A Thousand Words...HV20/Brevis short.
 

Ted Chung and Thomas Colett came up with this amazing short done on the HV20/Brevis combination using our CINEFUSE 1 imaging element. We developed our Pentax K mount for these guys specifically, and boy, was it ever worth it.

Discussion on the piece.

Dennis,

we are developing a 1/2" model of our camera, what are the imaging requirements for using it with Brevis? The optimal focus screen size and focus distance?

Thanks

The rolling shutter if very obvious in some shots.

The story is not exactly original. You might have seen this Patryk Rebisz film shot with a still camera:

http://patrykrebisz.com/stills/FINAL_movie.html

It is not exactly an unknown film. It was on abc news, filmmaker magazine, it received some awards and was screened at many festivals.

http://www.patrykrebisz.com/films_between.html

Are you referring to the banding when she leaves the train carrying the box (37s to 39s) and when he picks up the camera (50s to 53s)? Is that really rolling shutter from the HV-20 or something else freaking out?

If it's rolling shutter, then the HV-20 is a fundamentally flawed tool for serious work involving any kind of panning. And I can only hope other CMOS cameras do a better job.

Is there a way to fix this in post?

Global shutter cmos is as good as CCD so that's the CMOS feature you need for total camera motion freedom. Rolling shutter can be ok if you don't do much vertical camera motion. But you can't run with the HV20 or do fast vertical camera motion or hardmount on a vehicle on a rough road, it will not look shaky like film or ccd, it will look very distorted and unnatural. Shaky footage can be fixed in post but rolling shutter is almost impossible. Deshaker, a virtualdub plugin, has an option for rolling shutter but even with the intended camera model it does not work very well.

I don't think all viewers will notice it, but there are some rolling shutter issues at:

00:12, 00:42, 00:52 (a little)
02:14-02:25 (a lot)

There is nothing obvious at 0:37-0:39, perhaps your player does not double buffer? Step through frame by frame using the quicktime player. It shouldn't have any banding.

CMOS cameras are used in professional environments and are gaining popularity. CMOS and CCD sensors both have limitations and its important to be aware of them and know how to work with them. Just because a camera has a CMOS sensor does not make it a poor tool.

Generally unless filming in uncontrolled circumstances shaky footage, or footage containing distortions should be avoided and reshot. Relying on post work to correct errors is always a bad idea. However, I have read about compensating for shutter distortion in post by using a second, smaller camera.

Deshaker will cause a loss in resolution, and at times extreme. This is especially a problem when shooting 720p since it can easily lower resolution down to SD quality. I have also found that shaky footage is also prone to other problems for example lack of focus due to the shutter. In my opinion it should be only used as a final and last desperate resort.

Ahh you know what that is? I don't think that's rolling shutter actually. That's motion blur from the aftereffects of a shake-correction plugin. I've gotten that kind of artifacting a lot when smoothing out shaky dolly shots with after effects' tracking. What's happening is, after a shot is tracked to cancel out the motion of a shaky shot, the motion blur is still in the shot so even though an object stays in the same location in the frame, the motion blur from the object is still there and is unnatural.

If you ask the creator of this short I can almost guarantee he'll confirm this for you. The artifacting is very characteristic.

You'd have to do some digging to get to the points I'm referring to but on my website, "Broken Minds" has a scene towards the end when a character steps out of a car and the camera does a rotation around him. I added a de-shaker filter to this shot and the artifacts are readily present. Also, in "The Lifeguard" the first "fantasy" sequence (in sepia-tone) contains a lot of this artifacting as it was all done handheld and I decided it was too nauseating to watch without some kind of post-pro stabilizing present.

I tried very hard but could not find the issues you were talking about at 12, 42, or 52.

Also I do remember seeing a while back that stop-motion-esque film you referred to; this is indeed quite curiously similar.

John,

Thanks so much for your reply! It was a problem with QT. When I went through frame by frame (didn't know that left & right arrow could do that) what I saw went away.

I don't know if it's my computer: it's a 3.0 gig P4 with 2 gig of RAM and an Nvidia Quadro FX1100 AGP 8X video card. Problem solved: I moved the file from a 7,200 RPM IDE drive to a 10,000 RPM SATA drive and the banding went away. I belive the first drive just couldn't keep up.

Would it be possible for you to post a "classic example" of rolling shutter from some of your footage? I know it's caused by the image moving too much before the CMOS sensor has had a chance to completely write. I just want to know what I should be looking for, what it acutally looks like.

Thanks.

Peter,

This is cmos rolling shutter. Andrey shot it with the Elphel in a very difficult situation.

http://www.archive.org/download/Roll...rCMOS/cmos.wmv

You can also see the black borders on top and bottom go in and out of the frame sometimes, wrong setting:)

It looks life a fast shutter speed and deshaked quite well. On horizontal panning you get verticals lines to change angle back and forth depending on the direction of shake/camera motion. On vertical camera motion you get compression and decompression of actual object size. This is as extreme as it will get I believe.

You will sometimes see banding if the motion is combined with a refresh of the screen. This is not on the video of course. Stepping through shows it's not there and other player will play it fine. It has to do with drawing on screen while a refresh is being made. Normally you prepare the frame and draw in a single transfer so no refresh happens in the middle of the frame.

Ben,

I'm pretty sure it is rolling shutter. Deshaking will create blur that goes on and off, not distortion. That shot has a little flunctuatiing blur but is also visibly distorted on the vertical axis. Deshaking makes rolling shutter easier to see because it removes any camera motion that is masking it.

Daniel,

How is the second camera used to remove the shutter distortion? I haven't seen that.

Deshaker has parameters that allow to control the ammount of correction. It's not much if you loose 5 or 7% to get extra stability. With a fast shutter it can be artifact free in other aspects. With CCD and global shutter cmos you have an option at least.

If you step frame by frame from 2:20 to 2:22, it is pretty obvious that the house on the far left appears to be made of rubber and the frame is distorted bue to the (probably) diagonal movement.

Ben, in the original thread over at Cinevate, Ted confirms exactly what you had suspected but it was Final Cut's Smoothcam program. You have a sharp eye!

John, our upcoming relay lens will address your camera's needs if you're planning to use the B4 standard.

You can also see rolling shutter problems in this Silicon Imaging clip on the horse riding scene at 2:28-2:35: (2:31- on youtube).

http://www.archive.org/download/SirL...irLateALot.wmv
http://youtube.com/watch?v=H5yD2IutHFU

Better than a Sony HC1 or a Canon HV20 perhaps, but still quite severe vertical distortion.

A C mount lens should be ideal for a relay lens I believe. But what are the actual requirements so we can chose one?

That got me thinking, its been a while. I think I originally was introduced to the idea here:
http://wiki.elphel.com/index.php?tit...n_Compensation

And may have seen it referenced elsewhere. There are other methods used to reduce the rolling shutter effect. I think SI2k shoots at twice the frame rate and drops every other frame to help minimize distortions but I have not had a chance to test this for myself.

Some people don't seem to be bothered by it at all. When I watched Peter Jacksons trailer for "Crossing the Line" shot with RED I noticed a lot of distortions. Others that I played it for didn't mention a thing about it, they thought I was crazy. The newer footage I have watched from RED I have a hard time finding distortions.

The percent required to stabilize a shot will always be relative to how bad the footage is. 7% is kind of suddle, enough for an unstable dolly but In my case I was attempting to stabilize a hand held sequence shot by what seemed like a crippled camera operator. I tried using deshaker, combustion, and eventually steadymove pro. I had very poor results with each and ended up doing it by hand. Even after stabilizing a single shot, the loss in resolution can show as the options come down to cropping the rest of the film or upres the corrected shaky sequence. Nonetheless, software such as deshaker have their uses and are an important asset.

Removing rolling shutter will not be easy. Try to step through the sirlatelot clip and see how complex it actually is. Normal shake with horizontal and vertical distortion on top.

Where would the distortions come from on Red? It's global shutter, isn't it?

EDIT: I thought your link was a generic reference to rolling shutter. So the RED is actually rolling shutter. That's news for me. So, no flash photography wedding shoots with it:) So, the Red will not look very good in a club with a strobe or red carpet scenes or effects with flashes. Sounds like a very serious limitation. There are $200 camcorders that do not have a problem with these shots.

http://www.redhax.net/wiki/Understan...tter_Artifacts

Maybe, but comparing RED to a $200 camcorder is just silly.

I have watched quite a number of shorts and tests shot with the RED and honestly have not noticed any distortions on anything shot on a production model camera. It captures amazing footage and in my opinion is the benchmark for independent, and industry wide cinema (considering the price).

It is nice to see some competition entering and I look forward to see things develop. Competition is the driving force of technology and we all benefit from it.

In my opinion, the distortions in this video are partly from rolling shutter artifacts. I have done digital stabilization on dolly/handheld shots with both ccd and rollingshutter cmos. the skew/stretch/compress distortions are from the rolling shutter, but the blur distortions are from stabilizing bumps shot with a normal 1/48th sec shutter. both types of distortion are noticeable and together even more noticeable, but certainly a nice looking short film otherwise.

distortions throughout crossing the line (peter jackson's RED short)? i saw it in 4k at nab and was watching quite carefully for various artifacts and didnt see much besides the one really noisy shot. ive seen a lot of mention of skew/distortion artifacts but if they were there they werent nearly as noticeable as anything ive seen on the hv20 or si2k. then again maybe people are just more careful with their REDs, and ive never used one. ive only seen skew on the si2k when i was playing with it briefly at nab panning it back and forth really fast.

i dont necessarily think its silly to compare red to cheap camcorders. i mean red is supposed to be no compromises but to think you might be better off shooting a runway fashion show with a $200 camcorder than RED since red will be all half-frames all the time. then again if they support windowed SD resolutions with ridiculously fast framerates or at least the same ridiculously fast pixelrate, maybe then at least people shooting things normally hazardous for rolling shutters would have options. or maybe work in a mechanical shutter option.

I think those who buy Red could make a Canon HV20 look excellent:) They are usually quite experienced, the setups are very professional, lighting is good, etc.

Yeah, personally I noticed the distortions in "Crossing the Line" right away. The skew was there with panning cameras or fast moving objects such as wheels. Did not even have to look for it. Im guessing it was worse in that film because it was shot with a prototype camera. Either people really are more careful or the production model is improved.

What I meant by my comment about comparing RED to a camcorder is that I believe they are built for two completely different reasons, and markets. As far as I know RED is designed for cinema and controlled shooting situations. Furthermore, it is unclear how RED would perform in situations such as the ones mentioned. I have yet to see such an example.

As I mentioned earlier, just because a camera has a CMOS chip does not make it a poor tool. Its just knowing how to use it. Even with the rolling shutter distortions and other limitations of CMOS, carefully shot films with the hv20, RED, V1U, and others look outstanding.