Color subsampling

from Wikipedia, the free encyclopedia

With chroma subsampling (English chroma subsampling [ kɹoʊmə sʌbˌsæmplɪŋ ], color subsampling ), in the image pickup means a method of reduction of the required amount of data is used, for example, to reduce the required memory or transmission bandwidth required (u. A. In telecommunications systems ) to restrict . The prerequisite is the use of a suitable color model that can describe chrominance (color information) and luminance (brightness information ) separately. The chrominance is stored with a lower sampling rate than the luminance . The subjective optical loss of quality is small, since the human eye perceives color with a lower resolution than brightness.

Color subsampling can be found in video technology , in digital image formats such as JPEG and in the field of color television technology .

General

The color subsampling makes use of the physiology of the human eye, which, in contrast to the brightness information, can only perceive the color information with reduced resolution. In principle, all color models with color difference information separated from the brightness information, for example YCbCr or CIELab, offer the possibility of color subsampling .

Color image above and the splitting of this image into the components Luma Y '(brightness) and the two chrominance components Cb and Cr (colors).

The adjacent figure illustrates this effect: The image sharpness impression is essentially contained in the brightness information (2nd image). The two color components in the two images below are reduced in resolution. The information from these three images results in the above color image, the reduced color resolution of which is not noticeable.

In analog video technology, the color information is transmitted with a narrow bandwidth, which is mapped accordingly in digital video technology and image processing by reducing the sampling rate, i.e. increasing the spatial distance between the sampling points of the color components. Although color subsampling relates primarily to digital systems and has nothing to do with color television technology, some of the names come from analog color television technology, in particular from the NTSC color television system .

Designation scheme

Common color subsampling

There are various forms of color subsampling, which describe the subsampling of the color signals as a relation to the brightness signal. The following notation with three numbers is usually used:

This notation is based on a pixel block of A pixels wide and two pixels high and describes the sampling rates for the color information contained therein in relation to the value A.

  • The first digit A is now usually given as 4, since traditionally all color subsampling rates were low powers of two (even if 4: 2: 2 could also be given as 2: 1: 1, this is not common in practice).
    Originally, A stood for the factor of the sampling rate of the brightness signal Y ' in multiples of the NTSC video bandwidth of 3.375 MHz in the Y'CbCr color model. The sampling rate is typically four times the video bandwidth, which results in a pixel clock of 13.5 MHz in the ITU-R BT 601 digital video standard . The number 3 is also rarely used as a leading number, since otherwise fractions in the values ​​B and C would have to be used at sampling rates of one third.
  • The second digit B stands for the sampling rate of the two color channels C b and C r in the upper row of pixels in relation to A or originally to 3.375 MHz. In the context of ITU-R BT 601, it has the value two and represents the color undersampling.
  • The third digit C describes the same value as the second digit, but for the lower row of pixels in the pixel block.
  • Occasionally, a fourth number is given afterwards in the designation scheme, which indicates the sampling rate ratio of an alpha channel that is also transmitted .

The following color subsampling are common:

A: B: C notation 4: 4: 4 sampling
Y'CbCr 4: 4: 4 		A: B: C notation 4: 2: 2 sampling
Y'CbCr 4: 2: 2
A: B: C notation 4: 2: 0 sampling
Y'CbCr 4: 2: 0 		A: B: C notation 4: 2: 0 sampling MPEG-2
Y'CbCr 4: 2: 0 MPEG-2 sample positions

White circles symbolize the scanning positions for the luminance, light and dark gray circles the positions for the two chrominance signals.

4: 4: 4 (1x1.1x1.1x1)

In this case there is no color subsampling. Applications of 4: 4: 4 are in the area of ​​high-quality digital image processing and the RGB color space . In the RGB color space, due to the lack of separation between brightness and color information, color subsampling is fundamentally not possible.

4: 2: 2 (2x1.1x1.1x1)

This color subsampling originated from the analog color television standard NTSC and is used within the framework of the ITU-R BT 601 standard for digital video signals. A distinction is made between horizontal and vertical color subsampling: The scanning in the horizontal direction is only half as large as in the vertical direction. The Digital Betacam tape format uses the 4: 2: 2 method.

4: 2: 0 (2x2.1x1.1x1)

The 4: 2: 0 format is used for digital images in the JPEG standard or digital video material in the MPEG standard and has identical scanning in both spatial directions. As shown in the figures above, the scanning points of the color signal are slightly offset: With JPEG, scanning is carried out centrally, with MPEG in the same vertical orientation as the brightness signal.

Even if the reference to the NTSC video bandwidth does not make any sense for JPEG and MPEG, the designation scheme for color subsampling with digits was retained. The PAL-DV tape format (DV25-PAL) uses the 4: 2: 0 method.

4: 1: 1 (4x1.1x1.1x1)

Similar to 4: 2: 2, although the color channels in the horizontal direction only with 1 / 4 of the sampling rate of the luminance signal are sampled. The NTSC-DV tape format (DV25-NTSC) uses the 4: 1: 1 method.

The total color resolution corresponds to the 4: 2: 0 color scanning, but with a higher vertical resolution to the disadvantage of the horizontal resolution.

More color subsampling

In addition, there are some rarely used color subsampling. In the area of DVCAM (DV SDL), color subsampling with the designation 3: 1: 1 is used, which was coined by the Sony company . The pixel clock is not 3 times the NTSC bandwidth, but 12 MHz.

In the area of High Definition Television (HDTV) on the HD-SDI interface according to the SMPTE 292M standard , 22:11:11 occurs. However, the term is rarely used and is derived from the pixel clock of 74.25 MHz used in HDTV.

Individual evidence

  1. ↑ Chroma subsampling notation (PDF; 167 kB), Charles Poynton, 2008, engl.
  2. ^ Douglas A. Kerr: Chrominance Subsampling in Digital Images. (PDF; 114 kB) Accessed April 21, 2013 .

Literature sources

  • Charles Poynton: Digital Video and HDTV - Algorithms And Interfaces . Morgan Kaufmann Publishers, 2003, ISBN 1-55860-792-7 .