When Marketing is Lossy

 In Blog Post, Guest Post

I remember when I was a kid looking at a Hot Wheels TV commercial, boy I wanted that track! When I finally received one the next Christmas, my disappointment was noticeable. TV commercials are exaggerated – my Hot Wheels car could not do 3 back-flips and land on its four wheels. I expected too much because the product did not live up to the marketing.


Ok, it was long ago and I was a kid… Since then, I filter. I take a look at the marketing, do some independent research, and I make my own judgment. It works well and helps me avoid disappointment.

Now in 2019, pro AV companies have started to abandon matrix switches to go over IP instead. Hence, their marketing departments start bragging about their associated image compression technology.

WOW! In some cases, a new level of nonsense has been reached; distance from reality now measures in yotta-meters. My Hot Wheels ad seems real by comparison.

Lossless / (ˈlɒslɪs) /

adjective

Term used by marketing teams to fool users into thinking that their image compression technique will not produce artifacts. When paired with your CODEC name, any compression ratio becomes acceptable and customers will just buy millions of dollars of your equipment.

synonyms

Pixel Perfect

antonyms

No real antonyms exist since every CODEC technology in pro AV is marketed to be lossless.

Reality check

There is no such thing as a free lunch. In fact, it is possible to achieve lossless quality while still reducing of bit rate. Images and videos contain redundancies that can be exploited. But characteristics of the content itself determine the amount of compression that can be applied before loss occurs. The compression schemes for images and videos are basically the same, except that video signals can be compressed in space (x, y coordinates in an image) as well as in time (frame to frame).

There are two primary categories of image redundancies that CODECs exploit.

Statistical Redundancy

Statistical redundancy occurs due to the fact that a pixel within an image tends to have very similar intensity as its neighbors. This is true except at the object boundaries or illumination changes. For still images, statistical redundancies are spatial in nature. For natural two-dimensional images, redundancies are present along both the x- and y-dimensions. Video signals exhibit yet another form of statistical redundancy: temporal. For video, pixels in a certain position across successive frames tend to be very similar, unless there is large amount of motion present.

Psychovisual Redundancy

Psychovisual redundancy arises due to imperfections in human visual perception. Our eyes are more responsive to changes of illumination than to perceiving finer details and rapid changes of color. Hence, a strong CODEC balances preservation of the details of the source image against bit-saving image compromises that the human visual system is not well suited to detect. That is, exploiting the psychovisual redundancy.

Rule of Thumb

Working in CODEC technology for so many years, I have seen it all. For video-only content, a very high complexity H.265 CODEC (combining spatial and temporal redundancy optimizations) can yield very good fidelity at fairly reduced bit rates. But most H.265 encoding is non-real time, due to the complexity of the algorithm (think of a Netflix server farm spending six hours encoding a 2 hour movie).

Unfortunately, it gets more complex in pro AV when the content is not only video but can also be graphics, and the requirement is for real-time encoding with minimal latency. When both type of content are in play, I have never seen a CODEC that could compress 4:1 or more without noticeable artifacts. I have to admit that algorithms are getting more sophisticated, and sometimes I struggle in the lab to reproduce visible artifacts around the 4:1 ratio.

But rest assured, at 20:1 it is a piece of cake. Artifacts jump at you without even trying.

This past NAB I heard companies trying to fit an 8K – 50 Gbps signal down to a 1G pipe. Ouch! Why go 8K in the first place? Why spend so much on 8K source and displays just to compress the content over 1G? At a fraction of the price, a 4K system using 10G bandwidth would yield much better visual performance.

Oh wait – I see, the marketing department says it is lossless…