View Full Version : Aliasing and Moire in fine details question(s)


Lou Bruno
March 1st, 2011, 08:40 AM
I have noticed, at times, fine lines such as telephone wires tend to shimmer. Also, moire lines appear on patterned shirts and even leaves on hedges and bushes. Like the "old days" with a proc/enhancer amp.

I was able to correct this for the most part by using a -7 on the sharpness and bringing up the coring as well as lowering the sharpness on the HDTV. However, I have not observed this on any of my other cameras which are AVCHD and HDV. I have observed the above with the IAF off, no auto-focus and even with the stabilizer off as to eliminate these as potential issues.

My question is this:

Has anyone else observed this and what was your solution?

NOTE: It is more present when rendering to AVCHD Blu-ray but is present in the original material.

Could it be the MPEG-2 compression?

Does anyone have a menu setting for the Vertical and Horizontal detail?

Very little is actually written in the instuction manual/book. I found some information in the Canon 5D threads and am even wondering if this oversharpening is lens related as well.

Charles W. Hull
March 1st, 2011, 11:51 AM
Lou, I am just starting to look at aliasing with the XF100. I have been using a 5DMkII where aliasing is a big issue, so I would like to see it go away with this new camera. I shot a powerline scene that has shown significant aliasing with the 5D, and it is much better with the XF100. And the XF100 shows more detail as well; in one case it resolved two power lines where the 5D only showed one fat line. But there is some effect with the XF100, perhaps what you are calling shimmer. The XF300/305 undoubtedly resolves even better than the XF100, so it could be more pronounced with that camera.

But my tentative conclusion is that I am seeing simple pixelation, and as I scan the camera the pixel edges shuffle. In this case it shows with pixel peeping, but it does not show at a normal HD viewing distance. If you view the video at a normal distance, something like 2.5 x the diagonal screen width, do you still see shimmering? Chuck

Philip Lipetz
March 1st, 2011, 12:24 PM
With the XF100 I experimenting with reducing the red channel to lower shimmer. In cam display it works.

Alan Robert's XF300 BBC settings have a reduced red channel, but red is much stronger in the XF300. Will let you know if I can then properly compensate in post.

Glen Vandermolen
March 1st, 2011, 02:43 PM
Lou,
Are you referring to the XF300 or XF100?

Chris Hurd
March 1st, 2011, 03:13 PM
I suppose I should add a thread prefix selector to delineate topics pertaining to the XF300 series vs. the XF100 series.

Lou Bruno
March 1st, 2011, 03:34 PM
Yes...the XF-300. Both vertical and horizontal lines actually shimmer. Looks somewhat like what happens in downconversions, but not related to that process.

I am not the only person observing this anomaly. I was contacted via e-mail from another user observing the same thing.


Chris,: May be a good idea to differentiate the cameras via another topic. There will be many questions relating to the less expensive XF-100, though they are tied-in very much.

Lou Bruno
March 1st, 2011, 03:43 PM
Yes....I adjusted the red channel as per the BBC white paper on the XF-300 This is more related to what may be resolving issues-too sharp. The secret must be an adjustment encompassing the sharpen menu, coring menu. This would also include in the sharpen menu an adjustment to vertical and horizontal detail. The manual is pretty weak. I may have to contact CANON PROFESSIONAL shortly-this week.

NOTE: This shimmer does not appear om my professional JVC HD monitor, on a computer screen or in the actual display. It appears on a HDTV. Never appeared before on my other HD cameras.
I am good with this "stuff" and this has stumped me. The fact that it is being displayed on other owners HDTV's may help to isolate the problem and come to a plausable solution.

So far -7 sharpness in the menu BBC Horiz/Vert set-up, and increased Coring I also lowered the sharpening on my HDTV and all is passable. Trying to find a good default setting to avoid shimmering.

Lastly, the power-line example is a true example of the shimmering.

With the XF100 I experimenting with reducing the red channel to lower shimmer. In cam display it works.

Alan Robert's XF300 BBC settings have a reduced red channel, but red is much stronger in the XF300. Will let you know if I can then properly compensate in post.

Dave Partington
March 3rd, 2011, 04:52 PM
Yes....I adjusted the red channel as per the BBC white paper on the XF-300

I'm about to start testing the XF300, where can I find this white paper? No point re-inventing the wheel!

Glen Vandermolen
March 3rd, 2011, 04:59 PM
I'm about to start testing the XF300, where can I find this white paper? No point re-inventing the wheel!

http://thebrownings.name/WHP034/pdf/WHP034-ADD50_rev1_Canon_XF300-305.pdf

Dave Partington
March 3rd, 2011, 05:03 PM
Awesome - thanks!

Antony Michael Wilson
March 5th, 2011, 07:10 AM
Lou, this is definitely something to look into on this camera, I agree. It is so sharp out of the box AND with the BBC settings that you will often get aliasing on very fine detail, which will only get worse on SD deliverables. The BBC tests were done on static charts, as far as I know. While they are a great starting point, it seems that - when pictures start to move (!) - you will have problems, particularly on a baseband 1080/50i or 1080/60i output. You get interline twitter, which is what you are probably seeing. As you get further down the post-production chain (such as encoding for Blu-Ray in your case) what is actually very, very minimal aliasing will be exacerbated. I'm running a hardware down-convert from XF material shot with the BBC settings right now. I'm using a Teranex and even with all sharpening/enhancement disabled, I'm still getting exacerbated aliasing. The aliasing is certainly visible on our HD production monitor when you look carefully but could never have been spotted on the shoot. It didn't show up when we tested with charts to start with and this is a great example of how important it is to test/calibrate with charts and scopes AND by actually shooting something!!!! I'm very glad we remembered that before unleashing this camera on a big job!!

I'm confident this has nothing to do with the codec. We're seeing the same phenomenon going into the Avid and FCP uncompressed. Most existing AVC and HDV cameras just don't have the optics and/or the sensor specs to resolve this much detail. The issue here is that the lens and the sensor on the XF are pin-sharp. By disabling all sharpening that the menus will allow, it will be possible to get a better idea if the OLPF is doing its job well.

I have a 305 on test at the moment and I'm having the same problem. If I can find the time and a confirmed solution, I'll report back. A combination of sharpening reduction/elimination and noise reduction should do the trick.

FWIW, I was on an Alexa test a couple of weeks ago and it has the same problem on the 1080 output/internal ProRes recording. It will be interesting to compare to the ArriRaw footage...

Lou Bruno
March 5th, 2011, 03:29 PM
Antony,

Thank you for your comprehensive answer. I must agree with your findings. I did not think it was a CODEC issue.

I have lowered the sharpness and will contact CANON technical help as it relates to some of the other settings that will, hopefully, rectify this situation. Ex: Horizontal and Vert. detail adjustment.

I have been involved in video since 1982. This has to be the sharpest and crisp picture I have observed on a camera of this level. Too sharp. My AVCHD and HDV camera do not exhibit this anomaly.

Matt Ford
March 14th, 2011, 12:22 PM
Im shooting and reviewing footage in the field, no access to a decent screen.

BBC V CP. 25p. Sharpness -3. Looks great on the camera screen. Appalling on my Macbook Pro in XF Utility, shimmering, moire the lot! Hope its just the Macbook screen, if so why?

What coring levels are people using? Has anyone discovered the best CP settings to minimise this?

The XF is meant to be much better then the HD SLR, they must really suck!



Matt

Lou Bruno
March 14th, 2011, 06:59 PM
Here we go MATT:


Sharpness Level: -7 or even -9

H Detail 8

HV Detail 2

Coring Level 5

Lower the sharpness on a HDTV as well.

Good luck.

Matt Ford
March 15th, 2011, 12:15 AM
Thanks for sharing Lou, will give it a try tomorrow!

Matt

Antony Michael Wilson
March 15th, 2011, 08:39 AM
Hi Matt

What raster are you viewing at in the XF Utility? If it's not precisely 1920x1080, XF Utility is itself scaling the image and it's not going to do a good job!. It will introduce its own aliasing. XF Utility is in no way an appropriate way to judge aliasing. You need to view on a proper video monitor. A calibrated professional native 1080 display is best but a domestic 1080 TV will be a good start.

Lou, I've tested a lot on various monitors and with various charts and subjects. The bottom line is that you cannot entirely remove aliasing for a 1080i baseband viewing where there is motion on very fine detail. Detail at -7 or lower gets you close enough for most circumstances, though. The BBC settings are fine for a static zone plate test pattern but real life filming requires detail to be set at -7 or lower when shooting progressive. You don't need to go much lower than -3 shooting interlaced.

These cameras are very sharp indeed, so you do need to be extra careful with this issue, particularly if you need to keep SD deliverables in mind. I've had best results with SD down-converts keeping detail on acquisition at about -7 and then using AJA or Teranex hardware (not default settings) for the down-convert.

Lou Bruno
March 15th, 2011, 10:48 AM
Thank you Mr. Wilson. I have noticed that even at -3 with an interlaced setting, fine details, such as tree branches produce the interline twitter. There is minimal, if any movement in the branches. Same goes for any striped patterns such as fabrics and even bricks. To address this issue, I have lowered the sharpening a bit on my HDTV as well as the -7 sharpening setting.


I guess we all agree that this is some sharp camera!

Antony Michael Wilson
March 15th, 2011, 11:31 AM
To be honest, Lou, I have been doing nearly all of my testing shooting at 25P, so I defer to your experience shooting interlaced. My quick interlaced shooting test was nowhere near as thorough as progressive.

Of course, the aliasing/interline twitter I'm talking about is via a baseband 1080i output whether shot progressive or interlaced.

Bill Weaver
March 15th, 2011, 11:35 AM
Here we go MATT:


Sharpness Level: -7 or even -9

H Detail 8

HV Detail 2

Coring Level 5

Lower the sharpness on a HDTV as well.

Good luck.

worth a try -- but by doing this, are we subtracting from some of the image quality that attracted us to the camera in the first place?

Bill Weaver
March 15th, 2011, 02:55 PM
Another thought --

I've been doing some test shooting, shoulder-held (so as to accent any potential aliasing problem), in 60P.

I haven't tried rendering to SD yet, but on playback on 720 HD TV and on 1080 res apple monitor, the is the most pleasing yet, with less aliasing, moire, etc.

Isn't this an alternative to cranking down the sharpness in the higher settings? HAs a nice, smooth, buttery look.

Lou Bruno
March 15th, 2011, 06:28 PM
I never tried SD. I don't see the aliasing in HD that much defined on my PRO JVC HD studio monitor or even a VGA monitor. However, once rendered to a HD Blu-ray, I do see this sharpness on my HDTV. Perhaps it does smooth out due to less resolution in SD though.

On another note: This camera is so sharp that the -7 setting still produces a sharper image than my HVR-270 with the sharpen menu up to a +3.

Antony Michael Wilson
March 16th, 2011, 04:14 AM
worth a try -- but by doing this, are we subtracting from some of the image quality that attracted us to the camera in the first place?

No, not really. The sharpness of the XF300/305 out of the box is artificially enhanced (as is the case with nearly all cameras). Even with detail set to min (-10) you're still getting a very sharp, true HD image, as Alan Roberts says in his reports:

'Fig.1 shows a single quadrant of one pattern; for this exposure, the camera detail enhancement was turned down to minimum level (-10) which presumably means no correction, so this is probably the native performance of the camera. There are clearly no null zones, where the wanted lower frequencies mix with aliases produced by spectral folding of the unwanted higher frequencies, alias products. This is good evidence that the camera has 3 sensors of full 1920x1080 resolution, and that a “quarter-wave” filter (bi-refringent crystal or other fabricated filter) has been included in the optical path, and is well suited to the camera’s resolution.
There was no evidence of aliasing caused by out-of- band frequencies in the coloured zone-plate patterns. Usable resolution up to about 900 lines vertically and 1800 horizontally is clear. Also, there was no evidence through aliasing that “precision offset” (the spatial offsetting of the green sensor from red and blue by exactly 1⁄2 pixel spacing to improve luma resolution) is used in this camera.
It is rare to see resolution so clearly as this, which is very encouraging.'

I would suggest that it is a good idea to set detail to min/off and dial it back in to taste.

Jerry Laurence
March 16th, 2011, 06:30 AM
I am awaiting delivery of my XF-300 and therefore extremely interested in these posts. Looking at Alan Robert's report, as well as all these posts, it seems that much could be done just by adding more noise reduction, which reduces sharpness to acceptable levels anyway. Wouldn't this control alone reduce unwanted aliasing and noise at the same time?

Antony Michael Wilson
March 16th, 2011, 06:40 AM
Yes, NR will reduce aliasing but it's a bit of a blunt instrument, to be honest. We tested all this quite thoroughly and in order to reduce aliasing in fine detail with NR resolution takes a big hit. It's much better to reduce sharpening since the underlying (un-sharpened) image is plenty sharp enough - if that makes any sense!

It's also important to remember that NR doesn't just reduce effective resolution, it also introduces many other artefacts, so is best avoided where possible. Careful, gentle approach is necessary with all forms of DNR - even the best algorithms. Certain other cameras suffer because NR cannot be reduced or disabled. The XF300/305 is a well engineered camera, so you can control and disable NR. Also, the automatic NR is pretty good and is on the gentle side, so I would leave it off or on automatic unless there are real issues with noise and/or aliasing on specific occasions/shots.

Matt Ford
March 16th, 2011, 04:02 PM
Hi Matt

What raster are you viewing at in the XF Utility? If it's not precisely 1920x1080, XF Utility is itself scaling the image and it's not going to do a good job!. It will introduce its own aliasing. XF Utility is in no way an appropriate way to judge aliasing. You need to view on a proper video monitor. A calibrated professional native 1080 display is best but a domestic 1080 TV will be a good start.

Lou, I've tested a lot on various monitors and with various charts and subjects. The bottom line is that you cannot entirely remove aliasing for a 1080i baseband viewing where there is motion on very fine detail. Detail at -7 or lower gets you close enough for most circumstances, though. The BBC settings are fine for a static zone plate test pattern but real life filming requires detail to be set at -7 or lower when shooting progressive. You don't need to go much lower than -3 shooting interlaced.

These cameras are very sharp indeed, so you do need to be extra careful with this issue, particularly if you need to keep SD deliverables in mind. I've had best results with SD down-converts keeping detail on acquisition at about -7 and then using AJA or Teranex hardware (not default settings) for the down-convert.

Thanks Antony

Screen res on the macbook pro is only 1440 by 900, That could be it!

This is excellent info and Im sure helpful to all.

So it appears that turning down the sharpness to -3 and more is the way to go for real life use.
So what about the other settings? Im trying Bruno's out but will not have access to a decent screen until I return to the UK on the 28th. How are you setting you detail, I note your recommendation to turn it right down, do you mean both? and coring? Would I be right in presuming that this will help with gain noise, smooth it out slightly?

Its good to be approaching the best CP settings.

Matt

Bill Weaver
March 18th, 2011, 07:40 AM
I did a test shoot in a local cemetery (apple blossoms, bare trees, grass, rugged stone, etc) in 30p, with the BBC settings modified further with sharpness to -7 and coring at 5, etc.

Then, I went to DVD with the test footage via Ken Stone's formula (except that I started with native timeline - XDCam HD 8 50mps 422.)

Exporting HDV Video from the Timeline to Standard Definition DVD (http://www.kenstone.net/fcp_homepage/hdv_timeline_to_sd_dvd.html)

Really good results. Wide shots were a bit mushy as one would expect with so much challenging subject matter, but overall the footage was stunning on my 42" 720 HDTV.

This has been a great thread with a lot of useful info.

Would it make sense to start a general XF thread on post production, deliverables, etc, where we could keep editing, grading, etc, issues within our realm? Just a thought. And thanks again everyone, for all the excellent information and perspectives!