Color Spaces

Printer-friendly version

Moviola's Resource Center: 

  Tips, Tutorials, Articles & More »

Color Spaces

There are many different factors that affect the quality of recorded or transmitted video. One of these factors is the color space of the video. When comparing video cameras, tape formats and computer post production formats, the following terms and concepts are an important guide to aid in your decision-making about quality.

The Short Answer

There are two main color models that we work in with when images are converted into video:

• RGB
• Color Difference (generally but incorrectly referred to as YUV)

Within each of these models, there are several specific color spaces, in other words, the range of colors available. You can think of a color space as akin to a specific color resolution. This color resolution not only impacts the original filmed scene, but how much the video can be manipulated in digital post production. If you're going to shoot against a green screen for example, you'll want a video format that has good enough color resolution to handle the job of chroma keying out the green background successfully.

RGB: Generally reserved for high end applications in the digital realm, but rarely used for videotape. The RGB model employs the best color resolution, often referred to as full bandwidth color or the notation 4:4:4, implying that there is the same sampled resolution in each of the three channels (red, green and blue).

Color Difference: Color model used for practically all video, analog and digital. Within color difference systems, color information is often scaled down or reduced. When working digitally, color resolution is categorized by the following notations:

More Info

All image information starts out as direct proportions of red, green and blue intensities (RGB)- whether from a digital video camera or film through a film scanner. Since the beginning of television, this information is generally converted (encoded) into a different form before being recorded to a video format- a color difference system. On display, the video is decoded back into red, green and blue values. For TV, this is accomplished by the circuitry in the TV itself.

Color Difference video is also referred to as Luma/Chroma video, as the RGB values are converted into separate luma and chroma components for recording.

• Luma: describes the brightness portion of the video signal.
• Chroma: describes the color portion of the video signal. The chroma is separated into 2 color difference components, which we can generalize as "red data" and "blue data".

Since the 1950's (the introduction of color in NTSC), the luma/chroma model has taken on different shapes- many different individual color spaces, each handling the luma and chroma information a bit differently, with different quality results. Generally, when people refer to video, they are referring to one of these luma/chroma systems.

Video systems based on this luma/chroma model include:

• NTSC
• PAL
• Digital standard definition (SD) formats based on the ITU-R 601 standards (officially called recommendations)
• Various HD formats based on the ITU-R 709 standards (officially called recommendations)

Examples of a just a few of the various video formats (SD and HD) that record video using a luma/chroma system (color difference):

• Sony BETACAM SP
• DV
• Sony Digital BETACAM
• Sony HDCAM
• Panasonic D5 (and HD-D5)

One tape format that can record RGB video is:

Sony HDCAM SR

Note: HDCAM SR can record both models- RGB or Component (implying a color difference model of digital component video at 4:2:2 color sampling).

Bandwidth, Quality and Color Difference Models

One of the core elements of a color difference model is that during the encoding, the color information is usually scaled down or reduced. In digital terms, this is referred to as sub-sampling.

Benefits of color scaling:

• Reduces the bandwidth of the video signal
• Since the operation is similar to how the human eye perceives light and color, this color reduction is not noticeable under most viewing conditions.

Downsides to color scaling:

Although this scaling doesn't necessarily affect our perception of colors, it can hamper our efforts to manipulate video in post production- a prime example is chroma keying, which requires good color information.

Color Spaces (Color Difference Models)

A color space refers to the range of colors available in a particular color model. First, as we mentioned earlier, in these models, video is broken down into three signals:

 
When these signals are used in a component environment, in other words all three signals are kept distinct and separate, the signals would be expressed (and labeled) as: Y, R-Y, B-Y

 
(For analog decks and monitors, these symbols will often be the labels for component connections.)

There are several variations of luma/chroma (color difference) models, listed next.

YPbPr

Describes analog component video. Similar to Y, R-Y and B-Y. Commonly found on consumer equipment like DVD players connecting digital equipment to analog TV's.

YIQ

This is the color coding system used for NTSC video. This generates analog, composite video. For NTSC, after separation into luma, and two 2 color difference components, the color components are scaled down and combined into a single chroma signal, then the chroma is combined with the luma to create a single composite signal. Y stands for the luma component, and I and Q stand for the two scaled color components. I stands for intermodulation, and Q stands for quadrature.

YUV

Probably one of the misused terms in post production. Let's go over the incorrect use first, since this is generally how the term is used today.

Common Use of YUV

YUV is commonly used to describe color difference models in general, but usually is used to refer to component video, analog or digital. The proper term that should be used when referring to digital component video is Y'CbCr. In some systems, like Avid, it may be listed by one of the two universal digital standards: ITU-R 601 for SD video and ITU-R 709 for HD video. This also stands for Y'CbCr.

Correct Usage Of YUV

YUV, in its proper context, describes the coding system used for PAL analog video. After the video signal is separated into a luma and two 2 color difference components, the color components are scaled down and combined with the luma to create a single composite signal. Y stands for luma and UV for the two scaled color difference components.

Y'CbCr

A digital coding system where all three luma/chroma components are kept separate (although they are conveyed by means of a single cable). This is the basis for all digital, component video. This is standardized under ITU-R 601 for SD coding, and ITU-R 709 for HD coding. Y'CbCr is often (and usually) described inaccurately as YUV.

Color Sub-Sampling

In Y'CbCr, as with its analog counterparts, the color components are usually scaled down, referred to as sub-sampling in the digital world. The various sub-sampling schemes have a major impact on quality- and should be a consideration for post production decisions as color that is scaled down too much isn't as acceptable for various image-stressing operations like chroma-keying.

4:2:2: For every 4 samples of luma, there are 2 samples of each chroma component. This is full bandwidth in luma. Chroma components are scaled down to half the resolution. The most common sub-sampling technique. This is the standard recommendation in the international digital encoding standard ITU-R 601 for SD video. Many consider this the minimum quality for decent chroma-keying and similar effects. Many video formats use 4:2:2 sampling including Sony Digital BETACAM, Panasonic DVCPRO HD, AVC-Intra, D5, HDCAM SR (in component mode).

3:1:1: Similar in resolution to 4:2:2 sampling, used in Sony HDCAM (not the newer HDCAM SR).

4:1:1: For every 4 samples of luma, there is 1 sample of each chroma component. Full bandwidth in luma, quarter resolution in color components. Used in DV-NTSC. Not generally acceptable for chroma-keying.

4:2:0: Color is sampled differently than in the other sampling schemes. Sampling is done at 4:2:2 but only on every other line. 4:0:0 is sampled on the other lines (luma only). Color is derived by interpolation. This sampling is not acceptable for chroma-keying. Used in DV-PAL, all HDV formats, XDCAM HD and EX, SD DVD production, other MPEG-2 applications and recently in the consumer/prosumer format called AVCHD (to be detailed in another section).

RGB Color Model

RGB has actually been used in a number of specific color spaces. For the purposes of keeping this discussion simple, we'll look at an overview of RGB as its used in video and digital post production. RGB is completely different in nature than color difference systems, as the originally sampled red, green and blue values are not converted into a luma/chroma model. Full bandwidth color is maintained, resulting in very high quality. For this reason, it is usually labeled as RGB 4:4:4, implying that no sub-sampling is taking place.

RGB 4:4:4 has very high data rates and is currently reserved for very high end applications, including:

• Scanning film to digital files for digital post including effects, color correction and creation of digital intermediates(DI) for printing back to film.
• High end digital cameras: The following cameras are capable of generating RGB 4:4:4 signals: Panavision Genesis, Thomson Viper and others.
• One tape format that records in RGB 4:4:4 is Sony's HDCAM SR.
• RGB is also used as data files in some environments, not recording to tape at all.

This color space is ideal for post production effects like chroma-keying because of its very high color resolution. RGB 4:4:4 is referred to as dual-link HD, as two HD-SDI cables are required to transmit the full signal, one cable sending a 4:2:2 signal (full bandwidth of green, half red, half blue) and one cable sending a 0:2:2 signal (half bandwidth of red, half blue).