• Welcome to the Internet Infidels Discussion Board.

Smart TV artefacts

excreationist

Married mouth-breather
Joined
Aug 28, 2000
Messages
3,301
Location
Australia
Basic Beliefs
Probably in a simulation
I got a Hisense 4K TV as a replacement (old one had a problematic vertical line). It has occasional artefacts - so far I've only noticed it in Netflix. It seems to only happen where there is a large area of similar colour - usually on people's skin but one time with a lamp shade. Often it is a small spot that is pure green but sometimes the spot is black and one time it was light coloured or white. The spot moves around a bit and only lasts for about one second or slightly longer.

I was thinking that maybe it is a problem with the key frame - maybe key frames last about 1 or 2 seconds on Netflix? Maybe the TV has cores in its graphics card and one is faulty? Or maybe the AI upscaling has a problem?
 
It's hard to say what's going on here. My first thought is that these are data-compression artifacts. Data compression is, very roughly, looking for redundancies in one's data and then encoding the presence of those redundancies. There are a variety of algorithms for data compression, both lossless and lossy. Lossless means that the original can be completely restored, lossy means that it can't. Lossy compression can easily outperform lossless compression, but it's at the expense of losing some of the original data.

Text files one always compresses losslessly, for obvious reasons. The well-known archiver "Zip" does lossless compression.

But media files, audio, image, and video files, one often compresses lossily. One tries to lose details that are difficult to perceive, but if too much is lost, one will see compression artifacts.


For image files, PNG is a lossless format. For real-world pictures, it does not perform very well, because the number of bits one needs varies as the logarithm of the amplitude of variation. Thus, even a low-noise picture won't be compressed by very much. But diagrams can often be compressed much more, because they may contain large expanses of single colors.

GIF is an earlier format. It is lossless, but it is limited to 256 pixel colors. Getting a full-color image into GIF form requires loss of color resolution. If one does so naively, it makes gradients look banded. The solution there is "dithering", alternating between colors to get appropriate average colors. But like PNG, GIF can do well with diagrams.

JPEG is lossy, and not surprisingly, it performs better than PNG. JPEG has an adjustable quality parameter, and making it high makes a high-quality image with a large file, and making it low makes a small file with a low-quality image. Low-quality JPEG looks blocky, something that gives away how its image-compression works.

Similar considerations apply for audio and video. Lossy usually does better than lossless, and the smaller the file the worse the quality.
 
@lpetrich

Thanks... though I'm quite familiar with all of that...

This is what normally happens with loss in data for video compression:

Normally the artifacts are blocky... but what I saw was a round spot... and I tried rewinding the video and it didn't happen again... I think I'll call my TV manufacturer's hot line sometime...

Motion-vector-replacement-using-different-error-Concealment-methods-no-Concealment_Q640.jpg


47s27.jpg


BTW on the topic of PNG I highly recommend this free site:

https://tinypng.com/

It compresses the size to about 33% of the original size... it might be using 8-bit indexed colour though.... (which often looks the same) BTW PNG can perform better than JPEG for drawings without much colour... (as long as you compress the original - not try and compress a JPEG to PNG)

BTW GIF can be lossy in Photoshop:
GIF_preset.PNG


It adds noise that makes the image compress better....
 
Back
Top Bottom