What do you think this is? Rolling shutter issue?
 

After shooting around late at night with EX1, felt amazed with the low light situation (the best ever after various affordable cams), I thought this one one of the shot that demonstrates the amazing capabilities of EX1 with 1080/24p over 30 frame and the light sensitivities...., but I found out in the post, there is something wrong when a man with bag crossed in front of the camera. The shape of the taxi behind gets distorted.

Anyone can suggest to make it not happen?

http://web.mac.com/kakuito/KakugyoBl...1_study_1.html

I can't see what you are talking about. Is this the yellow taxi that changes shape? When? What guy with a bag?

Sorry I just don't seee anything. maybe i'm looking in the worng place.

I see what you're talking about, Kaku. It's got nothing to do with rolling shutter. It's actually the lens breathing because it's on auto-focus. The moment the man crosses in front of the camera, you can see the focus plane change from the taxi to the foreground as it searches for something to focus on. This is certainly weird looking, but again has nothing to do with the rolling shutter.

I've watched that clip many times and still haven't seen the artifact you have noted; difficult to precisely identify which is the man-with-the-bag, since there are many. Timecode is a useful thing. I do note that the taxi arriving isn't distorted and that the taxi in the rear moves as someone gets in or out. A stationary object cannot be distorted by a shutter (rolling or otherwise).

Are you speaking of taxi 1125, which bobs as it stops. Wasn't the camera on manual focus?

Ok, just watched it again. The area in question is early in the clip, just before the taxi enters the scene. You can see the focus plane changing at that moment, but it doesn't change as dramatically as I assumed... it does look to be manually focused. It is a strange anomaly, but I'm still quite sure it is unrelated to the rolling shutter.

Theory: the man that passes in front of the camera is, firstly, out of focus - which essentially blurs him. Secondly, he is swinging his arms as he walks in front of the car. The movement in the blurred foreground makes the car look distorted. It's similar to closing one eye, looking at the horizon, holding a pencil up horizontally in front of your eye, and moving it up and down while still focusing on the horizon... the horizon around the pencilo seems to bend and distort.

Just a guess that it was pure circumstance and not shutter-related. Essentially, light diffraction.

Here's a simplistic, but easy to understand example:
http://www.ngsir.netfirms.com/englis...ffraction2.htm

Light waves (or rays) react in the same manner when an obstacle is introduced.

And... one more cheesy example:
http://www.exploratorium.edu/snacks/diffraction.html

Okay, right now I don't have the time to read all of your responses, but this wasn't shot with autofocus. I was playing around with the focus between the people and the taxies. But no autofocus. As matter of the fact, I couldn't figure out to do autofocus until later (didn't know about the ring slide function).

If this thing gets figure out, EX1 should be really excellent.

I will look into more clips to see if any similar thing happened but over all, EX1 shows its ability to shoot video that other cams would have hard time.

And please, don't take it personal people. I'm not attacking EX1 particularly. I understand people don't want to hear about bad things to new star in the industry, but it is good to know how to avoid these things. The issue looks like similar problems I had with HV20 that the picture becomes wobbly, like elastic (excuse my limited expressions here).

I don't know what happened to all of that "you are greatest" and all that encouragements when I provided FX1 and Canon cam clips. I got a person posting "disturbing" on to my blog.
I think I'm going to quit all of this effort now.

Pasty, that's a good one.

Ito,

I think that this effect would happen with any camera, not just the EX1.

I think there is nothing to be concerned about.

No, diffraction isn't relevant at this scale! Certainly not the appropriate phenomenon.

Don't think anyone is upset if there is a problem with the EX, which of course people want to know about. It's just that this doesn't look like a problem.

Even if there's an issue with rolling shutter (I'm not saying there is at this moment then), EX1 seems to overcome many other affordable HD cam issues, so I don't think it will influence the popularity at all.

With my dog clip, you can see some skateboard video shooters would come up with some great footage, even at night at low light situation.

The noise level don't look that much problem when the gain was up and the stabilizer works really well, too. I think that should be helping the compression a great deal.

Kaku...

That EX1, from this first impression, looks very good in low light.

Right now I'm still shooting with an HVX-200 but was looking at the EX1 as a second camera for the fishing show. In fact, if it does well enough, it will probably be the primary camera since it records a lot more material per gigabyte at 1080p30 than the HVX.

BTW, I'm working with Audy to do a bunch of production work, and he is also joining the company that produces the fishing show. If we're lucky we can get a chance to look at a demo EX1 from Sony.

At the risk of sounding stupid. It looks to me that the camera is moving to the left simultaneously as the man passes in front and this makes the taxi look like its skewing because it stays in the frame longer then one would expect if the camera would have been stationary. If the man had not passed the camera at this point I doubt there would be anything funny looking in the taxi.

The footage looks really nice!! Would love to see more!

Sami

my bet is on the mpeg compression... that element on the frame may be interpolated.

Those are my feelings as well. There's a limit to the number of conclusions you can draw from a video clip posted on the web, but this clip does seem very clean, noise free, and crisp, and better than I'd have expected from any previous camera in this class.

I can't see any motion related problems that aren't caused by camera movement and/or focus shifting. But maybe it's because of looking in the wrong place?

Diffraction!
 

Hello Kaku

I guess that your problem is the geometric distortion of the number 87-39, is it right?

If my guess is right, I just succeeded to reproduce such a geometric distortion with my EX1, and also with my XH A1 too. It can be explained by the diffraction, as Pasty already explained. My experiments are as follows: The focal length is abount 10m, and an obstacle very near the aperture. If the obstacle moves, the distortion changes, but for stationary obstacle, the distortion freezes. Thanks.

It's diffraction.

Diffraction
 

Took me a few replays to spot the 'distortion' (easier when dragging the head frame-by-frame), but have to say I third (fourth?) the diffraction opinion... looks to me to be just plain ol' laws of physics/optics at work.

Can't blame a camera for behaving like that ;0)

Ah! I see what you're getting at. I'm not sure diffraction is quite the right term though. If you frame by frame it with a horizontal edge held by the image of the plate on the screen, the shape of the plate doesn't really change - what does happen is that it gets differentially shaded by the out of focus man, which I agree at normal speed does seem to give the illusion of distortion.

Quite easy to simulate without a camera - look at some writing about 3 feet away with one eye closed, and move your finger back and forth past the opened eye very close. Not a rolling shutter effect, just optics!

And under those lighting conditions, the pictures still look very, very good......

Oh, David, I can see distortion of several pixels! it really moves and is distorted. You can see the movement (or distortion : depends on the obstacle's shape) by your 'eye emulation' also.

Glad you got the "eye emulator" to work! What I think is happening is that the finger is is obstructing different parts of the iris with respect to different parts of the image over a narrow band. Over that band the effective f stop varies from max (not obstructed) to very small (nearly all obscured), and if you deliberately defocus the eye, you can see the sharpness of the writing vary over that band - the depth of field is effectively being varied over the relevant area. This is very clearly happening if you go between three frames on Kakus footage - where the "39" first appears it's sharp (and partially obstructed) but the "87" isn't, then in the next frame it becomes equally unsharp compared to "87" as the person moves by.

The other thing that is happening is that the effective centre of the iris moves as it is obscured from one direction, which may well shift the image slightly.

This effect would be expected to be most prominent at large apertures - and in this night scene (which doesn't look like any gain was used) a large aperture is likely to be exactly what was used.

It can all be predicted with assuming light rays travel in straight lines, which is why I don't think "diffraction" is the right term to use to explain it. Diffraction applies to effects which can only be explained by thinking of light as a wave motion.

yes as many people have mentioned, this is definitely not a rolling shutter or compression artifact. this would happen with any lens on any camera. i see this happen with my eyes all the time. anyone who has pretended to squish other people's heads with their fingers from a distance knows this. objects deform and wrap around your out of focus finger just as it begins to cover up the object from being visible to your eye.

too bad this camera has such shallow depth of field. if only they would come out with a deep depth of field camera so we wouldnt have to deal with issues like this. and that high sensitivity! what are we supposed to do with all our powerful lighting equipment! now we have to worry about blown highlights at night, too??

man i want an EX1

It's easy to nit-pick when looking for faults, but conclusions can be taken only from well setup tests. In this case the taxi has just stopped moving (has it?) and we see the stop lights activated a couple of times. People are walking across acting as stops in terms of image intensity, which affects flaring in the image. Light is not bending around people (remember your school diffraction physics about the influence of scale and the wavelength of light). I think, with all due respect, we can hear an echo of that saying "Move along folks, nothing to see here".

Serena, as far as I know, the geometric Optics was generalized by the Fourier Optics which is based on 'diffraction' and Huygens' principle. IMHO, The image distortion discussed here can well be expressed by the Fourier optics, and I'm sure the ray 'looks' bending near the obstacle in this case. Not only the ray bends, the obstacle also acts as some lens in this case. Without the obstacle, the Taxi number 9 is so blurred that we can not recognize it. But when the bag is very close to the number 9 (not physically close, but close to the ray), we can clearly recognize '9'. All these phenomena can be explained by the Fourier Optics, based on the 'diffraction', IMHO.

I'm not sure this helps...

I reckon I'll leave it there; reading on the subject is readily available.

Yes, but I think it's simpler than that, and based on larger geometries than are applicable for diffraction related effects.

The 9 appears sharp in the one frame because the partial obstruction of it by the bag effectively means that only part of the lens is used to image it, so effectively the lens is a smaller aperture for that part of the image. Hence greater depth of field, so it appears sharper (and somewhat darker). Try drawing ray diagrams to represent what happens frame to frame.

But we do all seemed agreed that it's an optical effect, and nothing to do with anything inherent any particular camera.

David,

I totally agree with your final sentence. Further discussion looks meaningless IMO, but I want to say just one thing : Your 'small aperture model' will result in image sharpen over large area, but in actual case, the resolution change occurs in very near vicinity of the obstacle.

I disagree - if we assume an iris diameter of 1 inch, the obscuring object 10 feet away, and the object being imaged 60 feet beyond that, then the affected region of the object will be 6 inches wide. To one side of that the image will be totally obscured, to the other side totally clear - rays from that part of the image will be received by all parts of the lens.

But for that 6 inch strip of the image, rays will only be accepted by part of the lens - effectively a smaller diameter iris and greater depth of field.

Oh, David, thank you for nice explanation. It is interesting. Up to now, I have thought the iris just on to the aperture. The fundamental theory of the Fourier Optics I remenber is the relationship between focal plane and the aperture plane.

But, still, I can't imagine the geometric distortion explained by your small aperture, ray tracing model. IMO, the diffraction will be the key in this issue.

Without a diagram I can't explain it any further, but diffraction effects tend to get worse as the aperture gets smaller, whereas what I'm talking about should get more pronounced as the iris gets larger - and being a night scene I'd expect the iris to be wide open.

If you're still curious, PM me an address I can e-mail a sketch to.

David, My point is somewhat different.. Kaku's clip clearly shows (and I can reproduce with my XH A1 any time,) that the pixels very near the obstacle in the image actually moves.(That made the distortion.) The ray theory can not explain this phenominon IMO.

I think it can if you think of half the lens being blocked off, so it's effective centre is shifted a quarter diameter, with consequent effects on the image.

Another way of thinking of it is to think of what happens to the circle of confusion of an out of focus point source as you obstruct half the lens. One half goes away, so the nett result is the remainder is smaller and it's AVERAGE position is shifted relative to the unobstructed image.

I think the effect is not totally dissimilar to crescent shadows during a solar eclipse, in which case the edge of the moon causes a lensing effect on sunlight.
http://www.astrospace.co.uk/gallery/...-crescents.jpg

The bokeh is essentially distorted back in to (or towards) focus as it bends around the edge of the foreground object, atleast thats how I like to think about it.

Particles... Waves... magic?

;0)
Dave.

What effect exactly are you all looking at? The only thing I see is a brief flash which could be attributed to the taxi turning slightly and having something reflect off its side - a few frames later it's clear the taxi is moving.

David, if you do not consider the diffraction, even if a half(or any amount ) of the aperture blocked, the traced ray falls at the same position in the imaging plaine. And, IMO, The 'circle of confusion' can be explained by the diffraction.

I guess you and I have both very similar concept, but I also guess that you and me are trying to explain the same thing in a different way.

One more thing I wan't to say is that the DOF and the aperture relationship is also based on the diffraction. The 'ray tracing theory' has nothing to say to this relationship.

Although one frame doesn't fully demonstrate the distortion, atleast it helps pin point the mid-point about which you need to scrub the play head to see the warping in action. It looks a bit odd out of context but plays fine as part of the whole scene, the EX1 is doing nothing 'wrong' here.

Hope you don't mind me posting the grab Kaku !?

Thanks Dave - yes, it's the difference between this frame and the one before that best demonstrates what's being referred to. Going back a frame, the unobscured "89" is still out of focus, but the now partly obscured "39" is sharper and darker.

The sharpening effect on the 39 (partial bokeh) seems to me to be caused by the foreground 'man with bag' object as it moves past, almost as a halo effect. I'd call this diffraction = light bending around the edge of an object.

My bad!
 

David and Serena, I'm sorry, I was wrong.

I re-tested it with my cams, and I realized that the image at the focus does not move. Only the images out-of-focus moves with the obstacle moving.

Maybe the 'diffraction model' can precisely describe these, but now I agree with the David's explanation. It's enough to explain all these. Thanks.