Tag Archives: motion

Temporal Aliasing – Beware!

As camera resolutions increase and the amount of detail and texture that we can record increases we need to be mindful more and more of temporal aliasing. 

Temporal aliasing occurs when the differences between the frames in a video sequence create undesirable sequences of patterns that move from one frame to the next, often appearing to travel in the opposite direction to any camera movement. The classic example of this is the wagon wheels going backwards effect often seen in old cowboy movies. The cameras shutter captures the spokes of the wheels in a different position in each frame but the timing of the shutter relative to the position of the spokes means that the wheels appear to go backwards rather than forwards. This was almost impossible to prevent with film cameras that were stuck with a 180 degree shutter as there was no way to blur the motion of the spokes so that they were contiguous from one frame to the next. A 360 degree shutter would have prevented this problem in most cases. But it’s also reasonable to note that at 24fps a 360 degree shutter would have introduced an excessive amount of motion blur elsewhere.

Another form of temporal aliasing that often occurs is when you have rapidly moving grass, crops, reeds or fine branches. Let me try to explain:

You are shooting a field of wheat, the stalks are very small in the frame, almost too small to discern individually. As the stalks of wheat move left, perhaps blown by the wind, each stalk will be captured in each frame a little more to the left, perhaps by just a few pixels. But in the video they appear to be going the other way. This is  because every stalk looks the same as all the others and in the following captured frame,  the original stalk may have moved  say 6 pixels to the left. But now there is also a different stalk just 2 pixels to the right of where the original was. Because both stalks look the same it appears that the stalk has moved right instead of left. As the wind speed and the movement of the stalks changes they may appear to move randomly left or right or a combination of both. The image looks very odd, often a jumbled mess, as perhaps the tops of the stalks appear to move one way while lower parts appear to go the other.

There is a great example of temporal aliasing here in this clip on Pond5 https://www.pond5.com/stock-footage/item/58471251-wagon-wheel-effect-train-tracks-optical-illusion-perception

Notice in the pond 5 clip how it’s not only the railway sleepers that appear to move in the wrong direction or at the wrong speed but notice how the stones between the sleepers appear to look like some kind of boiling noise.

Like the old movie wagon wheels one thing that makes this worse is the use of too fast a shutter speed. The more you freeze the motion of the offending objects or textures in each frame the higher the risk of temporal aliasing with moving textures or patterns. Often a slower shutter speed will introduce enough motion blur that the motion looks normal again. You may need to experiment with different shutter speeds to find the sweet spot where the temporal aliasing goes away or is minimised.  If shooting at 50fps or faster try a 360 degree 1/50th shutter as by the time you get to a 1/50th shutter motion is already starting to be as crisp as it needs to be for most types of shots unless you are intending to do some for of frame by frame motion analysis.

NEX-FS700 Significantly reduced shutter when in Super Slow Mo!

UPDATED WITH NEW FRAME GRABS FROM STROBE LIGHT AT BOTTOM.

One of the things that did concern me slightly about the FS700 was how would the sensor behave in Super slow Mo. The sensor is a CMOS sensor, so I expected it to exhibit rolling shutter artefacts, which it it does indeed do when in standard shooting modes and S&Q motion. It’s not bad, but you can make the pictures skew and when you try to shooting something like a spinning propellor you can get some weird effects, especially at higher shutter speeds. However when you switch the camera to Super Slow Mo the rolling shutter effects appear to go away. I was able to shoot propellors, do fast pans, shake the camera about etc and there was little sign of the usual rolling shutter artefacts.

FS700-Fan-Norm2-300x168 NEX-FS700 Significantly reduced shutter when in Super Slow Mo!
FS700 25P 1/100

Just take a look at these two frame grabs. One shot done at 25P with a 1/100th shutter, the other done at 100fps with a 1/100th shutter, so in both cases the shutter speed is the same, so you would expect the rolling shutter artefacts to be the same, but clearly they are not. In standard mode the fan exhibits a typically lop sided, asymmetrical look and the fan blades appear curved, the upper and lower fan blade both bent towards the right of the frame. But in Super Slow Mo mode the fan blades are straighter and the fan is a lot more symmetrical with noticeably less bias towards the right, notice in particular the differences in the lower fan blade.

FS700-Fan-SS2-300x168 NEX-FS700 Significantly reduced shutter when in Super Slow Mo!
FS700 Super Slow Mo 1/100th shutter

You can tell the shutter periods are the same as the amount of motion blur and spreading of the fan blades is near identical, so it’s not a shutter speed difference, this is clearly a sensor scan difference. This is very interesting and requires further investigation as it suggests that the sensor read out process is different in the high speed mode. It is probably just a significantly faster scan rate, but it could also possibly be a global shutter of some kind. It’s just a shame that you can’t access this read out mode for normal shooting.

UPDATE:

FS700-Norm-Flash-300x168 NEX-FS700 Significantly reduced shutter when in Super Slow Mo!
FS700 Flash band at 25fps 1/100th shutter.

Here are a couple more frame grabs done with the strobe focussing flash from a Canon DSLR. In both cases the shutter speed is 1/100th of a second so you would expect the width of the “Flash Band” to be the same. The narrower the band, the slower the sensors scan speed. These frame grabs suggest the scan speed is around twice as fast when in Super Slow Mo. It’s not a global shutter, but certainly a nice improvement. This is 100% repeatable.

FS700-SSM-Flash-300x168 NEX-FS700 Significantly reduced shutter when in Super Slow Mo!
FS700 Flash band in Super SlowMo 100fps, 1/100th shutter

You can take advantage of this for normal speed shooting by setting the camera to SSM and  recording the SDior HDMI feed to an external recorder.

Speculation: There is a little more aliasing when shooting in SSM. Is there some line slipping going on perhaps during SSM? This would allow a faster scan speed as fewer lines of pixels are read and thus might account for both the slight aliasing increase and the faster read out speed.

NEX-FS700 Slow Motion test clip.

I was lucky enough to be able to borrow a pre production Sony NEX-FS700 for an evening and of course the one thing I had to check out was the super slow motion function. So my good friend Den Lennie let me shoot from his balcony overlooking the Belagio fountains in Las Vegas. The video speaks for itself really. The slow motion function is incredibly easy to use and I was surprise how well it performed shooting at night at 240 frames per second (1/240th shutter). There are lots of other nice features on the FS700 which I’ll write more about in a later post.

 

Motion “Judder” on the FS100 and other video cameras at low frame rates.

There have been a number of threads in various forums about the way the images from the new Sony FS100 appear to judder or stutter when shooting at 25P or 24P. Most of the complaints appear to be coming from PAL areas where shooting 25P is common. This is not an issue unique to the FS100, in fact motion judder is often more noticeable with video cameras than film cameras even though the frame rates and shutter speeds may be exactly the same. Why is this?

One of the key issues here and I believe a very strong clue to what is going on is that most complain that the issue is most pronounced in areas of high contrast.

Our visual system picks up edges and other areas of high contrast to detect motion, in areas of high contrast any non-smoothness of the images motion will be more noticeable. The higher the resolution/contrast or more precisely the higher the MTF of the camera system the more we will notice judder and stutter.  Just take fast motion in an Imax film as an example, it stutters like crazy.

The FS100 and similar high contrast/resolution cameras will appear to stutter at low frame rates more than a low contrast/low resolution camera. Edges in film are almost never instant changes from black to white, there is almost always some smoothing or dithering caused by the grain structure of film. So when you consider the FS100’s near complete lack of noise, which through it’s random nature will help mask judder and stutter and you have a worst case scenario. A camera with sharp edges and no noise.

Another strong contributing factor is the use of detail correction that adds a very definite, hard, non-motion blurred black or white edge around any areas of medium to high contrast, so unlike the very slightly dithered edges we would see in film we have instant light to dark or dark to light transitions occurring over a single pixel. In the case of a pan that hard edge is going to step uniformly from one position to the next, it won’t have any motion blur and it will increase edge contrast compounding the images judder as our visual system will notice these hard edges jumping from one place to the next.

The PMW-F3 although it uses the same sensor is less prone to this effect as it has a more sophisticated DSP and uses less detail correction and more aperture correction for image sharpening. Aperture correction blurs with motion as it is a type of high frequency boost and as you pan the camera the motion blur of the image reduces high frequencies so the amount of correction also drops thus helping smooth edges as you pan.

You also need to consider the results of watching 25fps video on a computer monitor typically running at 60hz. You will get judder as 25 does not go into 60 evenly, this helps explain why this “issue” is getting more airtime in Europe than in the US where 24P with pull up to 30P is common and of course 30P will display on a 60Hz monitor with no additional problems.

So in the case of the FS100 (or other cameras exhibiting this effect),  I would suggest turning off the detail correction circuits or at the very least reducing the detail level if you are shooting high contrast images or anything with a lot of motion. It would also be interesting to compare similar pans at different speeds with some gain added to see if that helps.

I don’t think this is, as claimed by some, to be camera fault, more likely a result of a very clean, detail corrected image. Even an EX1 or EX3 will do similar things if your detail settings are too high. It’s not unique to the FS100, just one of those things that can happen when you have sharp pictures. When I watched the Sony F65 4k demo film “The Arrival” I noticed a similar increase in motion judder compared to film, again I put this down to high edge sharpness catching my eye and making me notice the cameras motion more acutely. Ohh that F65 stuff looked stunning!