60D RAW video - it's working !!!

Started by marekk, May 24, 2013, 09:27:26 PM

Previous topic - Next topic

0 Members and 3 Guests are viewing this topic.

DerekDock

How easy are these VAFs to take in and out? I'm planning on upgrading my 60D to a 6D in the next couple of weeks and the only Dow side I see to the 6D is the aliasing. I'm not upgrading for RAW but that is always a plus. The moire issues I've been having on the 60D are fine though as I'm used to them in the video I shoot on a dslr. Anything corporate or high end gets shot on a true video camera.  But if the VAF is a simple fix that then makes my "future" 6D a RAW beast it might be worth the $400.
www.vimeo.com/leftcoastdigital
www.twitter.com/derekdock

twillfast

Quote from: Audionut on June 26, 2013, 07:53:58 AM
This is where the VAF filter is useful.  It helps reduce moire/aliasing in the 18MP > 2MP process!

I think QUATRO's point is that it doesn't downscale to 1920x1080, but some esoteric resolution like 1728x526.
However, I'd imagine the filter is supposed to work whether you're shooting 1080, 720, or 526.

a1ex

If you crop an image in Photoshop, does that operation change moire or aliasing or sharpness or whatever?

Same here.

theollybanks

I've been testing Adobe's CinemaDNG Importer for Premier Pro, (Which I explain how to acquire in this post: http://www.magiclantern.fm/forum/index.php?topic=6769.0)

Although I go over in detail some of the advantages and disadvantages of the plugin in the post, my personal view is that for projects that are last minute and need to be done at speed it is excellent coupled with RAWMagic, I experienced no pixels and only a few magenta frames using an hourly build from yesterday (unsure of which) and despite being only 8-Bit (see the post for more on this) I was able to get some really nice grades using Neumann Film's Blockbuster Looks package!



Abstrak

Quote from: theollybanks on June 26, 2013, 03:15:21 PM
I've been testing Adobe's CinemaDNG Importer for Premier Pro, (Which I explain how to acquire in this post: http://www.magiclantern.fm/forum/index.php?topic=6769.0)

Although I go over in detail some of the advantages and disadvantages of the plugin in the post, my personal view is that for projects that are last minute and need to be done at speed it is excellent coupled with RAWMagic, I experienced no pixels and only a few magenta frames using an hourly build from yesterday (unsure of which) and despite being only 8-Bit (see the post for more on this) I was able to get some really nice grades using Neumann Film's Blockbuster Looks package!



Those grades are pretty nice to be presets thanks for sharing I havent followed Neumann Films in so months. They are only a couple bucks to thats not to shabby. I plan to do some test this weekend I'll pick these up.

theollybanks

For the price its phenomenal, you do have to do some considerable adjustments to make them work with the 14-Bit RAW as its perceived dynamic range and the Log to Linear Cineon converter do not play well together, and you often get unusual diamond shaped checkered patterns, easily solved by adjusting the white and black levels in the preset's cineon convertor

cthornhill

POSTSCRIPT - see revised test data below...my original conclusion was somewhat altered on closer inspection - I got a bit more out of the VAF60D than I first thought, but it was not perfect in raw.


I did a very rough test of the 60D with and without the Mosaic VAF-60D Optical Anti-Aliasing Filter. I tried two raw resolutions for this test, 1600 x 670 at 2.39:1, and 1728 x 972 at 16:9. I also shot the same target scene using H.264 with and without the Mosaic filter. For today's quick test I did not have much time or optimal conditions (did not have my lights, etc.). I used ISO 800, and my Canon EF-S 17-55mm f/2.8 IS USM lens at F2.8 and set to 17mm, and room light mixed with daylight. Color balance was set to 5600K. Yes it was a slapped together test! My target was a wool herringbone sport jacket in tan and black. My initial findings were:

1. The jacket showed lots of moire and aliasing without the Mosaic filter using H.264.
2. Using the Mosaic filter and recording in H.264 moire and aliasing were essentially removed - note that at ISO 800 there is some inherent chroma noise, so this is not the best evaluation of fine pattern noise,  but gross artifacts were clearly removed by the filter.
3. At raw 1600 x 670 with no filter, there was VERY pronounced moire and aliasing - much worse than H.264 (which was itself bad).
4. At raw 1728 x 972 with no filter, there was VERY pronounced moire and aliasing - much worse than H.264 (which was itself bad).
5. Using the Mosaic filter at 1600 x 670, there was still moire and aliasing. Subjectively I could not see much difference with or without the filter at this resolution.
6. Using the Mosaic filter at 1728 x 972, there was still some moire and aliasing. Subjectively I think it was a little better, but not good enough to suit me - it would have to be cleaned up to be used in a final output (by some means).

My tentative conclusion is that the Mosaic VAF-60D Optical Anti-Aliasing Filter is so closely optimized for the image size of HD (1920 x 1080) and default H.264 downsampling process on the Canon cameras, that at other resolutions it is not effective, or not as effective. Mosaic has indicated that alternate recording sizes might require unique solutions and I think this is the case.

I will try to post some pictures when I get time (may take a few days) and see what I find in further tests, but for now, I think that the problems of moire and aliasing in raw footage at sizes below full HD are not effectively treated by the filter. I suspect various image processing filters might be able to help reduce the problems, but those may need to be applied as part of the deBayer process to be most effective. This is just early speculation.

Nickbibs

Zeiss Prime Lens Test At B&H Photo. 21mm,35mm and 18mm.

1728x724 @ 24fps

Ungraded footage.


mva

Quote from: cthornhill on June 26, 2013, 11:39:18 PM
My tentative conclusion is that the Mosaic VAF-60D Optical Anti-Aliasing Filter is so closely optimized for the image size of HD (1920 x 1080) and default H.264 downsampling process on the Canon cameras, that at other resolutions it is not effective, or not as effective. Mosaic has indicated that alternate recording sizes might require unique solutions and I think this is the case.

Thanks for the test. It's interesting that forum member kgv5 has achieved some good results for raw with his VAF filter for his 6D. He has made available DNGS with and without the filter, shot at 1808x756: http://www.magiclantern.fm/forum/index.php?topic=6577.msg55134#msg55134 To my eye, it looks like most of the aliasing and moire has been removed by the filter.

I wonder if the different results you and kgv5 got is related to the fact that the product pages for most of Mosaic Engineering's VAF filters (including the 60D filter) describe the filter as providing a "solution for 1080p moiré and aliasing" (e.g., http://www.mosaicengineering.com/products/vaf/60d/main.html) whereas the product page for the VAF-6D just calls it "solution for HD moiré and aliasing," (The product page for the Nikon D800 filter says it's a "solution for 1080p and 720p moiré and aliasing.")

Makes it look like the filters for the various cameras may differ significantly from one another.

QUATRO

Quote from: cthornhill on June 26, 2013, 11:39:18 PM
5. Using the Mosaic filter at 1600 x 670, there was still moire and aliasing. Subjectively I could not see much difference with or without the filter at this resolution.
6. Using the Mosaic filter at 1728 x 972, there was still some moire and aliasing. Subjectively I think it was a little better, but not good enough to suit me - it would have to be cleaned up to be used in a final output (by some means).

My tentative conclusion is that the Mosaic VAF-60D Optical Anti-Aliasing Filter is so closely optimized for the image size of HD (1920 x 1080) and default H.264 downsampling process on the Canon cameras, that at other resolutions it is not effective, or not as effective. Mosaic has indicated that alternate recording sizes might require unique solutions and I think this is the case.

Thanks for the heads up!  :)

P.s.: @Audionut - I TOLD YOU SO!

cthornhill

POSTSCRIPT - see revised test data below - the results are a bit better than first glance showed, but correction was not total for raw. This may be a source of the apparent variable results - some subjects correct easier than others it appears.

Re: Mosaic, I am not trying to start a war! I am a happy user of the product for my 60D in H.264, but my test so far on raw at lower res have not shown the same level of correction. My test was a nightmare for moire - herringbone fabric... :)It would give trouble on most TV cameras (HD/SD) in my experience, and can give trouble to still cameras as well (again in my experience working on product shots). I think the closer to HD res the better the correction, but I have not had time to be more thorough. 6D results might be different.

cthornhill

I did a bit more analysis on my images. I think the filter did more good than I originally thought, but it did not totally correct the issues...this is sort of what I expected. This is just a theory, but the strength of the optical low-pass is tuned for the spatial acuity of the 60D recording in H264. I expect the engineers at Mosaic did not make it any stronger than they needed to, to minimize impact on resolution and overall IQ. Raw is higher res, and may need more 'umph' in a low pass to totally correct aliasing and moire. Anyway this is just a guess on my part. I think it may be less what res you shoot at and more what you get when you shoot - sharper pixels show more artifacts...that is my theory anyway.

I did see quite a bit of correction on my images but not as much as in H264. I put some of them on Dropbox and will post the URL once I test it. Please remember this was a very quick and dirty test in a very short time - it is at ISO 800, and I did nothing to 'make it pretty'. Better exposure (more light) would have allowed a better DOF, and lower noise...but the basic tests is still there.

here is the link (hope this works for all):
https://www.dropbox.com/sh/eq8pqvw8pv3di6m/IzK709GYiz

Revised conclusion - I see more benefit than I thought, but some artifacts remain. Check out the lapel as a good example to see the correction. - I expect to keep the VAF-60 in place and do more tests, but it appears to help some, just maybe not fix every situation. Heck, even the H264 in this case is not perfect and is MUCH lower quality in terms of acuity.

Audionut

Quote from: cthornhill on June 27, 2013, 09:02:29 PM
I did a bit more analysis on my images. I think the filter did more good than I originally thought, but it did not totally correct the issues...this is sort of what I expected. This is just a theory, but the strength of the optical low-pass is tuned for the spatial acuity of the 60D recording in H264.

H.264 goes through a ton of color space changes and resizing.  This kills fine detail.  The same level of aliasing is present in raw video as there is in the initial feed to the H.264 encoder.

H.264 hides the aliasing better because it kills (blurs) fine details.

If they want to "tune" the filter for raw, they would need to blur more detail.

Take this example.
Original raw frame captured by a 6D from this post


And this frame.  The image above simply downsized and resized back to original resolution.



To minimize aliasing caused by line skipping, you have to blur the detail.  This is what the VAF filter is doing.
With the VAF filter not blurring enough detail for raw captures (respecting that the H.264 encoder is going to further blur detail), I would suggest that the next test is to shot a scene that is very slightly out of focus.  You only need to kill the very fine detail that suffers aliasing.

cthornhill

Audionut - yes, I suspect that would be a good test. I spent some time with the engineers at Mosaic via phone discussing what I am seeing and the issues. I myself have done quite a bit of image processing (software for medical and broadcast products in my case,not optical, but of course digital processes mimic optical processes in many cases). One thing they said struck me as very true and important - interpretation of visual results are very subjective...charts, etc. often don't tell you as much as using the camera in real settings...:-)

One thing is very clear - raw provides a huge amount of extra detail over the H264 signal. That is going to come with drawbacks too, since it will show this sort of artifacts as well...every imaging system I know of has to tune these issues out (really they detune the system). What surprises me is how bad moire and aliasing on the 60D in H264...it really is quite bad in many situations - even though the signal (which we now can see is quite sharp in raw) has been hugely degraded in terms of resolution by the compression process.

To me, this is not so much a 'problem with raw', - the 60D has the issues of moire and aliasing and they are uncorrected in the camera as it comes from Canon. While I like Canon, I have to say that I would prefer to buy a camera that came with these video issues corrected out of the box. This is not just an issue for Canon - Sony and other vendors like Black Magic now sell products with real moire and aliasing problems in many scenes...and most of them sell products where these sorts of image artifact issue are better corrected, mostly at higher price points...I do sort of personally find that a bit of an issue, but hey, I can 'vote with my wallet' if it gets to me too much...:-)

The Mosaic filter does address the moire and aliasing issues on the 60D and other cameras with an aftermarket solution. That solution was developed and tested on systems that only offered H264 to test against. The current solution will perform as it was designed, and may or may not be as effective as we hope using raw, but it is just what we have access to now.Now that raw is available, it may be that a solution might be made that has also been tested  and optimized against raw data streams - if that is something Mosaic decides makes business sense.

I want to try and get some shots that will show different situations, and have much better exposure. I would also like to control the scene a bit more, and use some different lenses. I was running my Canon 17-55 wide open, and that is not the best situation (nor is the high ISO). Also, like I indicated, the fabric test I did is a 'worst case' I could easily stage - that fabric would be an issue for most broadcast situations, and it is quite troublesome even on medium format sensors in stills. I would like to try as scene more like what you showed with brick walls and power lines, etc...might be a more typical test.


Audionut

Quote from: cthornhill on June 28, 2013, 03:07:19 AM
One thing they said struck me as very true and important - interpretation of visual results are very subjective...charts, etc. often don't tell you as much as using the camera in real settings...:-)

Indeed.  It's easy to make a filter that hides all the artifacts from a test chart ;)
It's a fine balancing act between hiding artifacts and keeping detail. 

I'd be really interested to see how many artifacts can be minimized with focus, and then how many details can be brought back in post process.  It's generally only the very fine details that suffer from aliasing, so it should only take a very fine adjustment from perfect focus to hide the artifacts in most footage.

cthornhill

Audionut - yeah, the main reason I decided to get the Mosaic for my use was that I often find that on a wide shot I have subjects in the background (typically architecture) that have 'troubles' - especially the dreaded brick buildings...Even at F2.8 (about as low as my wides go) a lot of the background is sharp enough to give trouble in an establishing shot. It just gets worse if I am trying to fly the camera on stabilizer since I don't have a wireless follow focus and have to depend on hyperfocal technique.

Anyway, I will see what happens, and maybe get some shots that are actually worth posting.

mva

Quote from: cthornhill on June 28, 2013, 03:07:19 AM
I want to try and get some shots that will show different situations, and have much better exposure. I would also like to control the scene a bit more, and use some different lenses. I was running my Canon 17-55 wide open, and that is not the best situation (nor is the high ISO). Also, like I indicated, the fabric test I did is a 'worst case' I could easily stage - that fabric would be an issue for most broadcast situations, and it is quite troublesome even on medium format sensors in stills. I would like to try as scene more like what you showed with brick walls and power lines, etc...might be a more typical test.

If you can find the time (and I know how hard that is to do sometimes!), I would certainly appreciate tests of some more typical situations (bricks, fences, roof shingles, etc.). :)  Before shelling out over $300 for a filter I'd want to use primarily for raw on my 60D, I really need to see something of what it can do. (I sent an email to Mosaic Engineering a few days ago asking if they could offer any information on the performance I might expect from the VAF-60D with ML raw, but they haven't replied.) I'm sure lots of other 60D users would be very interested too. I'm not after perfection or success with especially hard cases, just a significant reduction in aliasing and moire in typical cases, as can be clearly seen e.g. in clips made from kgv5's "FILTER ON" and "FILTER OFF" DNGs from his 6D (http://www.magiclantern.fm/forum/index.php?topic=6577.msg55134#msg55134). It would be nice to know if the VAF-60D can produce results with raw that are close to that in similar situations.

cthornhill

It may be a few days or more before I can get a good set of clips,  but I will try and do that...I want to try to get downtown where there are more 'typical' scenes that might create situations like that (building materials and details as well as power lines, etc.)...my schedule and the weather are not too cooperative right now, but I hope to have time over the holiday (in the US anyway) weekend next week. If I can post before then I will.

I can also say that if you can call Mosaic (EST time zone in the US) they may be able to provide more data...I have found them very supportive...They told me they are a little taxed keeping up with all the changes on ML raw (small firm of specialist, lots of work to do to make a living). I think they are pretty responsive, but all this ML raw has happened really fast.

Sorry I haven't more time to do proper tests yet...it is a LOT more fun than some of the meetings I am in these days...I miss making TV for a living some days more than others...:-).

mva

Sounds great. Will look forward to anything you end up posting on the topic, whenever. I thought about calling Mosaic. I may do that. Cheers!

cthornhill

Post to no specific other comment...I just read the article on EOSHD about Skin (and watched the film)...I was telling the guys at Mosaic that using raw on the 60D feel like shooting film...I could not help but notice the comment from the director of Skin (Andrew Wonder) that using the 5DM3 raw reminded him of using his Bolex. I could not agree more - it feels VERY much like shooting on 16mm to me...I don't find it quite as consistent yet but I am sure it will stabilize as the code gets more mature. Overall, the short takes and heavy post process orientation just remind me a lot of small 16mm days...not as refined as an Eclair or Arri for sure (yet)....more like a spring wound Bolex...but those were fun...and made some cool films. The look is capable of much more refinement of course, but I really do 'feel it'...maybe no one else does, but I suspect some other old school (well old anyway) folks may know what I mean...:-)

Haliburton

Quote from: DerekDock on June 26, 2013, 08:21:39 AM
How easy are these VAFs to take in and out? I'm planning on upgrading my 60D to a 6D in the next couple of weeks and the only Dow side I see to the 6D is the aliasing. I'm not upgrading for RAW but that is always a plus. The moire issues I've been having on the 60D are fine though as I'm used to them in the video I shoot on a dslr. Anything corporate or high end gets shot on a true video camera.  But if the VAF is a simple fix that then makes my "future" 6D a RAW beast it might be worth the $400.

DerekDock, the VAF filter looks quite simple to re-and-re in Cinema5D's review
http://vimeo.com/44900401

The big downsides to me of the 6D are yes, the terrible aliasing, but also the card-speed bottleneck which pretty well eliminates any full-frame goodness while shooting raw:
http://www.magiclantern.fm/forum/index.php?topic=6215

$400 to fix the aliasing with a VAF, and require post-sharpening to try to regain some of the lost sharpness, is a big chunk of the price difference between the 6D and the 5D MkIII - and still, we're stuck with slow, SD write speeds.

As attractive as the pricing on the 6D may be, it seems to me that no amount of technical jury-rigging will bring it close to the native full-frame clean-ness of the 5D MkIII.

Though I doubt that I'll buy another crop-sensor Canon, I look forward to seeing if Canon answers their widely-criticized aliasing, with the new sensor on the 70D, slated for introduction Tuesday July 2:
http://www.canonrumors.com/2013/06/canon-eos-70d-spec-list/


DerekDock

I see your points but I'm finding that I don't need the record times afforded by the Cf card cams. With the 6D I can record 1080 2.35 for almost 20 seconds (in theory as I don't have it yet). Which is perfect for me as the data and cost of media gets out of control on the 5D mk3.

For $900 less than the 5D I can have the 6D with mosaic filter and have the same image quality, a nice form factor, and more manageable data. That works for me and gives me $ to spend on glass and creative cloud. I also have unlimited access to two Mark3s owned by my brother and a friend, so if I find I need one its a phone call and an hours drive away.

I have a 60D as well and have been shooting a good amount of RAW on it even with a SD bottleneck that is much worse than on the 6D. I'm sure ill be stoked.

Really now my only concern is getting the 6D to play nice with RAWmagic so my workflow can be smoother. But the VAF purchase is likely coming sooner than later.
www.vimeo.com/leftcoastdigital
www.twitter.com/derekdock

1%

Heh, 6D is 5d3 JR, FW is pretty much the same. But a pretty good camera, esp when people start selling them to get 5D3, etc.

cthornhill

My $.02 worth on the ease of use of the VAF is..it depends. The packaging and tools from Mosaic are nice and help a lot. In principle it is not hard. In practice, you have to have the camera supported, and you are handling a small piece of expensive glass around your open camera body. You lens is off, so it has to be put someplace too, and protected. All not a big deal if you are careful and have time and a place to put things...now try it outdoors...and at in the wind, and in the light rain, etc...you see what I mean. Just as you would not want to change lenses on a windy day at the beach, etc. there are times moving the VAF in and out is not fun. No one's fault, but you need to consider it...the filter costs $400 or close to it...so you need to be careful, and your $1,000 camera is pretty vulnerable with no lens and the mirror up...but the actual insert and removal is quick, so that is at least good.

I just don't take stills with my 60D so it is not an issue for me, but if I did, I would plan ahead a bit, and keep an extra cloth to put things on, and maybe an extra pouch. A bench or table might be your friend in this situation, or at least your camera bag.

Don't be afraid of it, but do realize it is not the most convenient thing in the world...if a 5DM3 is not in the cards, this is not too much trouble to go to, you just need to take care and plan ahead (find a place to 'change the baby' before you need to).