XL2 image problem
 

We have had the PAL 25p version of theXL2 for a few days testing for an indepth review in Showreel magazine (www.showreel.org) and have come across an issue which we were keen to see if possible to eliminate.

We have noticed that in 25p mode, there is a very pronouned morray pattern which we can't get rid of. It is visible in the viewfinder as well as in the final image on rooftops, brick walls and even road markings. Seems to become significantly less in 50i mode as you would might expect. We have given this feedback to Canon and have yet to have a response from Japan.

Does this appear in the US 24p and 30p modes?

We will be trying an Arri film lens to eliminate the standard Canon lens as a potential cause as we have yet to make custom or camera setting adjustments to get rid of it.

Any views or feedback would be appreciated.

Find it hard to believe that there would be problems already and it's not even officially released.

This isn't a joke I'm afraid.

Here are some shots to show clarity of image which is very impressive:

http://www.showreel.org/XL2/xl2stills.html

Will post morray pattern in a moment

hi clive
 

Nice to hear from you that you have XL2. Here in this region no one has NTSC version XL2 ,may be by the end of this month it may come out. People supporting canon is not revealing anything,only few of them are coming out. As I observed from this forum that the people who attended the East Expo, clearly explains that Canon Xl2 has not problems. While you are the first guy who revealed the camera. Can you please explain the morray pattern briefly.

I appreciate your post.

vamshi

I see what you mean. I also notice a bit of red bleeding or glow around the trees in the far bottom left of the image. This doesn't look good. As weird as this is I'd guess there is a defect in the unit or something. Something this obvious can't have slipped through could it? Here's the link just in case nobody can find it.

http://www.showreel.org/XL2/morray.html

Also....I see a pattern in the sign that looks bad! Is that in the real sign or is thatcaused by the pronlem you are reporting?

Here is the link to see what I'm saying

http://www.showreel.org/XL2/morray.html

First off, let me explain that this image is to show the patterning as seen on an interlaced monitor.

This could be a fields issue and easily explained. When you shoot in 25p and view on a TFT or CRT, of course, you are able to see the true prgressive scan of what has been filmed and the patterning doesn't appear as prominently. Regardless of this, both in the viewfinder during filming and when viewing on a video monitor or TV, this patterning becomes very evident. Changing field dominance doesn't change what you see on the video monitor. Indeed, setting your software to square pixels still shows this sort of patterning on an interlaced monitor as you would expect.

The question then remains that if this were to be used for broadcast, how would this patterning be rectified? We haven't yet had a chance to try all possibilities but at present, we haven't found a solution to eliminate it.

Something worth trying would be to change the detail setting....or whatever it would called on the new camera. I have seen similar happenings on my DVX100 in "thin" mode when I would film and the siding of my house got in the frame. The close horizontal lines would cause the frames to look jittery or flickery.

I don't know if that could be it but it does look really odd. Did you compress that "morray" image more? Cause compared to the seagull image it seems really grainy and pixelated. At least to my eye.

Visit http://www.showreel.org/XL2/morray.html again as I have put the progressive image on there too.

Difference is that the building was shot on a less sunny day.

We have tried many of the obvious settings. We've tried Vertical Detail, Sharpness, colour matrix, knee, black.. pretty much everything. Still the same effect which has given us the preclusion that it could either be a side effect of shooting on progressive or an issue with the lens.

As i mentioned, we have requested from Canon the exact settings needed to try and minimise this effect if possible. Currently, it appears that ourselves and the BBC are giving feedback which might be included in further revisions prior to product release. We're receiving a film lens tomorrow and will try that a a process of elimination.

Which program modes are you shooting in? I'm interested to see what the "Green Box" Full Auto mode produces, since it represents the camera's defaults and pretty much takes the shooter out of the equation.

You are working through and communicating with Canon UK, correct?

One possibility lies in the fact that you've received a pre-production market test (MT) sample for review. Could be that it's from one of the very first batches and is therefore perhaps 80% or 85% of its potential, and lacking any final tweaks. If there is in fact an issue here, the question would be whether or not it's been corrected for the mass production units which are all still at the factory in Japan.

Finally, have you been able to reproduce this Moire pattern effect by shooting different backgrounds which have a linear texture similar to the brick building in your frame grabs?

Are these samples from a static shot or panned? Not that either would justify this pattern but it might eliminate another variable i.e pan with long focal length etc.
Thank you.

What I'm seeing doesn't look out of line for any digital imaging device. As the frequency of photographed patterns approaches that of the CCD, it is typical to see these types of moire's. (unless the CCD has an anti-moire/aliasing filter). Your statement that the effect you are seeing is only visible (or more so) on a monitor and viewfinder...would indicate that this is really a product of the pattern frequency in relationship to the output device, not so much the camera...when I look at the strait 25p shot, I'm not seeing any appreciable moire in the area you circled.

The patterning in the red sign and railing is another thing. I think part of it is relative to macroblocking in the red (see adam wilt's article on the dvx100). Also it may be somewhat relative to the expansion of the image to 576x1024..it looks like classic stair stepping thats been interpolated.

One of the dreaded demons of progressive imaging is the detail level often causes problems with standard definition sets...as someone noted about the DVX. Moire's and twitter are part of the package. In the pro digital still camera world (backs without anti-moire filters), we learn to adjust our camera positions to avoid moire's that occur (of course we have the ability crop later).

Barry

If you look at the rest of the picture, both the heavy moire and light/no moire one you see all sorts of strange crosshatching and artifacts. It's everywhere - I assume this isn't just the JPG compression?

Now I see from the caption you say it's capture uncompressed. Are you capturing via video out or something, or are you really capturing firewire? While I thought this could be your problem you do mention that you can see this in the viewfinder which is scary.

Aaron

Dang. I'm sure glad you're here, Barry! Thanks for the knowledgeable insight,

hi
 

Hi everybody.

Its not the capture problem. since according to clive the effect is showing in the viewfinder. It clearly expresses some serious internal problem.

vamshi

Aaron..

those smaller artifacts are also pretty typical of digital sensors, most of which get filtered out by the anti-aliasing filters on almost all digital cameras. When I got my Imacon Ixpress last year (instead of the Saab), which doesn't have one, I was quite shocked to see all the artifacts that my canon 10d covered up with its AA filter.

I think the critical question will be whether these artifacts show up in actual footage, not just still frames (they tend to move around)...my guess is that they will not be visible on an SD monitor, but I could be wrong.

Maybe someone knows whether AA filters are typically used on video cameras...and if so, whether canon has chosen not to put one on the xl2.
(why would you put an AA filter on a camera?--simply put.. AA filters blur the image slightly to eliminate patterning artifacts)

Varnshi--this type of moire would be extremely visible on a camera viewfinder, due to its relatively low resolution...here's a test--take the image, the strait 25p one that shows no moire, and open it in photoshop. Use the zoom tool to zoom out step by step. Notice how the moire in the brick area will appear and disappear as the image gets smaller(even though there is no noticeable moire in the full sized image.

Barry

Barry, that's absolutely freaky that you suggest those artifacts are normal. I would hae thought I'd have noticed stuff like that even on my lowly Xm2. Maybe not.

Cheers
Aaron

Aaron

I'm not saying that they are absolutely normal for a video camera image (but all CCD's do produce exactly these types of artifacts)...

However, lets talk about your xm2...is it progressive scan...?...not really, it's frame mode, which as we all know uses a pixel shift strategy that effectively blurs the image slightly (25%) from what a progressive image off the same sensor would do. A better example would be to look at a highly detailed, patterned image off of the dvx100...I'll try to take a look in the next few days....

Regardless...remember this is a video camera...still frames don't mean much, and I'm doubtful that you would be able to see this kind of artifact on a TV....but I could be wrong.

Barry

There seems to be something seriously wrong with this image compared to those PAL ones from the Simply DV review - they looked superb.

I suspect quite apart from the moire patterns there is something wrong in the way this image has been prepared as an exported jpg still - the resolution looks awful, not at all like the other stills and grabs I've seen so far.

Or if the still has been prepared/exported correctly then I would think there is definitely something wrong with the camera.

Hi Everyone and thanks for your comments.

First of all, yes the patterning is visible on playback both digitised and playing straight out of the camera straight into a video monitor. As I mentioned before, going through the frames on a TFT screen doesn't show the pattern as we are seeing the true progressive scan.

All these frame were taken from hand held moving footage without a tripod.

Suggesting using the auto mode to shoot is something worth trying although I certainly wouldn't expect to be forced to sacrifice making manual setting controls which in essence are at the heart of true detailed progressive work for DOP's.

I would have also thought that offering such potential to get such lovely sharp images as shown in my seagull shot would seem pointless if you have to downgrade sharpness or blurring the image just to rectify moire.

If the footage was viewed purely in progressive, then the effect would not be there if your workflow involved printing to film. However, I would suspect that most people would ideally like to maintain the look and feel for broadcast which does involve interlacing at some stage. If the implication is that such patterning would appear regardless of settings in the production final model, I think there could be issue. Perhaps the moire in the DVX100 is less pronounced due to its lower pixel res and lower spec lens which in a sense doesn't improve matters.

I certainly think that the camera is still a work in progress as Canon UK have pointed out that further revisions will be made based on our and other peoples findings.

Also, no artifacting was added turning the stills into JPEGs. What you see on the weblinks are identical to the actual raw frame data. Even the resizing in Photoshop doesn't detract as it remains faithful to the 16:9 view on our broadcast monitor.

Clive,

Why is your still of such poor resolution when compared to those of the Simply DV UK PAL review?

http://www.simplydv.co.uk/Reviews/canon_xl2.html

It just doesn't look like we are seeing stills from the same camera.

Best regards,
John

Hi John

Take a look at these stills I put up earlier

http://www.showreel.org/XL2/xl2stills.html

The image of the gull looks much clearer than the images on the simplyDV site. In fact the long shot of the building towards the bottom of the review looks simply awful to me having used the camera. I think the image quality is superb with the exeption of this patterning.

Secondly, there is nothing in any of those shots which would cause a moire pattern. Where i saw the effect was on roof tiles, road markings and lines, contrast coloured bricks on walls and even front grills on cars

I agree, John It would be nice to see an image that hasn't been expanded from the native DV frame. In addition to the moire, and ccd artifacts, some of what I'm seeing appears to be relative to bad upsampling, or interpolation.


Aaron...just did a little checking on the DVX...if I had to guess, I would say that the panasonic does have an AA filter. Some artifacts are visible, but not to the degree we see on the xl2 image (overall the image looks quite a bit softer on the DVX)...of course I'm comparing apples and watermelons here. On some gl2 footage, I saw exactly the same type of artifacts...although to a lesser degree...again I don't have any shots of brickwalls and railings in my files (mostly just naked girls and gun toting thugs!)

Clive--the long shot of the building on the simplydv site was shot on the XM2. I agree that your other shots look great, yet there is something out of whack on the brick wall shot...the micro pattern artifact doesn't match whats going on on your other shots...any thoughts?


Barry

Admittedly, the light conditions were less but flicking through shutter speeds, ND, iris and menusettings didn't have an impact at all.

The only slight possibility here is the act of creating the pseudo progressive scan. The DVX100 is only able to do this by buffering the signal which is the reason why on the initial release, some auto functions such as gain were disabled and the audio had a slight synching problem. If the XL2 is using the same technology or method (I think it is) then there could be some form of issue during the deinterlacing stage prior to dumping to tape.

Clive, what exactly did you do to the image in PS? What lens were you using?

Changed image size from 720 to 1024 pixel width. Having said that, let me assure you that rather than get fixated on the images on the web page, the patterning is there on playback on a video monitor from footage captured via composite at 10bit uncompressed. The image quality is not the issue as shown in the other non-moire examples.

The lens was the standard Flourite x20 Canon lens as seen in the press photos. I have an Arri film lens coming tomorrow with an adaptor to try to eliminate the lens as a potential contributing factor.

Incidentally, we posted a message on the CML website asking for any questions which you would be interested in having answered in the review. Feel free to do the same here.

Showreel is a UK based magazine and the camera we have is a pre-production 25p model but many functions would remain the same. We go to press on tuesday so I'll make sure to check back before then to try and cover any particular issues.
(www.showreel.org)

<<<-- the patterning is there on playback on a video monitor from footage captured via composite at 10bit uncompressed. >>>

just so we're clear...are you talking about the rather course moire which began the thread (circled area), or the pixel level artifacts that were discussed later.

barry

The moire.

Artifacting is not that noticeable unless dealing with harsh light which is pretty normal.

Gotta go guys. Its 1am!

Post any questions here or email me at clive@showreel.org

Thanks

hi
 

to Barry:

Well barry. That depends on the image resolution in photoshop.I know if you take 640 vs 1280 picels, the ratio of the picels differs when zooming.

Alright.Although as per subject pal is 800000 picelled resolution. You have to imagine the quality. not even a dot should appear on the frame. To some extent i dont accept with your answer. I dont know weather i am correct or not .This moire effect may be when recording the ccd picels are not active or because of bit level or may be some adjustments in the settings of the camera. You come up with filters to use. In this regard this is not at all the solution. Let me tell you why , even if you record with filters, that effected picels may be darken. but the problem still persists. After all some people are making films with these cameras. May be if some one who has the xl2 pal version. Then we can know the proper reason, either camera defect or anything else.

may be i am not satisfactory to some people but i expressed what i think.

vamshi

Clive

Moire....

I'll go back to the post I made earlier regarding the moire. This is common, expected, normal and has been around since the earliest days of TV. Remember the old adage about not wearing a tweed jacket to an interview...you'll look like you've got ants crawling all over you...this was, and is moire, and it is cause whenever any input pattern approaches the the same frequency of the capturing matrix, as well as the output device matrix...Watch a little low end reality television and you'll see it all over the place...

the effect is heightened by increases in sharpness at the capture end and can be lessened by turning down camera sharpness, filtering the lens (or chip), or by changing the frequency of the offending pattern (by moving closer or further away from the subject).

Barry

Crying wolf? when will it ever stop? As Barry said, this is pretty commonplace in (raster)scanned images. BFO( beat frequency oscillations) are commonplace. There are simple fixes. not the least of which are changing zoom ratio, slightly. Even if it can't be fixed during the shot, it can readiy be dealt with in post, anti-moire filters. So, what's the big bruhaha? These images are astonishing for their quality. I'm already sold on an XL2.

I recall the first time a moire pattern really jumped out at me as kid and really held my attention. It was some fairly famous news video shot in the basement of the Dallas courthouse just after the Kennedy assassination and IIRC before Ruby was shot. Some reporter or court official is talking to the camera, and he wore these heavy metal frame glasses which produced the most bizarre reflective stroby pattern with every slight movement of his head, in my youth to me they looked like some kind of wierd X-ray specs. What's that? I thought, totally lost (still lost) in the hypnosis of that very bright pattern, which made me ignore the entire context of that moment. If I saw it again today it probably wouldn't be a big deal, but back then it really pulled me out of the moment.

Yeah I see Moire on commercial Movie DVD's. Moire doesn't bother me as I almost expect it in video. The other artifacting I saw was what made me sit up and take notice. The images were FULL of this artifacting.

Thanks Barry for trying out a stuff from your DVX etc.

Aaron

So everyone thinks that moire is acceptable and is commonplace or that it can be fixed by antialiasing filters. Mmmm.

1. If shooting a film, will you or your DOP be happy about always being compromised on zooming just because of moire?

2. Do you feel happy that to eliminate this level of moire would result in filtration effects which will soften the image to quality comparable to the DVX100 potentially worse? If so, what is the point of having a camera capable of such sharp imagery if its unusable?

3. Since the effect is dependant on what you're shooting, do you think its acceptable to have full res sharp images in your film intercut with certain shots which are softer to get rid of moire through post filtration? I don't think so.

I sense a complete contradiction with this camera. This camera is aimed at film makers. Progressive scan, interchangeable lens, yet its held back by video technology which has been around for years and yet somehow, JVC seem to have bypassed it with the HD10?? What surprises me is that judging by the posts here, its being forgiven for not fully delivering what its promising already.

"The image of the gull looks much clearer than the images on the simplyDV site. In fact the long shot of the building towards the bottom of the review looks simply awful to me having used the camera. I think the image quality is superb with the exeption of this patterning."

I have to disagree Clive. Whilst the gull looks superb - quite as good as anything I've ever seen from DV - the moire shot of the buildings has some serious artifacts that cannot be accounted for by the moire patterns alone.

I have imported the stills into Xpress Pro with Mojo and I watched them on a 700 line broadcast monitor with component in. In comparing the building shot to the gull and those from the Simply DV article, there is something very wrong. The windows and the sign have terrible resolution. The moire on the bricks I agree with others is not altogether an unusual phenomenon in all DV cameras - but the line break up in the lamp and sign seem to be interference of some kind other than moire.

I realise that you were probably hand held for this shot - but the resolution is so far away from the gull shot that I feel there must have been something else going on to have such a poor result.

Best regards,
John.

Clive: I would like to know if you can record an analog signal
through the XL2 over firewire to your computer without first
going to tape (as with the XL1S). Since the GL2 has this feature.
Would be great if you could add that test to your review.

Also, I'm seeing a dotted line on the top of your images. I'm
assuming this is coming from the camera as well? Can't remember
seeing it in any other images I've seen.

<<<<<1. If shooting a film, will you or your DOP be happy about always being compromised on zooming just because of moire?

clive- if I was shooting film...I wouldn't have this problem..If I was shooting video...any camera will produce this effect given the conditions I described earlier

>>>>>>2. Do you feel happy that to eliminate this level of moire would result in filtration effects which will soften the image to quality comparable to the DVX100 potentially worse? If so, what is the point of having a camera capable of such sharp imagery if its unusable?

Unusable?...all digital cameras produce moire of one type or another give the criteria I mentioned above...does this make these cameras unusable? You have shown us three photos, one of which shows moire only on your monitor--the moire doesn't exist in the actual image--does this make the xl2's sharpness unusable..what about the seagull, the carousel? My hasselblad/Imacon ixpress 132C 22 megapixal produces some of the most amazing moire I've ever seen (when I shoot a moire chart!), and I do have to be careful with certain fabrics...but does that make it unusable on my commercial advertising, portraiture, fashion, architecture projects etc? (my clients apparently don't think so...nor do I)

>>>>>>3. Since the effect is dependant on what you're shooting, do you think its acceptable to have full res sharp images in your film intercut with certain shots which are softer to get rid of moire through post filtration? I don't think so.

Again...ALL video cameras will exhibit this issue in some way shape or form given the conditions stated above. Our jobs as seasoned professional videographers is to recognize the inherant problems of the MEDIUM, and take actions to correct or avoid them in the first place.

Barry

I'm no expert but I would like to see this same set up locked off on a tripod it might eliminate a few things like handheld shake or oscilation, image stabilizer etc. The red brick in the 25p is too degraded to be moire alone.
Looks like a slightly off vertical movement, blur.
However, THANKS TO CLIVE for taking the time to report this, whatever it is deemed to be.

I don't know how it was shot, bill, but if I had to guess, it looks like the in-camera sharpness was turned way up, and the image was then expanded from the DV dimensions to the 576x1024, and that there was possibly some gain during shooting as well...the image quality of the building just doesn't match any other image I've seen off the camera, including clives other shots...so something is not right.

Barry