Moire and tweed jacket

[James Press]

[Mods--not sure if this is the best forum for this post--please feel free to relocate if more appropriate elsewhere]

We shoot a talkback style of show (with 2 Reds and 2 Canon XHG1s), with about 30 people in a studio. One of the main speakers, who in the last show was the host, wears a tweed jacket. It really suits him and a lot of thought has gone into it.

We are shooting Red so we can archive our footage and intend to use it for documentaries for broadcast and/or cinema release in the coming years; although our immediate use of it is web and DVD.

The Red footage looks gorgeous and we have no moire at all until we downsize the footage to DVD and web size (approx 700x393). We fixed the moire (at the kind suggestion of Jim Hare :hurray by applying a subtle Gaussian blur to a second video track, which was fantastic and very simple.

My question: moving forward, is there any risk/danger to continue using the tweed jacket? We really don't want to have to change it as it's the right clothing for the host. I understand that we could completely eliminate any possible moire issues by not wearing tweed, but we want the tweed if we can. What's the industry consensus on tweed? What happens in Scotland where every bloke and his dog wears tweed; or filming the British Royal family in tweed; or Sherlock Homes productions?

Given there is absolutely no moire in the 4k, 2k, 1080 or 720 versions of the footage, and we can solve the moire in the web/DVD resolutions with the subtle Gaussian blur, are we safe to continue with the tweed jacket?

Any thoughts greatly appreciated.

James

 

[Cail Young]

I would say so. Instead of the blur filter, you may want to experiment with different downsampling filters. How are you producing the DVD version ( and what show is it, I'd love to see it!)

Edit: on Hawke (lots of fine-detail fabric) we used the high quality scaling filter in Compressor to go from 1080 to the rushes size (bigger than SD but smaller than 720p IIRC) as otherwise we had lots of jagged edges everywhere. Took a little longer but it meant no hassles from wardrobe on review.

 

[Dan Hudgins]

Sharp

Sharp

Anytime you have sharp edges you can get aliasing with a resize that skips pixels. those are used because they are easer to write and work faster.

I made a good resample in my CC that uses 100% of the source pixels to resize the 4K RED footage down to SD or smaller because without it the aliasing was too much.

Even so, the blur from the lens used on the RED ONE plus the OLPF may not be enough to stop aliasing and moire problems with SD size images made from 4K frames. There are adjustments in REDCINE etc. for OLPF comp and sharpness, when you are going down in size quite a bit you can set those to make the images softer.

When you have any sharp digital image and resize it an edge can fall on a pixel boundry so one pixel is bright and the one next to it is dim, but if the edge falls half a pixel off you get one bright pixel, one or two dim pixels and then a dark pixel, in other words the edge takes 4 pixels rather than 2. When the edge moves over the raster you see a jump between sharp edges and ones that have half bright pixels.

When the image is not on/off sharp on the edges, when things move in the image the change between how many pixels define the edge does not vary as much there are some part bright pixels on edges all the time. When film was projected for television use it was not possable to get very sharp edges, so the aliasing and moire from soft film images limited the problem somewhat automaticly.

Because TVs now resize images and change the aspect ratio, if you make a super sharp program for HD and it is viewed SD it may not have enough softness to the edges to avoid aliasing and moire issues. The image re-sizer should take care of adding extra blur before resize, but it may not.

You can see aliasing on computer media viewers because the do a quick realtime resize rather than a good resample using 100% of the source image data, so the noise and aliasing may look worse if the media is not displayed using 1:1 pixel display size.

I would avoid subject matter that is known to cause problems, but otherwise you can hope that things down the line do not "bring back" the issues that you try to filter out. Because TVs have their own sharpen and re-compression by cable companies etc. can cause edge effects you can never be sure what is going to happen when fine patterns of high contrast are in the frame.

 

[jimhare]

Hey James, glad you got some results. Probably worth mentioning whether you are going back to the R3D level when downsizing, or from a ProRes 720P master etc.

My belief is the latter but I don't want to confuse things!

I think Cail's suggestion about the scaling filter is a good one. We spoke about this briefly but spent more time on the blur fix.

Dan, do you start at the R3D level with SD in mind, or can this be done after the video master is created, without going through an online?

Cheers,

Jim

 

[D Fuller]

To answer your original question, ditch the tweed jacket. It's going to be trouble for as long as you use it. While you can make it work through special processing, it will cost you far more in time and money to deal with it than it would for even a very expensive new jacket.

 

[James Press]

Thanks everyone for your constructive feedback.

Cail--would love to chat if you have a moment...Where exactly is the scaling filter in Compressor--I couldn't find it in the Filters tab?

Jim--yes good point. I transcoded the r3ds at full debayer in RED Rocket™™, to ProResHQ 720, and it was that material that we successfully applied the Gaussian blur to.

David--it seems the absolutely 100% safe option is to not shoot tweed--but that doesn't quite address the point I'm wondering about, namely, tightly woven fabrics and patterns get filmed every day: buildings; animal fur; tiles; stripes; blinds etc, so the broadcast/film world has to deal with moire, and is the consensus therefore "avoid problematic fabrics/textures if you can, BUT there will be times when you can't, and the way to cope with that is to soften/blur etc the footage, and that's normal operating procedure and it's OK, 'we all do it', sort of thing. We'd really like to keep the tweed if we can...But I get your point--cheers.

I presume the problem of moire is lessening as definition/resolution of TVs improves too.

 

[Dan Hudgins]

Target

Target

Dan, do you start at the R3D level with SD in mind, or can this be done after the video master is created, without going through an online?

In my DI system I built in 12 levels of resolution for each frame so that you can make output frames optimized for each target resolution.

To get the best results it may be needed to make more than one set of frames from the R3D files, its better NOT to blur an image that has been sharpened, rather not to sharpen it in the first place, if you sharpen then blur you get noise and ringing that would not be there.

If you don't have the time to have soft conversions from the R3D for low resolution targets then doing a blur on what you have before downsize it better than nothing. You want to reduce the 4K down to about 0.75K resolution on the 4K pixel count frames before conversion to SD so that you do not pick up noise and other artifacts. So you blur with a radius of about +/-2 to 3 pixels.

In my system you can make a 4K copy of the 4K color corrected frames and blur those, then do a resample pass to make the small size images. Then if you don't need the soft 4K you can archive or delete them. Its better maybe to resample from the 4K so you get more shades of color through pixel averaging, it takes about 9 pixels in a 4K to equal 1 pixel in SD, so you gain a few bits on tonal range.

 

[Luca Immesi]

Try Video Purifier
with a bit of threshold, it has a very good HD to SD scale capability.

 

[James Press]

Thanks Luca! Will check it out. and THANK YOU for always sharing all your generous knowledge re the Color bug etc--very helpful!

 

[D Fuller]

David--it seems the absolutely 100% safe option is to not shoot tweed--but that doesn't quite address the point I'm wondering about, namely, tightly woven fabrics and patterns get filmed every day: buildings; animal fur; tiles; stripes; blinds etc, so the broadcast/film world has to deal with moire, and is the consensus therefore "avoid problematic fabrics/textures if you can, BUT there will be times when you can't, and the way to cope with that is to soften/blur etc the footage, and that's normal operating procedure and it's OK, 'we all do it', sort of thing. We'd really like to keep the tweed if we can...But I get your point--cheers.

I presume the problem of moire is lessening as definition/resolution of TVs improves too.

I do understand your desire to keep the tweed, but my point really is that it's your host, so I presume that he's on camera a lot, and that'll mean you have to deal with any issues a lot. Over here, there's a host on PBS who wears a lot of fine herringbone suits, and in High def they buzz like crazy. It makes me nuts to watch him.
I'm not so sure the problem is going away with increasing definition. There's just so much you can't control once it goes to the broadcast chain -- will it buzz in 1080? in 720? in standard def? Will it make the cable company's mpeg compression produce digital blocking? Will it cause the mpeg to soften the whole image because it runs out it's data rate? And the answer is different for every frame size. You may well look at the thing differently, but for me, it looks like a really expensive and frustrating choice of wardrobe.

 

[Dan Hudgins]

Size in frame and zoom shots and focus

Size in frame and zoom shots and focus

Another issue is that if the pattern is different sizes in the frame, long shot, medium shot, close up you need different amounts of blur to control the issues, so you cannot just make a one pass blur on the whole project but should optimize the blur for each shot.

Then what do you do about a Zoom in or Zoom out shot, do you optimize for what focal length in the shot.

The f/ stop can affect the pick up of the pattern at very small stops like f/32 you my not see it, but at f/5.6 it may be worse, and at f/1.2 it may come and go as the actor moves in range of the focal plane.

 

[Cail Young]

James, the Compressor control is in the Frame Controls tab.

 

[James Press]

I do understand your desire to keep the tweed, but my point really is that it's your host, so I presume that he's on camera a lot, and that'll mean you have to deal with any issues a lot. Over here, there's a host on PBS who wears a lot of fine herringbone suits, and in High def they buzz like crazy. It makes me nuts to watch him.
I'm not so sure the problem is going away with increasing definition. There's just so much you can't control once it goes to the broadcast chain -- will it buzz in 1080? in 720? in standard def? Will it make the cable company's mpeg compression produce digital blocking? Will it cause the mpeg to soften the whole image because it runs out it's data rate? And the answer is different for every frame size. You may well look at the thing differently, but for me, it looks like a really expensive and frustrating choice of wardrobe.

All excellent points David. I will take them to the team and discuss. Thanks very much for your input.

Another issue is that if the pattern is different sizes in the frame, long shot, medium shot, close up you need different amounts of blur to control the issues, so you cannot just make a one pass blur on the whole project but should optimize the blur for each shot.

Then what do you do about a Zoom in or Zoom out shot, do you optimize for what focal length in the shot.

The f/ stop can affect the pick up of the pattern at very small stops like f/32 you my not see it, but at f/5.6 it may be worse, and at f/1.2 it may come and go as the actor moves in range of the focal plane.

Similarly, excellent points Dan. Thanks for your advice. We need to make a decision from a position of knowledge, which you're facilitating.

 

[James Press]

James, the Compressor control is in the Frame Controls tab.

Got it--cheers.

 

[Roger Crouse]

[Mods--not sure if this is the best forum for this post--please feel free to relocate if more appropriate elsewhere]

We shoot a talkback style of show (with 2 Reds and 2 Canon XHG1s), with about 30 people in a studio. One of the main speakers, who in the last show was the host, wears a tweed jacket. It really suits him and a lot of thought has gone into it.

We are shooting Red so we can archive our footage and intend to use it for documentaries for broadcast and/or cinema release in the coming years; although our immediate use of it is web and DVD.

The Red footage looks gorgeous and we have no moire at all until we downsize the footage to DVD and web size (approx 700x393). We fixed the moire (at the kind suggestion of Jim Hare :hurray by applying a subtle Gaussian blur to a second video track, which was fantastic and very simple.

My question: moving forward, is there any risk/danger to continue using the tweed jacket? We really don't want to have to change it as it's the right clothing for the host. I understand that we could completely eliminate any possible moire issues by not wearing tweed, but we want the tweed if we can. What's the industry consensus on tweed? What happens in Scotland where every bloke and his dog wears tweed; or filming the British Royal family in tweed; or Sherlock Homes productions?

Given there is absolutely no moire in the 4k, 2k, 1080 or 720 versions of the footage, and we can solve the moire in the web/DVD resolutions with the subtle Gaussian blur, are we safe to continue with the tweed jacket?

Any thoughts greatly appreciated.

James

I apologize in advance if this is a bad suggestion, I am not as experienced as most of the pros here. But what about setting crop marks for 700 x 393 resolution and shooting so you don't have to scale the end results? I suppose that's going to mess up the archives for later formats.

Good luck.