XL2 image problem

Clive Collier · August 15th, 2004, 05:50 AM

John

The moire was seen in the following ways:

1. In the viewfinder at acquistion

2. Playing out of the camera into a video monitor both s-vhs and composite

3. Moire was seen digitised into Media 100 both firewire and s-vhs and included drop out.

Please understand that we are professionals too and just because every fact hasn't been noted here doesn't mean they haven't been tried.

I'm going to leave this discussion now purely for the fact that until everyone sees the issue at hand with their own eyes, there's no point spending so much time trying to explain it. I was hoping that others had played with the camera and had possibly experienced the same problems we have. Clearly not yet.

Grayson L. Wideman · August 15th, 2004, 05:51 AM

As Barry has said the problem of moiré is inherent to some extent in all CCD cameras and was a problem in earlier tube cameras as well.

Part of the artistry of a DOP is to work within the bounds of the medium at hand to achieve great looking pictures. If you must have all the parameters of the “Film Look” shoot film.

When I worked at NBCTV-NY we had producers that hated video and complained about all the technical stuff they could not ignore. They hated the fact that they could not hold a piece of tape up to a light and see an image. I worked on one edit session for ‘Saturday Night Live’ that wasted 36 hours at $800.00 an hour trying to do a croma key for a skit because the producer would not admit that there was way to much green spilling onto the puppet. We were even using an Ultimatte. For another skit we had to “dirty up” footage from an L1 to make it look bad enough for the look the show wanted.

I don’t see that the XL2 is pretending to be anything. The spec sheet and the brochure say that it is a DV camera. What it does appear to be doing is pushing the limits of what can be achieved with a DV camera just a little further that any thing before it.

John Mercer · August 15th, 2004, 07:12 AM

Clive,

"Please understand that we are professionals too and just because every fact hasn't been noted here doesn't mean they haven't been tried."

I am in no way suggesting you're not, and you might have seen this moire problem any which way, however all you've supplied us with is a composite capture still that has its own inherent problems that obscure the moire - the resolution and artifacts on this one still are so bad that it is almost impossible to judge it.

If this were truly the only image I had seen from the XL2 I would be seriously worried about its quality as, and this is no reflection on your ability, it is one of the worst quality images from a DV camera that I have seen.

I agree as well with Barry - moire is a fact of digital video life under certain conditions, but until you supply us with a still from a the native DV capture we cannot comment on how bad the problem truly is with the XL2.

Best regards,
John.

Antoine Fabi · August 15th, 2004, 03:03 PM

Clive,

it is also possible that some of the artifacts are due to the upsampling process.

Is it possible for you to post an original JPEG 720X576 JPEG (best depth) ?

thanks

Stefan Scherperel · August 15th, 2004, 05:34 PM

Clive, I completely understand your frustration. However, this truely is not a problem with the camera, it is a problem with the current broadcasting standard, both PAL and NTSC. I have a NTSC DVX100a and have been deeling with this for months. In the DVX there are settings for Thin, mid, and Thick vertical detail settings. The thick is meant for broadcast standards and actually ends up reducing the image quality by about 1/3. WHen viewed on a standard definition television, however, the footage looks great. If you really want to look at it, this is not a problem with the camera at all, in that it captures more information than can actually be displayed on a televison, resulting in the moire patterning. I always shoot in thin mode, even if I know that it is only going to be shown on a television. Why, because it gives me more resolution to work with. THIS IS NOT A BAD THING. I have done some tests with the animorphic adapter for a short film we will be shooting next month, and I can't tell you how badly that will have the moire patterning on an NTSC display. Again, I look at this as a good thing not a bad thing. The easiest way to get rid of this is when viewing the clip on a display monitor, I identify the offending clip, and add a slight horizontal blur to the footage, in affect, turning the footage from the thin line resolution to the thick. On a standard defeniton TV you cannot tell the difference. Be happy that the camera has that kind of resolution, it really is not a bad thing. The XL2 looks like a great camera and I have seen some fantastic screenshots. IF you really think about the marketing of this camera, the moire pattern should never be topic. It is being marketed to higher defenition TV (ie 16:9 native resolution) and to film makers, (the more resolution the better).

Antoine Fabi · August 15th, 2004, 09:26 PM

Stefan,
I used a DVX100A for 9 months and with the anamorphic adapter for 3 months and NEVER experimented any bad artifacts like the ones i see on Clive's grab frames, NEVER.
I have used "thick" and "mid" and "thin" for broadcast applications without any problem. no joke.
"thin" v detail setting has the best resolution and i use it for broadcast applications without any problem.
So this IS NOT normal for a good DV camcorder.

So i think that there are only 2 possibilities in this case (XL2 grabs):

1-Or the XL2 have a real big moire problem.
2-Or the upsampling algorithm (converting 720X576 to 1024X576) is very bad.

i just can't imagine that Canon would release a camera with that kind of problem so i think it has to do with the upsampling algorithm that Clive used to post those grabs.

...but if it comes from the camera, then it is not funny at all...

that's why i want to see full quality JPEGs in native 720X576 res.

Stefan Scherperel · August 15th, 2004, 10:30 PM

Antoine
I would have to agree with you that I have never seen a patterning problem like the one that Clive has posted, however, Higher res footage ie adapter and thin settings do produce some strange flickering when shooting certain patterns, ie. bricks fences etc. I have never seen it look like it does as the grab is posted, however, like Clive said, there is a filter applied to it and that cannot truthfully portray the optical pheonomenon that is seen through an interlaced televion. I hope that this really is the case, as I doubt that canon would release a camera that looked that bad on interlaced television.

John Mercer · August 16th, 2004, 05:00 AM

Antoine,

I don't think it's anything to do with the the upsampling algorithm. I have no problems in exporting 720 x 576 to 1024 x 576 - this is normal to get 16:9 export still from anamorphically squeezed PAL - it should not look bad.

The reason this still looks so bad is given by Clive himself - he captured it into Media 100 via composite analogue - when he should have captured it as native DV.

Best regards,
John.

Jay Gladwell · August 16th, 2004, 05:39 AM

Clive, why haven't you posted an image captured in native DV as requested? It certainly would help put some issues to rest.

Jay

Milosz Krzyzaniak · August 16th, 2004, 06:51 AM

Well I guess the moire problem is not connected nor with upsampling of the image into 1024x748 nor with any TV or PAL issue.

I bet it has something to do with deinterlacing method. As interlaced material doesn't have a moire, and "progressive" has I would turn the spot on the fact that there's something wrong in the way in which the camera does "progressive" scan.

Look at the lamp in http://www.showreel.org/XL2/morray.html
The interlaced one is smooth, but the "progressive" one is jerky. The jerkeness is very similar to the look of interlaced footage treated with deinterlacing algorithm which IS NOT the same as true progressive scan of the image.

My bet is XL2 DOES NOT utilize the true progressive scan (as in DVX100), but is rather something like "frame mode" in previous models.

Robin Davies-Rollinson · August 16th, 2004, 07:02 AM

Can I refer you to a frame grab I made in Progressive scan and with Cinegamma?

http://www.simplydv.com/Reviews/cano...2_popup10.html

If there was to be a problem with moire, I would have thought it would show up on the slates on the roof.

However, I have heard from Canon that the beta version of the XL2 which people were playing with was probably only at 85% of the final picture quality that we can expect in the release models ;-)

Robin
SimplyDv.com

Antoine Fabi · August 16th, 2004, 08:30 AM

guys,

...captured via analog with media 100...OK, didn't know...
...and yes, i too believe that the deinterlacing method may produce such results....

John Mercer · August 16th, 2004, 08:33 AM

"My bet is XL2 DOES NOT utilize the true progressive scan (as in DVX100), but is rather something like "frame mode" in previous models."

I think there is little doubt that the XL2 uses true progressive scan - look at the resolution of Robin's shots.

Again it has nothing to do with deinterlacing either - the break up on the lampost and sign is diagonal. It is clear to my eyes that it is simply a low quality composite analogue capture, and that is confirmed by what Clive has said.

It is exactly the sort of artifacts I see on my monitor when the picture is turned to composite rather than component or s-video.

Best regards,
John.

Barry Goyette · August 16th, 2004, 09:42 AM

<<<If there was to be a problem with moire, I would have thought it would show up on the slates on the roof.

The funny thing about moire is that it only shows up a certain pattern frequency/ chip resolution frequencies..lets call it the M-factor...The way to test it with a video camera is to find a suspect pattern and zoom through a range of focal lengths....depending on the pattern, a moire pattern will appear, change, disappear, and possible reappear....the greater problem, is, as clive found...moires will occur at different times depending on the output device...the viewfinder, monitor, or computer screen...because they all have different resolutions (ie screen frequencies).

The thing to remember is that moire is a natural expectation given this type of image capture. The fact that the xl2 resolves at a slightly higher level make it more likely to "see" the offending patterns, and thus will perhaps, in some situations, create a moire pattern in situations where the lower resolution xl1s wouldn't have (and transversly, not produce moire in situations where the xl1s would have). The second thing to remember is that the problem is much worse at the output end, as monitor resolution is typically quite a bit lower than the aquisition resolution....as clive's image shows us....we see it on the monitor version(I still don't know how he created this), but not on the actual image.

Barry

Bill Anderson · August 16th, 2004, 09:45 AM

John, Antione, and all who are concerned with the method of capture. Read Clives comments and you will find that the method of capture is irrelevant because Clive wished to show an "Impression" of what he saw in the camera, and on the video monitor. And all other faults, problem areas are to be ignored because they ARE, as Clive admits, a product of the capture. This does not present us with any greater an oportunity to assess the footage because visual "impressions" or representations are not really telling us much. However, understanding this might help us to stop chasing our own tails.

August 16th, 2004, 06:51 AM	#55
Milosz Krzyzaniak Regular Crew Join Date: Mar 2004 Location: Germany Posts: 64	Possible explanation Well I guess the moire problem is not connected nor with upsampling of the image into 1024x748 nor with any TV or PAL issue. I bet it has something to do with deinterlacing method. As interlaced material doesn't have a moire, and "progressive" has I would turn the spot on the fact that there's something wrong in the way in which the camera does "progressive" scan. Look at the lamp in http://www.showreel.org/XL2/morray.html The interlaced one is smooth, but the "progressive" one is jerky. The jerkeness is very similar to the look of interlaced footage treated with deinterlacing algorithm which IS NOT the same as true progressive scan of the image. My bet is XL2 DOES NOT utilize the true progressive scan (as in DVX100), but is rather something like "frame mode" in previous models.

August 15th, 2004, 05:50 AM	#46
Clive Collier New Boot Join Date: Aug 2004 Posts: 18	John The moire was seen in the following ways: 1. In the viewfinder at acquistion 2. Playing out of the camera into a video monitor both s-vhs and composite 3. Moire was seen digitised into Media 100 both firewire and s-vhs and included drop out. Please understand that we are professionals too and just because every fact hasn't been noted here doesn't mean they haven't been tried. I'm going to leave this discussion now purely for the fact that until everyone sees the issue at hand with their own eyes, there's no point spending so much time trying to explain it. I was hoping that others had played with the camera and had possibly experienced the same problems we have. Clearly not yet.

August 15th, 2004, 05:51 AM	#47
Grayson L. Wideman Regular Crew Join Date: May 2004 Location: Honlululu, Hawaii Posts: 30	As Barry has said the problem of moiré is inherent to some extent in all CCD cameras and was a problem in earlier tube cameras as well. Part of the artistry of a DOP is to work within the bounds of the medium at hand to achieve great looking pictures. If you must have all the parameters of the “Film Look” shoot film. When I worked at NBCTV-NY we had producers that hated video and complained about all the technical stuff they could not ignore. They hated the fact that they could not hold a piece of tape up to a light and see an image. I worked on one edit session for ‘Saturday Night Live’ that wasted 36 hours at $800.00 an hour trying to do a croma key for a skit because the producer would not admit that there was way to much green spilling onto the puppet. We were even using an Ultimatte. For another skit we had to “dirty up” footage from an L1 to make it look bad enough for the look the show wanted. I don’t see that the XL2 is pretending to be anything. The spec sheet and the brochure say that it is a DV camera. What it does appear to be doing is pushing the limits of what can be achieved with a DV camera just a little further that any thing before it.

August 15th, 2004, 07:12 AM	#48
John Mercer Regular Crew Join Date: Jul 2004 Location: Spain Posts: 73	Clive, "Please understand that we are professionals too and just because every fact hasn't been noted here doesn't mean they haven't been tried." I am in no way suggesting you're not, and you might have seen this moire problem any which way, however all you've supplied us with is a composite capture still that has its own inherent problems that obscure the moire - the resolution and artifacts on this one still are so bad that it is almost impossible to judge it. If this were truly the only image I had seen from the XL2 I would be seriously worried about its quality as, and this is no reflection on your ability, it is one of the worst quality images from a DV camera that I have seen. I agree as well with Barry - moire is a fact of digital video life under certain conditions, but until you supply us with a still from a the native DV capture we cannot comment on how bad the problem truly is with the XL2. Best regards, John.

August 15th, 2004, 03:03 PM	#49
Antoine Fabi Regular Crew Join Date: Mar 2004 Location: sherbrooke (Quebec) Canada Posts: 108	Clive, it is also possible that some of the artifacts are due to the upsampling process. Is it possible for you to post an original JPEG 720X576 JPEG (best depth) ? thanks

August 15th, 2004, 05:34 PM	#50
Stefan Scherperel Major Player Join Date: Dec 2003 Location: Seattle WA Posts: 209	Clive, I completely understand your frustration. However, this truely is not a problem with the camera, it is a problem with the current broadcasting standard, both PAL and NTSC. I have a NTSC DVX100a and have been deeling with this for months. In the DVX there are settings for Thin, mid, and Thick vertical detail settings. The thick is meant for broadcast standards and actually ends up reducing the image quality by about 1/3. WHen viewed on a standard definition television, however, the footage looks great. If you really want to look at it, this is not a problem with the camera at all, in that it captures more information than can actually be displayed on a televison, resulting in the moire patterning. I always shoot in thin mode, even if I know that it is only going to be shown on a television. Why, because it gives me more resolution to work with. THIS IS NOT A BAD THING. I have done some tests with the animorphic adapter for a short film we will be shooting next month, and I can't tell you how badly that will have the moire patterning on an NTSC display. Again, I look at this as a good thing not a bad thing. The easiest way to get rid of this is when viewing the clip on a display monitor, I identify the offending clip, and add a slight horizontal blur to the footage, in affect, turning the footage from the thin line resolution to the thick. On a standard defeniton TV you cannot tell the difference. Be happy that the camera has that kind of resolution, it really is not a bad thing. The XL2 looks like a great camera and I have seen some fantastic screenshots. IF you really think about the marketing of this camera, the moire pattern should never be topic. It is being marketed to higher defenition TV (ie 16:9 native resolution) and to film makers, (the more resolution the better).

August 15th, 2004, 09:26 PM	#51
Antoine Fabi Regular Crew Join Date: Mar 2004 Location: sherbrooke (Quebec) Canada Posts: 108	Stefan, I used a DVX100A for 9 months and with the anamorphic adapter for 3 months and NEVER experimented any bad artifacts like the ones i see on Clive's grab frames, NEVER. I have used "thick" and "mid" and "thin" for broadcast applications without any problem. no joke. "thin" v detail setting has the best resolution and i use it for broadcast applications without any problem. So this IS NOT normal for a good DV camcorder. So i think that there are only 2 possibilities in this case (XL2 grabs): 1-Or the XL2 have a real big moire problem. 2-Or the upsampling algorithm (converting 720X576 to 1024X576) is very bad. i just can't imagine that Canon would release a camera with that kind of problem so i think it has to do with the upsampling algorithm that Clive used to post those grabs. ...but if it comes from the camera, then it is not funny at all... that's why i want to see full quality JPEGs in native 720X576 res.

August 15th, 2004, 10:30 PM	#52
Stefan Scherperel Major Player Join Date: Dec 2003 Location: Seattle WA Posts: 209	Antoine I would have to agree with you that I have never seen a patterning problem like the one that Clive has posted, however, Higher res footage ie adapter and thin settings do produce some strange flickering when shooting certain patterns, ie. bricks fences etc. I have never seen it look like it does as the grab is posted, however, like Clive said, there is a filter applied to it and that cannot truthfully portray the optical pheonomenon that is seen through an interlaced televion. I hope that this really is the case, as I doubt that canon would release a camera that looked that bad on interlaced television.

August 16th, 2004, 05:00 AM	#53
John Mercer Regular Crew Join Date: Jul 2004 Location: Spain Posts: 73	Antoine, I don't think it's anything to do with the the upsampling algorithm. I have no problems in exporting 720 x 576 to 1024 x 576 - this is normal to get 16:9 export still from anamorphically squeezed PAL - it should not look bad. The reason this still looks so bad is given by Clive himself - he captured it into Media 100 via composite analogue - when he should have captured it as native DV. Best regards, John.

August 16th, 2004, 05:39 AM	#54
Jay Gladwell Join Date: Jul 2002 Posts: 1,244	Clive, why haven't you posted an image captured in native DV as requested? It certainly would help put some issues to rest. Jay

August 16th, 2004, 07:02 AM	#56
Robin Davies-Rollinson Trustee Join Date: Dec 2002 Location: Gwaelod-y-garth, Cardiff, CYMRU/WALES Posts: 1,215	Can I refer you to a frame grab I made in Progressive scan and with Cinegamma? http://www.simplydv.com/Reviews/cano...2_popup10.html If there was to be a problem with moire, I would have thought it would show up on the slates on the roof. However, I have heard from Canon that the beta version of the XL2 which people were playing with was probably only at 85% of the final picture quality that we can expect in the release models ;-) Robin SimplyDv.com

August 16th, 2004, 08:30 AM	#57
Antoine Fabi Regular Crew Join Date: Mar 2004 Location: sherbrooke (Quebec) Canada Posts: 108	guys, ...captured via analog with media 100...OK, didn't know... ...and yes, i too believe that the deinterlacing method may produce such results....

August 16th, 2004, 08:33 AM	#58
John Mercer Regular Crew Join Date: Jul 2004 Location: Spain Posts: 73	"My bet is XL2 DOES NOT utilize the true progressive scan (as in DVX100), but is rather something like "frame mode" in previous models." I think there is little doubt that the XL2 uses true progressive scan - look at the resolution of Robin's shots. Again it has nothing to do with deinterlacing either - the break up on the lampost and sign is diagonal. It is clear to my eyes that it is simply a low quality composite analogue capture, and that is confirmed by what Clive has said. It is exactly the sort of artifacts I see on my monitor when the picture is turned to composite rather than component or s-video. Best regards, John.

August 16th, 2004, 09:42 AM	#59
Barry Goyette Trustee Join Date: Oct 2001 Location: San Luis Obispo CA Posts: 1,195	<<<If there was to be a problem with moire, I would have thought it would show up on the slates on the roof. The funny thing about moire is that it only shows up a certain pattern frequency/ chip resolution frequencies..lets call it the M-factor...The way to test it with a video camera is to find a suspect pattern and zoom through a range of focal lengths....depending on the pattern, a moire pattern will appear, change, disappear, and possible reappear....the greater problem, is, as clive found...moires will occur at different times depending on the output device...the viewfinder, monitor, or computer screen...because they all have different resolutions (ie screen frequencies). The thing to remember is that moire is a natural expectation given this type of image capture. The fact that the xl2 resolves at a slightly higher level make it more likely to "see" the offending patterns, and thus will perhaps, in some situations, create a moire pattern in situations where the lower resolution xl1s wouldn't have (and transversly, not produce moire in situations where the xl1s would have). The second thing to remember is that the problem is much worse at the output end, as monitor resolution is typically quite a bit lower than the aquisition resolution....as clive's image shows us....we see it on the monitor version(I still don't know how he created this), but not on the actual image. Barry

August 16th, 2004, 09:45 AM	#60
Bill Anderson Regular Crew Join Date: Jul 2004 Location: Ottawa. Canada Posts: 163	John, Antione, and all who are concerned with the method of capture. Read Clives comments and you will find that the method of capture is irrelevant because Clive wished to show an "Impression" of what he saw in the camera, and on the video monitor. And all other faults, problem areas are to be ignored because they ARE, as Clive admits, a product of the capture. This does not present us with any greater an oportunity to assess the footage because visual "impressions" or representations are not really telling us much. However, understanding this might help us to stop chasing our own tails.