patch densities are read from the negative with a densitometer and the data is used to modify the intensity of the film recorder’s light source to achieve the target grey scale density. Assuming that the aim values of the grey scale were selected so that a print visually matched the digital data viewed on a broadcast monitor we have a calibrated video to film system. We have also found a lot of colors (all grey) that visually match between video and film.
A film recorder controls color by applying these same grey scale intensity settings to each color channel of an image. Let’s look at the simple case of two 8-bit color patches that have RGB color values 128, 128, 128 and 128, 0, 0. The same red light source intensity would be applied to both of these images to produce a negative that printed to a medium grey in one instance and a deep red in the other. In this example the grey patch should visually match the video monitor because it lies on the grey scale that was calibrated. What about the deep red patch? What about other colors that lie far away from the grey scale? If video responded to color mixing the same way a film recorder controlled color everything would match visually. This is not the case. The further color varies from our calibrated grey scale the more visual infidelity between our two media develops. In most cases the colors that exist on video also exist in the gamut of film however, the film recorder’s color control system can not compensate for variations off of its calibrated grey scale to reproduce them accurately.
The good news
It
is now possible to correct these off grey scale color variations through color
gamut remapping. This has been achieved through measuring and mapping the
color gamut of both a standard broadcast video monitor and the color gamut
of motion picture print stock printed from a digitally recorded negative.
The two gamuts were converted to a common color system and a sophisticated
translation applied to the data so that all the video colors that are possible
on film are aligned and produced during film recording. This type of translation
is often referred to as a 3D data transform – it allows us to treat
color channels differently for all possible combinations of red, green and
blue. An 8-bit per channel image can have over 16 million different combinations
of red, green and blue each of which is remapped during film recording using
this technology!
What about those
out of gamut colors?
Fortunately,
the colors that can be produced on video that are not possible on film are
in the overwhelming minority. Out of film gamut colors can be captured on
video or created through color correction. It is not possible to reproduce
these colors accurately on film.
The results
Side by
side comparisons of video displayed on broadcast monitor with projected answer
prints are spectacular.
Alan
Bak (P. Eng., MBA)
President
In 1998, ARRI, a German company with a long and well respected history in the motion picture camera and equipment industry, turned its efforts to producing the world's most sophisticated film recorder. The result is the ARRILASER, recognized in 2001 by the Academy of Motion Picture Arts and Science with an Academy Award for Scientific and Technical achievement. The Academy stated: "The ARRI Laser film recorder demonstrates a high level of engineering resulting in a compact, user-friendly, low-maintenance device while at the same time maintaining outstanding speed, exposure ratings and image quality."
Shooting on intermediate stock also allows the use of polyester based films which are incredibly tough, scratch resistant, and difficult to tear. As a result, release prints can be made directly from the original polyester based digital negative, providing first-generation quality. Rather than creating an IP and dupe negative for release printing, many of our clients just make an IP for protection. This delivers significant cost savings when producing a limited release movie.
