how to avoid video shearing at DVinfo.net

Lisa Lee · October 28th, 2003, 05:28 PM

I'm wondering how slow do we have to pan / zoom etc to avoid the video "shearing" effect. And other types of artifacts introduced with any rapid movement with this camera. Or is this camera only useful for still shots....

Alex Raskin · October 28th, 2003, 06:04 PM

Special shooting technique must be used, as with all progressive cameras (including film), to prevent ghosting images while panning.

Namely, pan the camera very slowly; or track the subject you're focusing attention on (background will still smear).

If your camera is stabilized with gyros (see www.ken-lab.com, for instance), then the gyros will naturally slow down the maximum speed you can safely pan with, making your moves very film-like.

Now, *zooming* has nothing to do with this phenomenon - it normally propagates only while panning.

Gints Klimanis · October 28th, 2003, 06:15 PM

Why wouldn't this happen with zooming? From my limited understanding, For panning, the motion is nearly horizontal. For zooming, the motion is both vertical and horizontal.

Alex Raskin · October 28th, 2003, 07:39 PM

There's no "motion" with zooming. Only lense's focal length is changing. Zoom doesn't make objects move against the camera, thus there should be no visible artifacts like strobing (or ghosting - I still don't know what would be the universally accepted term for the effect).

On contrast, the object crossing the camera's view is changing its position against the camera. Because camera only captures 30 frames per second, each frame registers an object in a different position. The faster the relative movement of object against the camera, the harsher the strobe.

Of course this looks even worse with 24p than with Hd10's 30p.

Please note that Shutter speed also contributes to the perceived severity of the artifact. With 1/30 shutter, the severity is less than with 1/60 or faster.

This is because 1/30 shutter simply makes images blurry, thus "smearing" the ghosting effect to the extent that it becomes less visible.

Steve Mullen's HD10 Shooting Guide (recommended!) has a scientific explanation of the phenomenon.

Gints Klimanis · October 28th, 2003, 08:04 PM

>There's no "motion" with zooming. Only lense's focal length is >changing. Zoom doesn't make objects move against the camera, >thus there should be no visible artifacts like strobing (or >ghosting - I still don't know what would be the universally >accepted term for the effect).

The motion is the subject moving relative to the CCD array, which will happen during panning (horizontal), tilting (vertical) or zooming (horizontal and vertical). You're describing an artifact that seems to worsen with the frame rate (sampling rate). The artifact that is caused by panning too fast can also be caused by zooming too fast. Another area of videography that clearly illustrates both issues is titling, especially noticable on characters with small features and sharp, high contrast edges.

Gints Klimanis · October 28th, 2003, 08:23 PM

Actually, the motion is that of the projected of the image relative to the CCD array. The subject need not move. Likewise, as we've seen to great effect in some movies, one can zoom out and move the camera toward the subject such that the subject doesn't appear to change size/move.

Brad Hawkins · October 28th, 2003, 09:53 PM

Lisa,

Unfortunately, I don't think your going to be able to find an answer to "how slow you need to pan" here in the forum. I think this is one of those things where you just take the camera out for an after noon and practice and then review your results later.

But I really wouldn't spend too much time worrying about the whole thing. Personally I don't think this "ghosting" should be an issue at all. I mean, film cameras have had this effect for over 100 years and no one is complaining.

Brad

Gints Klimanis · October 28th, 2003, 10:22 PM

It's easy provide a formula for a minimum of aliasing for a particular feature size for a particular video frame resolution sampled at a particular frame rate. However, the result will results in pannning/zooming that is too slow.

Steve Mullen · October 29th, 2003, 01:56 PM

The table you want is in my Shooting Guide.

Darren Kelly · October 29th, 2003, 03:09 PM

The higher the shutter speed, the less likely you will notice the motion when you pan.

this is true to a point. If you get too fast, you will loose the reality in the picture.

I would recommend you use the camera at 100th if a second or 125th of a second.

I'm not sure how fast you want to pan, but take a look at TV and film. Not a lot of fast pans unless they use it as an effect (like The Practice for example), in which case, you won't care.

DBK

Steve Mullen · October 29th, 2003, 05:57 PM

<<<-- Originally posted by Darren Kelly : The higher the shutter speed, the less likely you will notice the motion when you pan. -->>>

The higher the shutter-speed the WORSE it looks. You want it below 1/60th.

Shuter-speed control helps, but pan speed is critical.

A super fast pan or a slow pan are both fine.

Darren Kelly · October 29th, 2003, 06:36 PM

1/100th or 1/125 does not look bad Steve, faster than that and I agree, but come one, it looks terrible at 1/30 due to the juddering.

DBK

Charles Papert · October 29th, 2003, 06:39 PM

The nice thing about digital is that you can see the results of strobing in the viewfinder to an extent, and certainly on playback on a monitor. With film, you don't really get a feel from it in the eyepiece--you just develop a knack for what is the "no fly zone" for pans. Be aware that different focal lengths will affect the pan speed.

As Brad points out, this phenomenon has been present for many years in the film world. However, I have seen too many instances of strobing and I find it very annoying when it does happen, especially watching a movie in a cinema; sometimes I have to look away from the screen. I remember seeing "The Nightmare Before Christmas" when it came out, and was astonished that during the opening sequence, there was an extended pan that strobed terribly, pulling me right out of the movie. I would have thought that the obviously expert animators on that movie would have been conscious that they were programming a move that would have had those results, but...!

Just yesterday I was asked to perform a pan across a set simulating a character's point of view; the director was pushing me to get through it faster, but I had to warn him that doing so would risk strobing, and I would have to jump up to double the speed to avoid this. It's a regular occurence.

Alex Raskin · October 29th, 2003, 08:02 PM

Charles, I'm with you.

Last movie I saw in the theaters was Matrix Reloaded.

I also was shocked at how stroby many scenes looked. Exactly as you said - I found myself turning away from screen when the effect became real bad!

Going back to HD10 filming: of course Steve is right that *faster* shutter makes strobing more apparent, while *slower* shutter makes strobing less apparent.

The practical difference is between 1/30 (smears the shot to make strobe less visible) and 1/60 (clear, sharp video but strobing is showing.)

Faster than 1/60 shutter makes no difference in 99% of the cases.

Remember that frame rate is just 30p *regardless* the shutter. Therefore "amount of reality" captured in any given frame is *not* increasing with faster shutter. As a matter of fact, it is decreasing with faster shutter speeds for the obvious reason (thus more strobing).

I did some rather extensive tests on strobing with HD10 camera, and finally decided to go with 1/60, sharp picture, and (unfortunately) some strobing.

Of course, strobing is significantly *worse* with film, which has lower temporal resolution (with equal pixel dimensions) due to the lower (24 vs. 30) frame rate. (Of course film always wins in clarity as it's 2000p to 4000p per each frame...)

So in terms of strobing, we are in a better position with HD10's 30p than film camera operators.

I'd of course wish for 720p60 cameras to arrive... Now that would really increase temporal resolution and lessen amount of strobing, comparing to today's HD10 720p30 - but at expense of DOUBLE the data bandwidth to capture/transfer/edit.

Talk about slow real-time editing with 720p60! :)

So taken post-production into consideration, 720p30 seems to be tops what we can realistically edit/process at this time using high-end PCs.

Steve Mullen · October 29th, 2003, 11:36 PM

<<<-- Originally posted by Alex Raskin :

I did some rather extensive tests on strobing with HD10 camera, and finally decided to go with 1/60, sharp picture, and (unfortunately) some strobing. -->>>

Alex is right. 1/60th is the optimum. Anything higher is horrible. 1/60th is a 180-degree shutter just like film. (Darren, i hope you have this right on your DVD as it is crucial. You have a copy of my Shooting Guide with a detailed explanation of this aspect.)

Of course, 1/30th is necessary when light goes too low.

Alex is also right about p60. Since the HDV spec. limits 720 to 19Mbps, doubling the pre-compression data rate is going to increase compression artifacts terribly. As will 1080i -- which is also 2X the data rate -- even through the spec allows 25Mbps.

2X the data rate also dramatically increase heat and power consumption.

October 28th, 2003, 05:28 PM	#1
Lisa Lee Regular Crew Join Date: Sep 2003 Posts: 86	how to avoid video shearing I'm wondering how slow do we have to pan / zoom etc to avoid the video "shearing" effect. And other types of artifacts introduced with any rapid movement with this camera. Or is this camera only useful for still shots....

October 29th, 2003, 01:56 PM	#9
Steve Mullen HDV Cinema Join Date: Feb 2003 Location: Las Vegas Posts: 4,007	The table you want is in my Shooting Guide. __________________ Switcher's Quick Guide to the Avid Media Composer >>> http://home.mindspring.com/~d-v-c

October 29th, 2003, 03:09 PM	#10
Darren Kelly Major Player Join Date: Aug 2003 Location: Vancouver Posts: 351	The higher the shutter speed, the less likely you will notice the motion when you pan. this is true to a point. If you get too fast, you will loose the reality in the picture. I would recommend you use the camera at 100th if a second or 125th of a second. I'm not sure how fast you want to pan, but take a look at TV and film. Not a lot of fast pans unless they use it as an effect (like The Practice for example), in which case, you won't care. DBK __________________ Darren Kelly

October 29th, 2003, 05:57 PM	#11
Steve Mullen HDV Cinema Join Date: Feb 2003 Location: Las Vegas Posts: 4,007	<<<-- Originally posted by Darren Kelly : The higher the shutter speed, the less likely you will notice the motion when you pan. -->>> The higher the shutter-speed the WORSE it looks. You want it below 1/60th. Shuter-speed control helps, but pan speed is critical. A super fast pan or a slow pan are both fine. __________________ Switcher's Quick Guide to the Avid Media Composer >>> http://home.mindspring.com/~d-v-c

October 29th, 2003, 06:36 PM	#12
Darren Kelly Major Player Join Date: Aug 2003 Location: Vancouver Posts: 351	1/100th or 1/125 does not look bad Steve, faster than that and I agree, but come one, it looks terrible at 1/30 due to the juddering. DBK __________________ Darren Kelly

October 28th, 2003, 06:04 PM	#2
Alex Raskin Trustee Join Date: Sep 2003 Posts: 1,435	Special shooting technique must be used, as with all progressive cameras (including film), to prevent ghosting images while panning. Namely, pan the camera very slowly; or track the subject you're focusing attention on (background will still smear). If your camera is stabilized with gyros (see www.ken-lab.com, for instance), then the gyros will naturally slow down the maximum speed you can safely pan with, making your moves very film-like. Now, zooming has nothing to do with this phenomenon - it normally propagates only while panning.

October 28th, 2003, 06:15 PM	#3
Gints Klimanis Inner Circle Join Date: Jun 2003 Location: San Jose, CA Posts: 2,222	Why wouldn't this happen with zooming? From my limited understanding, For panning, the motion is nearly horizontal. For zooming, the motion is both vertical and horizontal.

October 28th, 2003, 07:39 PM	#4
Alex Raskin Trustee Join Date: Sep 2003 Posts: 1,435	There's no "motion" with zooming. Only lense's focal length is changing. Zoom doesn't make objects move against the camera, thus there should be no visible artifacts like strobing (or ghosting - I still don't know what would be the universally accepted term for the effect). On contrast, the object crossing the camera's view is changing its position against the camera. Because camera only captures 30 frames per second, each frame registers an object in a different position. The faster the relative movement of object against the camera, the harsher the strobe. Of course this looks even worse with 24p than with Hd10's 30p. Please note that Shutter speed also contributes to the perceived severity of the artifact. With 1/30 shutter, the severity is less than with 1/60 or faster. This is because 1/30 shutter simply makes images blurry, thus "smearing" the ghosting effect to the extent that it becomes less visible. Steve Mullen's HD10 Shooting Guide (recommended!) has a scientific explanation of the phenomenon.

October 28th, 2003, 08:04 PM	#5
Gints Klimanis Inner Circle Join Date: Jun 2003 Location: San Jose, CA Posts: 2,222	>There's no "motion" with zooming. Only lense's focal length is >changing. Zoom doesn't make objects move against the camera, >thus there should be no visible artifacts like strobing (or >ghosting - I still don't know what would be the universally >accepted term for the effect). The motion is the subject moving relative to the CCD array, which will happen during panning (horizontal), tilting (vertical) or zooming (horizontal and vertical). You're describing an artifact that seems to worsen with the frame rate (sampling rate). The artifact that is caused by panning too fast can also be caused by zooming too fast. Another area of videography that clearly illustrates both issues is titling, especially noticable on characters with small features and sharp, high contrast edges.

October 28th, 2003, 08:23 PM	#6
Gints Klimanis Inner Circle Join Date: Jun 2003 Location: San Jose, CA Posts: 2,222	Actually, the motion is that of the projected of the image relative to the CCD array. The subject need not move. Likewise, as we've seen to great effect in some movies, one can zoom out and move the camera toward the subject such that the subject doesn't appear to change size/move.

October 28th, 2003, 09:53 PM	#7
Brad Hawkins Regular Crew Join Date: Jul 2003 Location: Edmond, OK Posts: 38	Lisa, Unfortunately, I don't think your going to be able to find an answer to "how slow you need to pan" here in the forum. I think this is one of those things where you just take the camera out for an after noon and practice and then review your results later. But I really wouldn't spend too much time worrying about the whole thing. Personally I don't think this "ghosting" should be an issue at all. I mean, film cameras have had this effect for over 100 years and no one is complaining. Brad

October 28th, 2003, 10:22 PM	#8
Gints Klimanis Inner Circle Join Date: Jun 2003 Location: San Jose, CA Posts: 2,222	It's easy provide a formula for a minimum of aliasing for a particular feature size for a particular video frame resolution sampled at a particular frame rate. However, the result will results in pannning/zooming that is too slow.

October 29th, 2003, 06:39 PM	#13
Charles Papert Wrangler Join Date: Jan 2002 Location: Los Angeles, CA Posts: 6,810	The nice thing about digital is that you can see the results of strobing in the viewfinder to an extent, and certainly on playback on a monitor. With film, you don't really get a feel from it in the eyepiece--you just develop a knack for what is the "no fly zone" for pans. Be aware that different focal lengths will affect the pan speed. As Brad points out, this phenomenon has been present for many years in the film world. However, I have seen too many instances of strobing and I find it very annoying when it does happen, especially watching a movie in a cinema; sometimes I have to look away from the screen. I remember seeing "The Nightmare Before Christmas" when it came out, and was astonished that during the opening sequence, there was an extended pan that strobed terribly, pulling me right out of the movie. I would have thought that the obviously expert animators on that movie would have been conscious that they were programming a move that would have had those results, but...! Just yesterday I was asked to perform a pan across a set simulating a character's point of view; the director was pushing me to get through it faster, but I had to warn him that doing so would risk strobing, and I would have to jump up to double the speed to avoid this. It's a regular occurence. __________________ Charles Papert www.charlespapert.com

October 29th, 2003, 08:02 PM	#14
Alex Raskin Trustee Join Date: Sep 2003 Posts: 1,435	Charles, I'm with you. Last movie I saw in the theaters was Matrix Reloaded. I also was shocked at how stroby many scenes looked. Exactly as you said - I found myself turning away from screen when the effect became real bad! Going back to HD10 filming: of course Steve is right that faster shutter makes strobing more apparent, while slower shutter makes strobing less apparent. The practical difference is between 1/30 (smears the shot to make strobe less visible) and 1/60 (clear, sharp video but strobing is showing.) Faster than 1/60 shutter makes no difference in 99% of the cases. Remember that frame rate is just 30p regardless the shutter. Therefore "amount of reality" captured in any given frame is not increasing with faster shutter. As a matter of fact, it is decreasing with faster shutter speeds for the obvious reason (thus more strobing). I did some rather extensive tests on strobing with HD10 camera, and finally decided to go with 1/60, sharp picture, and (unfortunately) some strobing. Of course, strobing is significantly worse with film, which has lower temporal resolution (with equal pixel dimensions) due to the lower (24 vs. 30) frame rate. (Of course film always wins in clarity as it's 2000p to 4000p per each frame...) So in terms of strobing, we are in a better position with HD10's 30p than film camera operators. I'd of course wish for 720p60 cameras to arrive... Now that would really increase temporal resolution and lessen amount of strobing, comparing to today's HD10 720p30 - but at expense of DOUBLE the data bandwidth to capture/transfer/edit. Talk about slow real-time editing with 720p60! :) So taken post-production into consideration, 720p30 seems to be tops what we can realistically edit/process at this time using high-end PCs.

October 29th, 2003, 11:36 PM	#15
Steve Mullen HDV Cinema Join Date: Feb 2003 Location: Las Vegas Posts: 4,007	<<<-- Originally posted by Alex Raskin : I did some rather extensive tests on strobing with HD10 camera, and finally decided to go with 1/60, sharp picture, and (unfortunately) some strobing. -->>> Alex is right. 1/60th is the optimum. Anything higher is horrible. 1/60th is a 180-degree shutter just like film. (Darren, i hope you have this right on your DVD as it is crucial. You have a copy of my Shooting Guide with a detailed explanation of this aspect.) Of course, 1/30th is necessary when light goes too low. Alex is also right about p60. Since the HDV spec. limits 720 to 19Mbps, doubling the pre-compression data rate is going to increase compression artifacts terribly. As will 1080i -- which is also 2X the data rate -- even through the spec allows 25Mbps. 2X the data rate also dramatically increase heat and power consumption. __________________ Switcher's Quick Guide to the Avid Media Composer >>> http://home.mindspring.com/~d-v-c