Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

Video encoding requires using your eyes

redvice.org

Post reply on HN
Print view View original

41–50 of 53 posts

Re: Video encoding requires using your eyes

#41

Post by cvz »

xmprt wrote: post #11

Is it just me or does the author here sound like they're hating just to hate? The writing doesn't sound that terrible. Maybe it's a bit amateur but isn't that what you're expect from an engineering blog post written by people whose day job is to write code. And for a layman, the image comparison isn't as bad as they make it out to seem.

I am unsurprised that the author would hold a video distribution company which supposedly pays good money for experts to a higher standard than the average hobbyist blogger. I don't think it's hating for the sake of hating.

Re: Video encoding requires using your eyes

#42

Post by astrange »

cvz wrote: post #39
astrange wrote: post #34

Earlier quoted context omitted.

That is what it should mean, but in image compression people use it to mean "mosquito noise" artifacts, which come from quantizing DCT compression applied to edges. (Nyquist theorem = DCT is bad at edges because they're made of an infinite number of frequencies.)

It's the exact same phenomenon in both. Not sure where you're making the distinction.

It doesn't come from "oversharpening".

Re: Video encoding requires using your eyes

#43

Post by bestham »

astrange wrote: post #34
dist-epoch wrote: post #7

Earlier quoted context omitted.

Ringing is pretty obvious to me. It has a specific meaning in this context, it means edges are over-sharpened to the point that "fake" extra edges appear. https://en.wikipedia.org/wiki/Ringing_artifacts

That is what it should mean, but in image compression people use it to mean "mosquito noise" artifacts, which come from quantizing DCT compression applied to edges. (Nyquist theorem = DCT is bad at edges because they're made of an infinite number of frequencies.)

That is the exact same phenomenon. Artificial sharpening is introducing high bandwidth components to a signal. If you bandwidth limit (low pass) a signal to fit below the Shannon-Nyquist limit you will get ringing as the signal cannot be represented accurately and will smear in the time domain. Given a bandwidth constraint, artificial sharpening above a certain threshold will result in ringing.

Re: Video encoding requires using your eyes

#44

Post by bestham »

HPsquared wrote: post #25
crazygringo wrote: post #13

Earlier quoted context omitted.

I agree the "negative" artifacts are almost impossible to see, and came here to the comments to see what the heck the author was talking about. > People who work deeply with codecs are usually hypersensitive to these sorts of issues that mere mortals like us need to try to see. I think that kind of shows that the author is unfairly critical. They're saying "this should not have shipped", when it seems just fine to us…

Also video is viewed in motion, not as static frames. And end-users watching on low bitrates aren't going to freeze-frame and zoom in.

That is true. But why would you replace a simple known to be working filter with this thing that yields worse performance? There are no upsides to this, and that is why the domain expert is baffled. We have been making filters like this for a long time, it is known how they should be evaluated and yet this thing is still touted as state of the art but its creators. If the image quality is lower with this filter and you say that is does not matter then the quality of the stream is to high. This filter is not going to solve that.

Re: Video encoding requires using your eyes

#45

Post by HPsquared »

bestham wrote: post #44
HPsquared wrote: post #25

Earlier quoted context omitted.

Also video is viewed in motion, not as static frames. And end-users watching on low bitrates aren't going to freeze-frame and zoom in.

That is true. But why would you replace a simple known to be working filter with this thing that yields worse performance? There are no upsides to this, and that is why the domain expert is baffled. We have been making filters like this for a long time, it is known how they should be evaluated and yet this thing is still touted as state of the art but its creators. If the image quality is lower with this filter and y…

It apparently did better with users' subjective evaluations. I guess they liked the "extra detail" look, even if it's fake. End users aren't zooming in to freeze frames or, heaven forbid, taking screenshots (think of the copyright!!)

Re: Video encoding requires using your eyes

#46

Post by ZeroGravitas »

okramcivokram wrote: post #2

I don't see (or maybe don't recognize) any issues with the image that they're talking about (ringing, color shift, or fake details). It most certainly doesn't look awful to me, it looks exactly the same, only a bit sharper.

I don't immediately see the issue either.

Though, even if I could, this is a new way to preprocess an image before feeding it into an encoder, and the examples have both been fed through the new downsizer, then the standard encoder, presumably at the standard Netflix bitrates and then (I think) upscaled back to the original size.

So if this didn't look a little compressed then that would be a methodological mistake, as you don't use downscaled encoding unless you've already decided that a full size encode has too much quality for your task.

And Netflix generally has incentives set up to reduce quality until their customers notice. That's why they quote stats based on that.

In web dev terms it's like reducing the size of your product images until it hurts sales. You're almost guaranteed to have artifacts visible to image compression experts before you hit the point that it affects your bottom line. If you are targeting customers on slow internet (and again if you are downscaling then you basically are) your sales are likely to initially rise as you get usable pictures to people faster.

Re: Video encoding requires using your eyes

#47

Post by soulofmischief »

HPsquared wrote: post #45
bestham wrote: post #44

Earlier quoted context omitted.

That is true. But why would you replace a simple known to be working filter with this thing that yields worse performance? There are no upsides to this, and that is why the domain expert is baffled. We have been making filters like this for a long time, it is known how they should be evaluated and yet this thing is still touted as state of the art but its creators. If the image quality is lower with this filter and y…

It apparently did better with users' subjective evaluations. I guess they liked the "extra detail" look, even if it's fake. End users aren't zooming in to freeze frames or, heaven forbid, taking screenshots (think of the copyright!!)

End users aren't domain experts. That's like if I was allowed on the factory floor where my favorite products are made so that I could make subjective, uninformed decisions about the manufacturing process which also affect everyone else.

Re: Video encoding requires using your eyes

#48

Post by AlotOfReading »

HPsquared wrote: post #45
bestham wrote: post #44

Earlier quoted context omitted.

That is true. But why would you replace a simple known to be working filter with this thing that yields worse performance? There are no upsides to this, and that is why the domain expert is baffled. We have been making filters like this for a long time, it is known how they should be evaluated and yet this thing is still touted as state of the art but its creators. If the image quality is lower with this filter and y…

It apparently did better with users' subjective evaluations. I guess they liked the "extra detail" look, even if it's fake. End users aren't zooming in to freeze frames or, heaven forbid, taking screenshots (think of the copyright!!)

Yeah, but there are classical algorithms with tunable sharpening that are cheaper and widely known (e.g. catmull-rom or the "magic" kernel sharp algorithm by John Costella).

My suspicion is that none of this mattered though, because the evaluation was probably "perceptual equivalence" vs bitrate. I can easily believe it might be a marginal win over traditional algorithms from that perspective.

Re: Video encoding requires using your eyes

#49

Post by cvz »

astrange wrote: post #42
cvz wrote: post #39

Earlier quoted context omitted.

It's the exact same phenomenon in both. Not sure where you're making the distinction.

It doesn't come from "oversharpening".

Sharpening and bandwidth-limiting have the exact same effect, because the maximum sharpness of an image (like any other signal) depends on its bandwidth. There is no difference in the type of artifact produced. That's why the artifact from both has the same name of "ringing".

Re: Video encoding requires using your eyes

#50

Post by trehalose »

soulofmischief wrote: post #47
HPsquared wrote: post #45

Earlier quoted context omitted.

It apparently did better with users' subjective evaluations. I guess they liked the "extra detail" look, even if it's fake. End users aren't zooming in to freeze frames or, heaven forbid, taking screenshots (think of the copyright!!)

End users aren't domain experts. That's like if I was allowed on the factory floor where my favorite products are made so that I could make subjective, uninformed decisions about the manufacturing process which also affect everyone else.

Is it? Or is it like if you were allowed into a focus group where they let you try out two manufactured products, so that you could make subjective, uninformed decisions about the product lineup which also affect everyone else?
Post reply on HN