Giorgianni, Edward J., and Thomas

E. Madden (2008), Digital Color

Management: Encoding Solutions,

Second Edition (Chichester, U.K.:

Wiley).

Examine the flowers in a garden at noon on a bright, sunny day. Look at the same garden half an hour after sunset. Physically, the spectra of the flowers have not changed, except by scaling to lower luminance levels. However, the flowers are markedly less colourful after sunset: Colourfulness decreases as luminance decreases. This is the Hunt effect, named after the famous colour scientist R.W.G. Hunt. Images are usually viewed at

a small fraction, perhaps 1⁄100 or 1⁄1000, of the luminance at which they were captured. If the image is presented with luminance proportional to the scene luminance, the presented image would appear less colourful, and lower in contrast, than the original scene. To present contrast and colourfulness comparable to the original scene, we must alter the characteristics of

the image. An engineer or physicist might strive to achieve mathematical linearity in an imaging system; however, the required alterations cause the displayed relative luminance to depart from proportionality with scene luminance. The dilemma is this: We can achieve mathematical linearity, or we can achieve correct appearance, but we cannot simultaneously achieve both! Successful commercial imaging systems sacrifice mathematics to achieve the correct perceptual result.

Image state

In many professional imaging applications, imagery is reviewed and/or approved prior to distribution. Even if the image data originated with a colorimetric link from the scene, any technical or creative decision that results in alteration of the image data will break that link. Consider the movie Pleasantville. Colour is used as

a storytelling device. The story hinges upon characters depicted in greyscale and characters depicted in colour. (See Figure 2.2.) The R’G’B’ values of the final movie do not accurately represent what was in front of the camera! This example is from the entertainment industry, however, examples abound whereever colour is adjusted for æsthetic purposes.

Picture rendering is ordinarily a nonlinear operation, not easily described in a simple equation or even a set of equations. Once picture rendering is imposed, its parameters aren’t usually preserved. In many applications of imaging, image data is manipulated to achieve

ISO 22028-1 (2004), Photography and graphic technology– Extended colour encodings for digital image storage,manipulationandinterchange.

High-end D-SLR cameras have provisions to capture “raw” data that has not been subject to picture rendering operations. These cameras are capable of capturing “science.”

an artistic effect – for example, colours in a wedding photograph may be selectively altered by the photographer. In such cases, data concerning picture rendering is potentially as complex as the whole original image!

The design of an imaging system determines where picture rendering is imposed:

•In consumer digital photography and in video production, picture rendering is typically imposed in the camera.

•In movie making, picture rendering is typically imposed in the processing chain.

If an imaging system has a direct, deterministic link from luminance in the scene to image code values, in colour management terminology the image data is said to have an image state that is scene referred. If there is a direct, deterministic linkage from image code values to the luminance intended to be produced by a display, then image data is said to be display referred.

Video standards such as BT.709 and SMPTE ST 274 (both to be detailed later) are at best unclear and at worst wrong concerning image state. Consequently, video engineers often mistakenly believe that video data is linked colorimetrically to the scene. Users of digital still cameras may believe that their cameras capture “science”; however, when capturing TIFF or JPEG images, camera algorithms perform rendering, so the colorimetric link to the scene is broken. What is important in these applications is not the OECF that once mapped light from the scene to image data values, but rather the EOCF that is expected to map image data values to light presented to the viewer.

If a movie is “in production,” then principal photography is not yet complete.

The word reproduction, taken literally, suggests production again! I propose presentation.

presentation of programs. I’ll use the generic word “program” as shorthand for a movie, a television show, or a short piece such as a commercial. Figure 2.4 above presents a sketch of the entire chain.

Production refers to acquisition, recording, and processing. In a live action movie, the term production may be limited to just the acquisition of imagery (on set or on location); processes that follow are then postproduction (“post”). In the case of a movie whose visual elements are all represented digitally, post production is referred to as the digital intermediate process, or DI.

Production culminates with display and approval of a program on a studio reference display – or, in the case of digital cinema, approval on a cinema reference projector in a review theatre. (If distribution involves compression, then approval properly includes review of compression at the studio and decompression by a reference decompressor.) Following approval, the program is mastered, packaged, and distributed.

Professional content creators rarely seek to present, at the viewer’s premises, an accurate representation of the scene in front of the camera. Apart from makers of documentaries, movie makers often make creative choices that alter that reality. They hope that when the program completes its journey through the distribution chain, the ultimate consumer will be presented with

a faithful approximation not of the original scene, but rather of what the director saw on his or her studio display when he or she approved the final product of postproduction. In colour management terms, movie and video image data is display-referred.