It isn't edge enhancement... (1 Viewer)

[DaViD Boulet]

http://www.hometheaterforum.com/uub/...ML/028325.html
As Morgan Holly points out in this thread, many of the "edge enhancement" problems we see are actually idiosyncracies of particular MPEG encoders. Many studios are NOT deliberately using an "edge enhancement" feature.

My suspician is that many MPEG encoders may filter out fine detail and noise in an attempt to aid compression efficiency, and then add a touch of "sharpness" to try and compensate.
This would all happen without anyone intending it, would certainly be boni-fide "EE", and could also be described as artifacts of "compression" by studio execs since what went into the encoder was free of EE and what came out the other end had halos and ringing.
-dave

 

[Philip Hamm]

David,
GREAT point! MPEG artifacts are a function of the encoder software, the encoder person, and the decoder software. As we have recently found out with the whole "chroma bug" issue, mpeg decoders can make a great deal of difference in the picture coming out of DVDs.
I know for a fact that MPEG can create artifacts that appear to look exactly like edge enhancement. I also know for a fact that Edge Enhancement has been used for years even for some high quality transfers (I can see it in many LaserDiscs).
As I said before, I have no idea what is causing the problems these days with EE-like artifacts.
------------------
Philip Hamm
AIM: PhilBiker

 

[TedD]

I still want someone to identify the source of this problem and fix it!
This should not be rocket science here, just look at the image at each step in the mastering process and identify where it is introduced.
If it's the telecine process, educate the telecine operators or fix the equipment or software.
If it's the compression process, involve the manufacturers of the equipment and get the problem resolved.
Just don't try to tell us that it doesn't exist or that it was present on the film, because thats an out and out lie.
If anyone is really curious how bad this can look and lives in the DFW area, I'll be glad to arrange a demonstration and a comparison of good vs bad transfers.
Ted
[Edited last by TedD on October 03, 2001 at 09:01 AM]

 

[Michael Dueppen]

quote: My suspician is that many MPEG encoders may filter out fine detail and noise in an attempt to aid compression efficiency[/quote]
This is exactly what I think. It is often said that the Sony encoder is the most efficient encoder (i.e. it uses a low bitrate). Since Columbia Tristar are some of the worst "EE" offenders I always thought that this might be because the encoder artificially sharpens the picture and reduces grain to allow for more efficient compression. Unfortunately it seems that this cannot be turned off since even the Superbit discs seem to have the halos etc.
I hope that someone will find out what this is and stop it.
------------------
- Michael
My DVD collection
[Edited last by Michael Dueppen on October 03, 2001 at 09:54 AM]

 

[Dan Hitchman]

damn double post.
Dan
[Edited last by Dan Hitchman on October 03, 2001 at 11:50 AM]

 

[Dan Hitchman]

Is this showing up even in the master HD transfers or just the MPEG encoding stage for DVD?
I'd weep if EE and other compression anomolies start cropping up on HD-DVD as well.
Looks like we need lossless compressed HD 1080p video from new archival 1080p HD masters on HD-DVD so we don't have to put up with this nonsense anymore.
Why would Hollywood bitch about that? 1080p is still not even 35mm quality let alone 70mm or anything else!! Of course, this is Hollywood we're talking about. Money men, not savvy filmmakers and tech. geeks.
Dan
------------------
Stop HDCP and 5C-- Your rights are at risk!

 

[Jim A. Banville]

If you watch the scenes in the beginning of THE THING DVD, where you see the "rope" handrails along the footpaths between the buildings, you can see black lines directly above and below the ropes. Is this EE or MPEG artifacts?
thanks
Jim

 

[Bjoern Roy]

Dave,
quote: My suspician is that many MPEG encoders may filter out fine detail and noise in an attempt to aid compression efficiency, and then add a touch of "sharpness" to try and compensate.
This would all happen without anyone intending it, would certainly be boni-fide "EE", and could also be described as artifacts of "compression" by studio execs since what went into the encoder was free of EE and what came out the other end had halos and ringing.[/quote]
I agree 100% with what you and Michael said!
Philip,
i think you missed Dave's point. Only because the output of a certain MPEG-encoder "black-box" (in this case from Sony) shows halos around edges doesn't mean AT ALL that it is an inherent flaw of the basic MPEG encoding process.
It simply means exactly what Dave posted above: The Sony box seems to reduce the entropy of the signal by filtering the high frequency information (which needs the most bits) a fair bit, so that a lower bitrate is sufficient for the following encoding. To compensate for the slight softness that this introduces, it applies a certain amount of aperture correction (edge enhancement) to the image. AFTER THIS comes the actual MPEG encoding process, which is NOT the culprit for the EE artefacts in itself.
So, you could say: "That MPEG encoder box throws out images with halos, so it has to be a MPEG artefact." Thats what Dave thinks studio execs and telecine ops do. The result of this would be to accept that DVDs can't look better, which is of course nonsense, there are some excellent EE free transfers (some with low bitrate, some with high!).
But in reality, you would have to say: "That Sony BOX seems to apply EE automatically before it starts the actual MPEG encoding process." So a solution would be not to use the Sony encoder, but one of the 2 that Morgan mentioned in the other thread. He did 'Remember the Titans' with one of them and its one of the best, detailed, non-EE transfers every released.
Ted,
i totally agree. We should investigate what equipment was used on the transfers we don't like EE-wise and try to isolate which machine(s) is at fault.
________
And please folks... the idea that the EE artefacts (especially the ones as heavy as in DHWAV) are originated in the film print is the most ridiculous thing i have heard in years!
After the DHWAV disaster, Peter only quoted his telecine folks "they aren't aware that they added any EE on purpose". I never said i doubt this! But "if they didn't add it, it has to be in the film print" is simply the most miseducated conclusion imaginable.
The only conclusion is: If they didn't do it and its there, it has to be introduced by one or more machines in the telecine/compression chain. Nail the sucker and get rid of it!
And its not as if FOX would have to buy new equipment for millions of dollars. A lot of FOX transfers are breathtaking. So they seem to have access to equipment that isn't flawed!
Peter, what equipment/team was used for "Independence Day", "X-Men", "White Men Can't Jump" or "Inventing the Abbotts", to name a few sensational FOX transfers without any EE?
And what was used on the new "DH 3", "Star Wars TPM" to name 2 recently criticized transfers?
Another interesting thing is, that there are just as many non-major (lower budget) movies among the great looking Fox transfers as there are big hits. That implies that they don't have an "excellent but expensive" telecine bench next to an "crappy but cheap" one! Because you would think that if this was the case, they would use their best equippment on the biggest hits, no? Its seems more like HIT OR MISS with their efforts! So please lets get at the bottom of this.
------------------
"Never underestimate the predictability of stupidity" (Bullet Tooth Tony in 'Snatch')
My HT in action (Screenshot Page) | My Ultimate 'Edge Enhancement' Guide | My DVD/LD SPL page
[Edited last by Bjoern Roy on October 03, 2001 at 05:43 PM]

 

[Bjoern Roy]

Cris, Philip, and others,
let me tackle the "Most of the EE that people complain about is in fact only MPEG artefacting" issue again in more detail.
Short answer: No its not.
Long answer will take 2 approaches trying to convince you. First through simple logic and second with a concrete example.
I already presented the 'locical' approach in another thread, so i quote myself here:

...the myth that the EE we see on these titles isn't actually aperture correction (EE) but rather MPEG compression artefacts is simply not true.
While low bitrates can also induce edge artefacts (called mosquito noise), those look distinctively different and only happen at bitrates FAR below the ones that are used on normal DVDs.
If the speculation on this issue won't settle down soon, i will have to preper an article about the different artefacts. But even without presenting the technical details, you can easily tackle the topic with plain logic and some experience. Take a look at this trivia:
a) 'Titanic', which has among the least amount of EE i have ever seen, is at the same time probably one of the best examples of severe compression artefacts (mosquito noise)! The long running time and the incredible unfiltered detail in the transfer yields lots of MPEG artefacts throughout the movie.
b) 1.78:1 transfers use 120 more lines of resolution than 2.35:1 transfers. Thus, if they are encoded at the same bitrate, the data has to be compressed MORE on 1.78:1 material. Yet, the lesser compressed 2.35:1 transfers have more EE on average than the 1.78:1 ones! Col/Tri and NewLine are prime examples.
c) 'Starship Troopers' from Col/Tri is an early heavily compressed single layer transfer. There are several scenes with compression artefacts. Yet, its probably the best Col/Tri transfer ever with no EE at all!
d) DVDs are encoded with a variable bitrate. The more movement in the picture, the more bitrate is needed to avoid artefacting. If the ringing in the picture were indeed due to the compression, things would have almost no halos if no movement is in the frame. Once the camera would move or an explosion would occur, halos would all of a sudden appear. Yet, everyone who has ever seen EE on a title knows that it is fix. While the amount and thickness can change from scene to scene, it stays the same throughout that scene, no matter how much happens on screen!
e) I have some MPEG1/2 movies on my PC that have 100 times as much compression artefacts as any DVD i have ever seen (yes, even including Highlander) and use a fraction of the bitrate used on DVD. Yet, they only show severe macroblocking and softness, but nothing closely resembling the distinct constant halos on DVD transfers.
f) If you simply take 50 DVDs, some with lots of EE, some without, some with high bitrate/dual-layer, some with low bitrate/single-layer, you won't find any correlation between EE and bitrate AT ALL.
If someone doesn't see the discrepancy after the simple logic presented above, i doubt a technical explanation would convince him.

Now, since the logical, experience based approach above might still not convince the skeptical folks, i prepared a little example:

Explanation:
- T2 Ultimate Edition is a title with lots of excessive edge enhancement that i also use in my EE guide (link in signature).
- The top-left image is a 200x200 pixel crop of a screenshot from T2. You can see 2 things here. First, there is a bright halo around the letters. Second, in the proximity of the letters, there is a bit more noise than in the rest of the image. The halo is due to edge enhancement, the noise (also called mosquito noise) is due to insufficient bitrate, thus an actual MPEG artefact.
- Analyzing this capture alone there could be arguing that the halo is ALSO a MPEG artefact, as some here seem to apply. To show that it is not, i prepared 5 more images.
- The bottom-left image is a simulation of a similar text on similar noisy background. This shows how terrible the DVD transfer actually is and how crisp, clean and yet smooth the letter edges could be if the DVD transfer would use the DVD resolution to its fullest. No halo, no nothing. Now this is uncompressed, so its not fair, you say. Well, lets see then.
- The bottom-middle image shows the effect of MPEG encoding with a compression ratio that is close to what is used on DVDs. If you look closely, you will see slight mosquito noise around the letters, but no halos where introduced at all!
- The top-middle image shows the effect of ultra heavy MPEG encoding with a compression ratio that is FAR above what is used on DVD (more like the worst of the worst digital TV sources). The mosquito noise is very apparent now and you can easily see the 8x8 pixel sized blocks which are used in JPEG/MPEG encoding. Basically, every 8x8 pixel block represents a 'pattern' that approximates the real picture content. Still, you don't notice anything that would look like the EE that we complain about on DVD. And again, this is with a bitrate that is FAR lower than the lowest used on DVD (0.5-1Mbit instead of 4-8Mbit).
- On the top-right, i applied EE that produces ringing/halos similar (a bit worse) to what we see in the DVD transfer.
- On the bottom-right, i also compressed the above, again with a compression ratio similar to a high bitrate on DVD you see slight mosquito noise around the letters again. This one resembles the issues with the T2 DVD crop: Halos from EE + noise from MPEG artefacting
So, MPEG compression is not the culprit for EE. Its guilty of other things, though. Blocky artefacts especially in darker areas, loss of fine detail, an effect called MPEG 'wobbling' etc...
I will try to get another example up where the difference between EE and MPEG ringing is not quite as apparent as above. But this should give you a first impression.
------------------
"Never underestimate the predictability of stupidity" (Bullet Tooth Tony in 'Snatch')
My HT in action (Screenshot Page) | My Ultimate 'Edge Enhancement' Guide | My DVD/LD SPL page

 

[Cris K]

I have seen examples like the one in the middle mistakenly called "minor examples of edge enhancement"
Obviously, like you say, if it is constant- it cannot be related to the variable rate compression.
-cris

 

[Nathan_H]

Excellent reasoning about why it's not MPEG compression that's causing the ringing.
But it sounds like the problem is that *after* the MPEG compression, but before leaving the "black box" where the MPEG compression occurs, the unit may ALSO try to "sharpen" the picture. So it's this "sharpening" (done, apparently, at a fairly constant rate), after the variable-rate MPEG encoding occurs, that may be introducing the ringing.

 

[TedD]

Thank you, Bjoern, for seconding my comment on the totally ridiculous statement that "it must have been in the print".
This should be a simple problem solving exercise involving a little bit of logic and some critical viewing.
I cannot understand why we can't get the people responsible for this process (film to DVD)to identify the problem for us and then get something done about it.
Ted

 

[Chris Maynard]

it must have been in the print

That was never said by Peter or anyone else at Fox.

 

[Philip Hamm]

quote: Philip,
i think you missed Dave's point. Only because the output of a certain MPEG-encoder "black-box" (in this case from Sony) shows halos around edges doesn't mean AT ALL that it is an inherent flaw of the basic MPEG encoding process.[/quote]Bjoern, I think you may have misunderstood my post. I'll add emphasis so that it's more clear.quote: MPEG artifacts are a function of the encoder software, the encoder person, and the decoder software. As we have recently found out with the whole "chroma bug" issue, mpeg decoders can make a great deal of difference in the picture coming out of DVDs.
I know for a fact that MPEG can create artifacts that appear to look exactly like edge enhancement. I also know for a fact that Edge Enhancement has been used for years even for some high quality transfers (I can see it in many LaserDiscs).
As I said before, I have no idea what is causing the problems these days with EE-like artifacts.[/quote]The word "can" is key. (though it seems that I was incorrect. Your reply seems to accuse me of stating that all MPEG video has these problems. Nowhere do I say that, nor did I, because it is not true. Sorry if I was not clear.
Thanks for the exhaustive post. It seems that I was ill-informed about MPEG causing this type of artifact. The last sentence in my post it seems is no longer valid (though It was valid yesterday). I wonder why there wasn't more of an outcry about this back in the LaserDIsc days? Did everyone assume it to be an NTSC artifact perhaps?
Is that information on the "Edge Enhancement" web site?
I'm archiving this thread.
------------------
Philip Hamm
AIM: PhilBiker
[Edited last by Philip Hamm on October 04, 2001 at 08:02 AM]
[Edited last by Philip Hamm on October 04, 2001 at 09:41 AM]

 

[TedD]

From an earlier post in this thread:
"Peter Staddon contacted the people that did the transfer on Die Hard WAV. It is NOT edge enhancement, but part of the actual physical print."
I stand corrected. Peter Staddon was told "that it's in the print".
By the way Chris:
It's obvious that excessive edge enhancement doesn't bother you as much as it bothers some of us, based on a search of many of the previous threads on this topic.
I personally don't believe that any 65" rear projector can possibly show a DVD to its full potential. That takes a CRT FPTV with a scaler at at least 720x1280 progressive.
With that said, I still can't understand your ardent defense of transfers that clearly could have and should have been free of the extreme amounts of edge enhancement being discussed here.
However, each of us has different sensitivities to different issues. For example, to someone who might have grown up with TV as a reference, edge enhancement may be considered a normal part of watching DVD at home.
In my case, critically watching and projecting film for 50 years makes me very sensitive to anomalies like edge enhancement that can't possibly be originated in the film process and relatively insensitive to things like grain and dirt, which are part of the film process.
I know that Frank Manrique shares my sensitivites and opinions, and it appears that Bjoern Roy also does.
Chris, if you ever get to the DFW, Texas area, please contact me and we'll set up a viewing session.
Ted
[Edited last by TedD on October 04, 2001 at 08:41 AM]

 

[Bjoern Roy]

Nathan,
the 'sharpening' can't be AFTER the mpeg compression, because at that time you have a compressed stream that you can't alter in any way. It has to be before the mpeg compression.
Chris,
i can't remember exactly whether Peter mentioned that its in the print, or if it was assumed by other members otherwise. I think the latter, thats why i defend Peter in my post above. The idea is ridiculous either way.
Its like a audiophile going into a music store with a Bose system, placed completely wrong no less, that plays a song you are familiar with and it has terrible one-note boomy bass from hell. You complain to the store owner and he says "Oh its all in the soundmix, all of the CDs are mixed like that". You would be rolling on the floor!

Philip,
oh sorry if you think i was disagreeing with your whole post. I should have quoted what i was referring to. Let me emphasize:
quote: I know for a fact that MPEG can create artifacts that appear to look exactly like edge enhancement.[/quote]
My point was, that the MPEG artefact doesn't look EXACTLY like EE, not even close in the example i presented and still easy to seperate from it in another example i will post.
Ted,
while in agree 100% that equipment plays a big role in this debate, in case of Chris, i highly doubt its his system. He just doesn't seem to be as sensitve to the issue like some of us. He mentioned that among the few EE cases that really bother him is 'True Lies', while he only found several instances of 'DHWAV SE' to be noteworthy. I find 90% of the movie unwatchable and i am sure i also would on his system!
Thats perfectly fine with me, though. I am, for example, absolutely insensitive to film defects! The dirt in e.g. Braveheart doesn't bother me AT ALL. And in other, even worth cases, it doesn't as well. And that although i am sure its a lot more apparent on my resolving system than on that of most others. I would go as far as saying, i love it

Makes the transfer look that bit more film like. Shouldn't be TOO excessive, of course, Braveheart is just perfect in that regard.
(Pssst, i often think it would be nifty if transfers had the cigarette burns and such still in it

)
Regards
Bjoern
------------------
"Never underestimate the predictability of stupidity" (Bullet Tooth Tony in 'Snatch')
My HT in action (Screenshot Page) | My Ultimate 'Edge Enhancement' Guide | My DVD/LD SPL page
[Edited last by Bjoern Roy on October 04, 2001 at 09:44 AM]

 

[Chris Maynard]

Ted -
quote: Peter Staddon contacted the people that did the transfer on Die Hard WAV. It is NOT edge enhancement, but part of the actual physical print[/quote]
No that is not what was said either. Where does this stuff come from? It reminds me of the "grapevine" in Johnny Dangerously.
All that was told to him was that little or no edge enhancement was put into the transfer.
BTW: My little ol 65" setup reveals everything that people see on 120" screens. Yes it may not be as obvious but I see the same things that everyone else does.
Like I also said...I watched the TPM DVD on a 30+ foot screen last week and while I noticed the ringing and could point it out...it was not that distracting. In fact most people did not see it at all. Was that system not revealing enough?
People are blowing this problem out of proportion.
Bjoern my friend - 90% unwatchable? Good grief!
I am glad that I am not overly sensitive. With me I notice the edge enhancement, think about it for a few seconds, then I proceed to enjoy the movie.
I think I am enjoying this hobby more than a lot of you.

[Edited last by Chris Maynard on October 04, 2001 at 10:14 AM]

 

[DaViD Boulet]

Bjoern Roy,
You are the man. That visual comparison of EE and MEPG artifacting is incredible. WOW. If you ever run for president of world-wide DVD mastering I'll vote for you

Thanks for the exhaustive post. It seems that I was ill-informed about MPEG causing this type of artifact. The last sentence in my post it seems is no longer valid (though It was valid yesterday). I wonder why there wasn't more of an outcry about this back in the LaserDIsc days? Did everyone assume it to be an NTSC artifact perhaps?

Laserdisc looks like mush on a big screen no matter what you do. Chroma noise. Comb-filter artifacts. It also doesn't have enought inherent sharpness to really let that EE come through to its full degree (which is why it was added in the first place--to try to sharpen up the soft image from laserdisc and VHS).
I remember seeing state-of-the-art Faroudja-line-quadrupled LD on the world's reference LD player projected by Stacked Runco projectors at the stereophile convention in NY in '96. It looked like crap. I honestly couldn't understand how anyone could have bothered with all that $$$ projection eqiupment when the end result was clearly sub-16mm quality.
Even now when I "zoom" a letterboxed laserdisc transfer to fill my 16x9 screen (Pioneer CLD-99 Elite player) I'm painfully aware how quickly I'd swap that disc for a DVD! (now if only we could get studios to start using PCM and high-bit-rate DD encoding for the 2.0 and 1.0 soundtracks...so then my DVDs could SOUND as good as my blurry LDs...)
-dave

 

[Philip Hamm]

Laserdisc looks like mush on a big screen no matter what you do. Chroma noise. Comb-filter artifacts. It also doesn't have enought inherent sharpness to really let that EE come through to its full degree (which is why it was added in the first place--to try to sharpen up the soft image from laserdisc and VHS).

Your opinion I guess. During the summer when I went to the Scooterplex I was pretty impresse with the LD material we saw. I'm also impressed with LD in my house, but then again I have an older 50" NTSC set.
Anyway, yes, believe it or not, edge enhancement jumps out just as much on LD as it does on DVD. Take a look at the "Beatles Anthology" set for the worst example I can think of. The Edge Enhancement is almost intolerable on that title if you're sensitive to it.
EE has always bothered me to some extent, but I guess I've assumed that it's an NTSC and / or MPEG artifact.
------------------
Philip Hamm
AIM: PhilBiker

 

[Ken_McAlinden]

The 1997 VHS release of George Stevens' Giant has the most gruesome enhancement I have ever seen. The characters look like homicide victims with police-drawn chalk lines around them. I can only imagine how much more noticeable it would have been at LD or DVD resolutions.
Regards,
------------------
Ken McAlinden
Livonia, MI USA