Wednesday, May 7, 2014

Distribution vs. Contribution Encoding: Part 1

As television technology continues to improve, broadcasters want to ensure that viewers are getting a crisp, clear image. This is where special contribution encoders come into play, as distinct from the normal sort of distribution encoder, used to render images for end-user viewing.

The main difference between the two is how color, the essential attribute of a digital picture element or “pixel” is treated. There are two separate issues here. One issue is the number of bits used to denote the color of a pixel, generally referred to as “8-bit” color, “10-bit” color, and so on.

The other issue is the number of pixels that are retained in an encoded picture. The terminology used here is a bit more obscure. 4:4:4 encoding means that 4 out of every 4 (i.e., all) pixels are retained in the encoded image.  As a rule, encoders do not support this mode, since the essential function of an encoder is to reduce the number of pixels needed to reproduce an image. 4:2:2 means that 2 out of every 4 pixels are retained in the encoded image. As a rule, this mode is only supported by source encoders. Distribution encoders instead support 4:2:0 mode, in which only 1 pixel is retained out of every 4.

Perhaps surprisingly, retaining just a quarter of the pixels in an image is sufficient to allow the image to very accurately reconstructed during the decoding process; indeed, few people can detect any difference between a 4:4:4 image, a 4:2:2 image, and a 4:2:0 image. As a consequence, for purposes of viewing, the extra pixels retained in 4:4:4 and 4:2:0 can be regarded as simple overhead, better dispensed with in a distribution encoder, since the fewer the bits that need to be transmitted or stored, the lower the cost of digital video.


Source encoders, however, are not designed merely for viewing pictures, and so the rules for them are different than the rules for distribution encoders. In our next blog, we will consider this issue in more depth, along with the issue of “8-bit” vs. “10-bit” color.