skip to main content
OSTI.GOV title logo U.S. Department of Energy
Office of Scientific and Technical Information

Title: Color spaces in digital video

Conference ·
OSTI ID:358797

Whether it`s photography, computer graphics, publishing, or video; each medium has a defined color space, or gamut, which defines the extent that a given set of RGB colors can be mixed. When converting from one medium to another, an image must go through some form of conversion which maps colors into the destination color space. The conversion process isn`t always straight forward, easy, or reversible. In video, two common analog composite color spaces are Y`tjv (used in PAL) and Y`IQ (used in NTSC). These two color spaces have been around since the beginning of color television, and are primarily used in video transmission. Another analog scheme used in broadcast studios is Y`, R`-Y`, B`-Y` (used in Betacam and Mll) which is a component format. Y`, R`-Y`,B`-Y` maintains the color information of RGB but in less space. From this, the digital component video specification, ITU-Rec. 601-4 (formerly CCIR Rec. 601) was based. The color space for Rec. 601 is symbolized as Y`CbCr. Digital video formats such as DV, Dl, Digital-S, etc., use Rec. 601 to define their color gamut. Digital composite video (for D2 tape) is digitized analog Y`UV and is seeing decreased use. Because so much information is contained in video, segments of any significant length usually require some form of data compression. All of the above mentioned analog video formats are a means of reducing the bandwidth of RGB video. Video bulk storage devices, such as digital disk recorders, usually store frames in Y`CbCr format, even if no other compression method is used. Computer graphics and computer animations originate in RGB format because RGB must be used to calculate lighting and shadows. But storage of long animations in RGB format is usually cost prohibitive and a 30 frame-per-second data rate of uncompressed RGB is beyond most computers. By taking advantage of certain aspects of the human visual system, true color 24-bit RGB video images can be compressed with minimal loss of visual information. For example, humans `see` more white-to-black (luminance) detail then red, green, or blue color detail. Also, the eye is most sensitive to green colors. Taking advantage of this, both composite and component video allocates more bandwidth for the luma (Y`) signal than the chroma signals. Y`611 is composed of 59% green`, 30% red`, and 11% blue` (prime symbol denotes gamma corrected colors). This luma signal also maintains compatibility with black and white television receivers. Component digital video converts R`G`B` signals (either from a camera or a computer) to a monochromatic brightness signal Y` (referred here as luma to distinguish it from the CIE luminance linear- light quantity), and two color difference signals Cb and Cr. These last two are the blue and red signals with the luma component subtracted out. As you know, computer graphic images are composed of red, green, and blue elements defined in a linear color space. Color monitors do not display RGB linearly. A linear RGB color space image must be gamma corrected to be displayed properly on a CRT. Gamma correction, which is approximately a 0.45 power function, must also be employed before converting an RGB image to video color space. Gamma correction is defined for video in the international standard: ITU-Rec. BT.709-4. The gamma correction transform is the same for red, green, and blue. The color coding standard for component digital video and high definition video symbolizes gamma corrected luma by Y`, the blue difference signal by Cb (Cb = B` -Y`), and the red color difference signal by Cr (Cr = R` - Y`). Component analog HDTV uses Y`PbPr. To reduce conversion errors, clip in R`G`B`, not in Y`CbCr space. View video on a video monitor, computer monitor phosphors are wrong. Use a large word size (double precision) to avoid warp around, the0232n round the results to values between 0 and 255. And finally, recall that multiplying two 8- bit numbers results in a 16-bit number, so values need to be clipped to 8-bits.

Research Organization:
Lawrence Livermore National Lab. (LLNL), Livermore, CA (United States)
Sponsoring Organization:
USDOE, Washington, DC (United States)
DOE Contract Number:
W-7405-ENG-48
OSTI ID:
358797
Report Number(s):
UCRL-JC-127331; CONF-9708191-; ON: DE98050923
Resource Relation:
Conference: Special Interest Group on Computer Graphics (SIGGRAPH) `97 conference, Los Angeles, CA (United States), 3-8 Aug 1997; Other Information: PBD: 1 May 1997
Country of Publication:
United States
Language:
English