RGB, YUV & Media Ambit

RGB is the convenient and simple designation of additive color model used for more than 100 years in colorimetry,  photography, cinematography, color TV and other areas. Does it means something particular?

Not at all!

There are so many variants of hardly compatible systems, using RGB as convenient label for light levels, screen pixels, analog or digital signals, or even the group of sub-channels within multi-channel video data processor.

Adding subscript and superscript indices, extra explanatory letters and  exponential values are not necessarily helpful.

The academic pedants insist on these cumbersome additions and they are keen to append more and more symbols  all the way from the camera lens to the display screen, thus making video engineers’ life really difficult.

The terminology committees have one very good rule:
If some feature of the term is well known and assumed by default, then such feature should be excluded from the term itself. The scientific name for this procedure is ellipsis  (from the Greek: ἔλλειψις, élleipsis, “omission”).

E.g. if the whole exhibition is dedicated to electric cars, then the presenter may say “this wonderful car”, omitting “electric”.

Thus, we have full right to designate the Red channel signal simply R  if we know for sure, that we are talking about some particular flavor of this wonderful signal.

For example: Full Range 10 bit , decompressed, full bandwidth, gamma-corrected, color-corrected HDR-PQ RGB data, originated in assumption of BT.2020 Primaries and D65 White Point, taken at the output of the color enhancement stage of the production workflow using Master Display in the  DCI-P3 Primaries mode.

The YUV signals group consists of the luminance signal Y (weighted sum of  R, G and B signals) and two color difference signals  B-Y, and R-Y. Color difference signals are scaled to fit the available  data levels range, e.g. 8, 10, 12  or 16 bit. The scaled versions of B-Y and R-Y signals are traditionally designated as Cb & Cr or U & V.

In comparison with the RGB signals typical processing and compression distortions of YUV signals are less noticeable.

However, it should be noted that in some conditions, e.g. if bit depth is not high enough, the  RGBYUVRGB conversion may go wrong, i.e. the reconstructed RGB images may significantly differ from the original RGB images.

The scope of the widely used terms Y,B-Y,R-Y,  YUV, YCbCr,  YIQ, YDrDb, YPbPr, etc., is ambiguous and overlapping.

We recommend to use widespread abbreviation YUV,  thus ignoring the academic pedants claim that the only legal name for this group of digital signals is pretty awkward Y’CbCr.

However, if necessary, we may and should specify extra details, e.g. slightly longer name YUV420p10le designates 420 sampling structure of YUV data, planar memory array format and use of 10 bit little-endian code-word per component. 

By some obscure reasons, rather weird term “pixel format” (pix_fmt) is often used to designate this group of parameters.

However, even such “extended label” as YUV420p10le does not cover all vitally important parameters of YUV data.

To get good images at the workflow output we should include in the YUV data specification the specification of RGB data from which they are derived.

For example: HDR-PQ RGB data originated in assumption of BT.2020 Primaries and D65 White Point, taken at the output of the color enhancement stage of the production workflow using Master Display in the  DCI-P3 Primaries mode.

YUV data specification must also include description of RGB to YUV conversion parameters. 

For example: 12 bit Full Range RGB data converted to 10 bit Narrow Range YUV data using BT.2020 Color Encoding Matrix in its linear Non-Constant Luminance version.

Shall we accept this large group of interrelated parameters as ‘simple, easy and comprehensive description’ of YUV data stream?


On condition that all these important  “birth certificate and life history” details are not included in the YUV label itself, but:

  •  embedded as Metadata within the Video Data Stream
  • and/or included in the Media Title Multi-file Package
  •  and/or stored separately in the Media Ambit database.

See next:

Tables and numerical examples of reference values, such as coefficients of Color Encoding Matrices and Color Conversion Matrices, as well as Color Bars Reference Levels for a variety of Color Spaces.

Learn more:

See posts in the Video Science Fundamentals and Media Ambit Tools categories.

Leave a Reply

Your email address will not be published. Required fields are marked *