What is the truth about EDGE ENHANCEMENT? (1 Viewer)

[Vince Maskeeper]

Let me start this little editorial off by warning you that I have very pedestrian knowledge of video signals, their creation, and their limitations. I am, as many of you know, an audio guy primarily with a mind that tends toward geeky concepts and technical mumbo-jumbo. I like computers, recording equipment, home theater, candle light dinners and long walks in the park...
But I digress.
I have been very interested in the idea of "edge enhancement"- and have followed the discussions, indictments, screen shots and examples very closely. While I never had much to directly add to the conversation- I lurked nonetheless in hopes of gather more knowledge on the topic, in the unending pursuit of perfect video experience...
In the course of reading the extensive discussions about Edge Enhancement, the one issue which really stuck in my crawl, so to speak, was the fact that even when it was as obvious as the nose on your face-- studios and technical types denied it was applied. As a natural cynic, I considered that these folks were simply flat out lying- but a logical consideration of why they would bother lead me to no answers.
I started to think this whole thing might be a misunderstanding- an issue of syntax or semantics which has thrown confusion into the mix and resulted in poor communication.
Often, in the discussion of audio concepts, there is much confusion because people improperly apply terms and concepts which they don't understand. Standards vary, terminology gets misused and misunderstood, and even basic processes are often tough to grasp-- as a result sometimes helping someone with an audio problem starts with a 20 minute process of educating them on the terms they're using and the concepts they're employing before getting to the meat of the problem.
In order to address the issue of edge enhancement- and specifically possible misunderstanding which is leading to misinformation- I think we have to establish the popular concept of what this is and how it happens.
I think I understand the idea as follows:
Edge Enhancement is an intentional artifact, introduced during the telcine process, which causes halo-ing of edge lines. The telecine operator uses some sort of optical process (or possibly digital process, I'm still unclear) to skew edge lines so they look "doubled". This results in an illusion of added resolution on smaller displays- as the smaller picture size doesn't display the added "enhanced edge" as a second line- rather they blend together to add crisp edges to things which originally had none.
Unfortunately, on larger displays- the enhanced edge becomes far more obvious, and appears more as a pronounced halo surrounding the original image.
Edge enhancement is applied specifically during the film to video transfer process (again, I'm not sure if this is the argument, but I believe it to be)- and is done on purpose to create an illusion of extra crispness and resolution to a video transfer.
So, back to the issue of syntax and confusion. As far as I can calculate, the term "edge enhancement" is not one used by the industry at large. I have spoken with one telecine operator, and have seen a second post here on the forum, and both referred to the process as simply "enhancement". The process of "enhancement" as they described it, sounded essentially the same as the concept discussed by observers here as "edge enhancement".
Now, here comes my main reason for posting this piece:
I worked on an independent movie called Dev/Null, I served as audio designer and essentially performed every duty required of the audio department. The movie was shot on Digital video using Canon excellent prosumer camera, the XL-1. The XL-1 is a very nice 3 CCD mini-DV camera which allows interchangeable lenses and manual controls like aperture which go very far in creating a profession image.
The movie is "finished" and the producers sent the DV tape off to a place to have some DVDs produced to use for festival submissions (and as the master for VHS dubs for festival submissions). These folks are quite budget minded, so instead of using the professional studio I suggested- they used some company from the back of a magazine which offered DVD copies from DV tape for $25 a piece.
I borrowed the DVD from them, so I could watch the finished cut with the DP (who had never seen the final edit)- and noticed something very interesting.
Edge Enhancement
A movie which was shot on Digital Video, edited in a PC, and mastered to DVD had very clear edge haloing, nearly identical in some scenes to the obvious examples presented from the Phantom Menace. Heck, this DV sourced movie also presented what I've seen called "Adaptive" enhancement- which supplies enhanced edges only on certain surfaces in particular directions.
So the question, for me, becomes- if edge enhancement is the same as the telecine process of "enhancement"- why am I seeing an identical artifact of a movie which never once had anything to do with a photo-chemical carrier like film, never experienced a telcine process, and stayed 100% digital from the moment it was shot?
I did some close study on the matter. Went back to the DV source tape for the DVD master, and would A-B them on my projector. The enhancement that was clear on the DVD was not at all present in DV tape. Going back to the DV tapes on which the material was shot- again no edge artifacts.
So I busted out my Phantom Menace DVD, since it seems to have become the standard for how bad edge enhancement can be, and started to notice some common elements between he Inge problems on the TPM DVD and the edge problems on the Dev/Null DVD.
Namely, it seemed the edge enhancement was most clear and pronounced when a moving foreground object moved across as static, or mostly static background. Scenes in TPM, such as space ships flying across a pale blue sky seemed to display this artifacting horribly. Also, the royal guard's hat against the desert backgrounds- again showed it badly.
On the Dev/Null DVD- one scene featuring an very dark skinned african-american actor who has a shaved head standing against a white wall in an office displayed extensive edge artifcting.
While these weren't the exclusive location of visible enhancement, they were certainly the most common and obvious locations to find the artifact.
So, the finding of "edge enhancement" type artifacting on the Dev/Null DVD cause me to wonder what the source of the artifact really is... and this above info on moving images on fixed backgrounds made me seriously consider the concept that maybe this enhancement is caused by MPEG compression. It stands to reason that certain types of compression would more tightly compress static, homogeneous background images-- and thus the point where a moving complex image passes across that- where they meet would be a potential problem for the compression scheme.
These seems to be suggested as a distinct possibility as I recently saw a thread in the TV area suggesting that similar "edge enhancement" had been applied to the live HD sports feeds- which is either a result of a similar compression problem- or a misnomer of the idea of "edge enhancement" as it has been discussed as a telecine operation.
In addition- I believe that currently most DVD releases are getting a "downconverted" master from a HD telecine master-- yet when reviews I've seen of DVHS HD material from (assumably) the same transfers as the DVD material- the issue of edge enhancement hasn't been mentioned.
So, any one of these seems to be a resonable possibility- if not likely possibility in a large number of the thinks illustarted as "edge enhancement". I think as MPEG compression techniques have developed, new programs and schemes have been used- the "quality" of the image vs the amount of the compression have shifted around quite a bit.
I think what I'm trying to say here is this- The possible reason why we've gone round and round with technical people from studios who deny edge enhancement was applied stems from a misunderstanding between our views and theirs. Some have suggested that it is possible that a few telecine machines apply enhancement by default- so the operators deny applying it, because it was applied without their knowledge. I think it seems even more realistic that some sort of process related to MPEG compression or the filtering occurring from a downconversion from a HD or higher resolution master is causing these edge artifacts. Or, even, that the filtering is done once the transfer is completed and in the digital realm, and thus when asking about "edge enhancement" applied, there becomes some confusion in what this means exactly...
As a result, when the operator who struck the transfers for Die Hard swears that no edge enhancement was applied... or when THX swears that no edge enhancement was used on TPM-- I can't help but wonder if they are literally telling the truth- but because we are misguided as to the source of the artifacts or their exact cause, and as a result we are asking the wrong questions.
-Vince

 

[Chad Ellinger]

I can't seem to find the thread, but I distinctly remember Robert Harris commenting about the misconceptions about edge enhancement. Mr. Harris specifically mentioned that many of the examples of edge enhancement we see are unintended side effects of certain types of video encoders and are NOT added intentionally.
Of course, this is all from memory, so someone please correct me if I'm wrong.
EDIT:
Found this thread where Morgan Holly discusses some of the "edge enhancement" in various encoder hardware.

 

[Matt DeVillier]

Vince,

the main problem I have with the "EE is a MPEG artifact" theory is that there are several instances that suggest otherwise, the most prominent being the lack of EE on the Phantom Menace trailer on the DVD compared to the feature itself. And how can we explain that some transfers have absolutely no EE at all, whereas others like TPM make one cringe?

I also have been silently following this (and even experimenting some with different MPEG-2 software encoders), but I have been unable to come up with anything that will produce the kind of EE we are seeing. I hope that we figure this out soon though, as it's a shame to see an otherwise outstanding transfer ruined by EE

 

[Michael St. Clair]

It shouldn't be hard to track down, wherever it comes from.
The academy screener DVD of 'Cast Away' is basically EE-free. But the version in stores is not.
Same for Episode One DVD versus the DVD trailer of the same.
Analyze and compare between how examples like these were produced (hardware chain, parameters), and you will find the answer.

 

[Troy LaMont]

Vince,

Excellent read and analysis. You are pretty much spot on from my what I'm understanding as well.

A few of my own notes;

It has been pointed out before that this may be an anomaly of the MPEG2 conversion process.

It has also been reported over at Widescreen Review that the T2, D-VHS tape also exibits some EE but not as much as the DVD, both are MPEG2 compressed.

A question that comes to my mind that has also be brought up before is the 'digital' nature of certain scenes and how they transition to the final video format.

I also think the Bjoern has pointed out that certain MPEG2 encoders produce better output than others (I remember the Sony encoder/decoder being mentioned but I'm not sure if it was good or bad).

I also agree on the fact that we may be talking apples while the studio tech guys are talking oranges (or maybe just a diffent apple; red vs green).

Your post really just leads to the fact that somewhere along the lines of production, artificial enhancement is introduced to the final product. We need to find out where in the process and why and eliminate it, period.

Troy

 

[PhilipG]

Steve Roberts (head of the much-lauded Doctor Who Restoration Team) had this to say about EE on a recent discussion:

Edge enhancement is routinely added as part of the telecine process in order to make up for the losses in the scanning system. The correct amount of correction should just make up for the optical roll-off. Over-correction results in the sort of problems you noticed. It's added in both the horizontal and vertical planes, but diagonally filtered to prevent diagonals getting a double dose of correction from both the vertical and horizontal correctors adding together.
It can also be added in the DVD mastering stage - particualrly if they have rolled of the high frequencies to allow easier compression. If you fileter the hf off with a brickwall filter, but then add a little bit of HF lift at the brickwall point, it gives the illusion of some hf detail still being there.

The thread in question is at Link Removed
As far as I am concerned, the (over)use of EE is almost as bad as pan/scam.

 

[Bjoern Roy]

Vince,
its no MPEG artefact. I explained this many times, at least once so foolproof that the chance that the earth is a disc is higher.
Simple example, the trailer on the TPM DVD has a lot more compression artefacts than the feature presentation but lacks ANY edge artefacts.
I will dig out my older post since i am too tired explaining the same thing over and over again. Its like when i would ask you to explain the LFE DD/DTS differences for the n-th time again. Then again, watching you doing that is a lot of fun, so maybe you want to tease me.
More later.
Regards
Bjoern

 

[Shayne Lebrun]

You'll also see similar haloing on streaming video feeds of live events. I remember watching the realvideo feed fo CBC news in the office, on 9/11, and noticing the halos.

 

[Vince Maskeeper]

the main problem I have with the "EE is a MPEG artifact" theory is that there are several instances that suggest otherwise, the most prominent being the lack of EE on the Phantom Menace trailer on the DVD compared to the feature itself.


I think that is actual an indicator that the EE is something other than a telcine artifact.

I think compressionists would rather shave a little digital space off the feature (where it will have a real impact to vailable dataspace) versus knocking up the compression on a 1 minute trailer which will gain very little.

I'm not involved in the pro world of video transfers- but I would think it would be possible and maybe likely that the transfer for all film elements for a potential DVD release would happen at the same time (specifically for new releases which don't have old trailer transfers lying around). If you had to do a transfer for the feature and for the trailers and a few other things- why would you bother booking time to do it on different occassions? Why not just thread up the machines and get all the elements done at the same time?

Again- it seems more likely, to me, that when shaving off data to meet a bit budget, upping certain compression settings on the feature would gain much more that squashing the holy hell out of a trailer- thus the trailers get a better compression configuration (or more specifically less time is spent on compression tweaking, so default settings are used) and less edge artifacts result.

I think the confusion arises because there are other "stereo-typical" artifacts in the TPM Trailer which seems to point to the fact that it is "more compressed" than the feature. But, Mpeg compression at that authoring level is far more complex than a simple cmpression ratio... There are literally dozens and dozens of elements to control in a given set of parameters for MPEG compression (which is why it is often seen as much art as science).

As MPEG encoder technology has advanced, this edge enhancement phenomenon sees to have also increased- which makes me wonder if it is possible to apply certain settings of MPEG compression which when over used would cause one type of artifact (stereotypical mpeg artifacts with us from day one) while excessive application of another type of compression setting would create different types of artifacts.

What I've been thinking is that there is some newer compression packages with built in filtering and filter settings. These specific authoring tools offer better controls and tighter compression without the stereotypeice compression artifacts associated with MPEG2. INstead, these new tools, when over used display their own artifacts- and that is the enhancement we see.


Of course- this issue could also be going back to the issue of a filter element in the HDTV downconversion process as well. If the feature was transfered Hi-Def, but the trailer and other materials were not- it's possible that the edge enhance artifact was introduced to the feature in downconversion.

This makes me wonder if the PAL disc came from the same film to video transfer HD downconvert (or if that can even happen)-- and if so, why would it not display the same problems (if the artifact was present in the transfer process).


The academy screener DVD of 'Cast Away' is basically EE-free. But the version in stores is not.


This seems to create some questions immediately- like was the screener anamorphic? Was there any reason to believe or not believe that the two transfers (screener and finished dvd) were different at the telcine process- or rather were the same film to video (which seems likely as it's an expensive process) with different compression.

Again- I'm just curious if we go all the way back to when the transfer from film was made- if the enhancement was applied there. And if it wasn't, I think the confusion happens because addition processes to fake resolution might not be technically known as "edge enhancement"- and so asking a professional operator will get a different answer than the one we seek.

If, like with the T2 example, there is less "enhancement" in the DVHS version- and if we could some how prove or assume they came from the same struck master- then it immeidately points to a process occurring after the actual transfer.

While this question of where exactly the artifact occurs seems like nit-picking-- I think it is an important issue to understand in order to open a meaningful dialog with professionals in this field. We seem to be able to quantify what we're seeing... however I think it will be equally imposrtant to be able to communicate it in terms open to what the cause is, rather than assuming and declaring it a film to video artifact- which only seems to illustarte our non-professional ignorance on the matter.

My bottom line in making this post is NOT to offer an answer to what is causing this problem (as I'm not nearly qualified).

Rather, I wanted to suggest that it seems possible from my examples that some and possibly much of these problems are caused by something other than a specific artifact from the transfer process and being incorrectly identified by the term "edge enhancement". By using an accepted and defined industry term ("enhancement") it makes it difficult to get past the simple answer that "enhancement" was not applied.

I think we need to move away from using this specific term in hopes that complaints on the matter can't be easily dismissed as consumer ignorance.

-Vince

 

[Jeff Ulmer]

I'm very interested in getting to the cause of this glaring problem. When a man wearing black pants stands against a neutral grey wall there should be no white line separating the edges, yet all too often there is. I suspect part of the problem is that, while they are using expensive, color calibrated monitors, the size of the image most of the work is done on nowhere approaches what many of us see in our homes. I would also like to get to the bottom of "floaters" where portions of the image become dissociated from their backgrounds, which I believe is an artifact of DVNR. With better and better end user displays, these kinds of unnecessary defects should be eliminated.

 

[Ken_McAlinden]

I am convinced that ringing artifacts are not inherent to MPEG compression (which Bjoern has capably shown has its own set of artifacts such as "mosquito noise" and the like), but I am also pretty sure that they are an artifact inherent to certain MPEG encoders.
Regards,

 

[Jeff Kleist]

Ken's figured it out, I wish others would

I've had too many independant people on all levels of DVDs tell me that it's MPEG Encoder artifacting. I've witnessed it myself with footage I've shot.

Vince has offered some great evidence as well

Believe it or not, if it bothers you, go watch "Wings of Honneamise" and then watch another feature. You'll notice it a lot less.

 

[Scott_MacD]

Which transfers are compressed with the Sony MPEG2 encoder?
I know Terminator 2 is an example, according to the DVDReview production diary. This title has an nasty amount of EE. And I don't know if it'd be safe to say that Columbia-Tristar uses it. (although Bjoern mentions that EE tends to be applied to 2.35:1 Columbia transfers, with very little, if none applied to 1.85:1 transfers)
Mr. Staddon assures us that there was no EE applied to the Die Hard With a Vengeance SE transfer, and I'm quite willing to believe him. So, this tends to suggest that some encoders internally add edge enhancement before compression.
So, what movies that exhibit EE use Sony's encoder? And please note, I'm not pointing fingers, I want to get to the bottom of this as much as everyone else.

 

[Vince Maskeeper]

See this quote reveals my own ignorance about the process. I would assume "encoding" and "compression" to be the same step.

I always understood it as:

1) Master is struck- physical film is processed through a machine which transfers those images to video... modern era one transfer is struct in High Def format (I assume a digital master @ 1080).

2) The digital signal is transferred into a computer system raw from the digital master struck in step 1.

3) The master is downconverted from the 1080 format to 480 in the computer.

4) The 480 material is processed into a MPEG2 format file, in the process all compression settings for the file (or section) are set... so the change from an uncompressed 480 video signal into a MPEG2 format compressed file happens in essentially one step.

5) Finished material is written to DLT masters for DVD production.

So is this a misunderstanding on my part? I would assume that the master from film is struck in a digital format- and kept that way from that point forth. At no point is a video signal fed into an "enoding machine" which spits out and encoded stream- instead it would be handled as a digital file which has conversions and filters applied to it in the digital format.

-Vince

 

[GregoryM]

Interesting piece. As a consumer, I don't really care about the cause of the ringing on T2 (and it's quite obvious--look at the T101's leather jacket inside Miles Dyson's house; it's like a halo has been added) or other movies, I just want it gone. The above examples make it quite clear that it's possible to encode discs without it. The studios know that it's happening; it's their responsibility to track down the cause and eliminate it. Until they do, they deserve any criticism for the existence of such artifacts (whether caused by enhancement or mpeg encoders or gremlins) that they get.

 

[Sean Oneil]

It's not in the MPEG encoder. Watch disc 2 of the 'Atlantis' CE, and you will see footage from the film (with slight EE identical to the presentation transfer on disc 1) intercut numerous times with beautiful EE-Free interview footage of the people involved in the making of the film.

Disney could have used separate MPEG encoders for the different types of footage on Atlantis' Disc 2, and then intercut all of this footage together and then overdubbed a separate audio track and then ran it all through the MPEG encoder a second time (which would cause even more artifacts). But more than likely they just edited footage from the film transfer and the interviews together with the audio before doing any MPEG encoding. This would save them an extra unnecessary step, and avoid double compression artifacts.

Also, having fiddled with MPEG2 a bit, I have yet to see any kind of artifact introduced by it that resembles EE.

EE is edge ringing caused by over-sharpening.

 

[Vince Maskeeper]

I'm still skeptical that this is an artifact introduced in the film to video transfer portion. I'm not arguing that it's there- nor that it is bad-- rather I think by declaring it "edge enhancement" we allow the complaint to be ignored.
Take that as you will.

 

[Sean Oneil]

Wouldn't it be entirely possible that the clips from the film in the documentary were taken directly from the finished compressed MPEG2 footage for the DVD?

I mean, if you were doing a doc on the film, and you already had finished video files of the movie- why create new ones? If the MPEG compression was already done, and done improperly- it's esily possible that when this footage was reused, the same problem would present itself.

What you suggest is possible, I suppose.. but it would seem to be a lot more work, and the interview footage would still use the same MPEG encoder as the film if the same authoring house handled it, yes?

Also, I think that Atlantis was a direct digital transfer.. so telicine transfer problems would not apply if that is the case. Disney did Toy Story 2, and Dinosaur, which were also direct digital, and EE free.

 

[MichaelAW]

Are there screenshots of the EE examples of The Phantom Menace out there on the net?

 

[Vince Maskeeper]

Sean,
[sean changed his initial post quite a bit, so this text replies to those original points]
I think you misunderstand the process. If the material from the feature would have been in MPEG 2 format ahead of time, there would be no "second pass" through the "enocoder", as the proper format would already be in place.
For example, if I use a system to capture video as MPEG2, I could send it direct to DVD without encoding it again. If I have video as MPEG2, it would simply carry the existing settings over. Even if I were to work in an editing package like Avid, it when exporting to MPEG2, it would simply carry the MPEG2 sourced files over, bit by bit, without "re-encoding" them.
There would be only "one" necessary encoding process per clip, once the footage was in MPEG2 format, it would never need compressed again to be compatible with DLT for DVD (although it could be done again- there would be no reason to do so)... There would never be an issue with double compression artifacts.
http://home.t-online.de/home/bjoern....TPM/TPM_01.htm