h a l f b a k e r yThunk.
add, search, annotate, link, view, overview, recent, by name, random
news, help, about, links, report a problem
browse anonymously,
or get an account
and write.
register,
|
|
|
Please log in.
Before you can vote, you need to register.
Please log in or create an account.
|
With the advent of relatively inexpensive image processing in digital video cameras, it should now be possible to fix the long standing temporal aliasing artifact known as the Wagon Wheel Effect. [link]
By varying the precise frame exposure time, {maybe on selected areas - per [jinbish]}, but keeping
the average frame rate constant, one could prevent the apparent "reverse rotation" of moving wheels.
The director could specify the apparent motion desired, typically forward rotation, and rapid enough to convey the effect of continuous rotation. Frame-to frame motion analysis (as is accomplished in some types of MPEG-2 video compression) can take place "on the fly," and image capture timing adjusted accordingly.
I'm almost surprised that this doesn't seem to have been implemented yet, though my search was pretty perfunctory.
Wagon-wheel Effect
http://en.wikipedia.../Wagon-wheel_effect
Why wheels in movies sometimes seem to be rotating backwards [csea, Jul 29 2007]
Cake Effect
http://www.cakemusic.com/tour.html
[normzone, Jul 30 2007]
[link]
|
| |
Other than by human intervention, how can you identify that aliasing effect is happening? |
|
| |
Aliasing is normally mitigated against by filtering to ensure that sampling is done below the Nyquist rate. I wonder what would happen if you did some kind of filtering over a set of frames to remove/smooth out high frequency components... |
|
| |
[jinbish], machine vision today can readily identify round shapes, and spokes within. |
|
| |
One way to prevent aliasing is to up the sample rate _above_ the Nyquist frequency (lots of frames per second.) But this eats up bandwidth, storage, etc, and isn't directly compatible with standard display rates. |
|
| |
Filtering to reduce high frequency content tends to "blur" the image, wich actually might be a pretty good idea if done locally to just the wheels! Good suggestion, thanks. If you did this on the entire image, you'd get a blurry/fuzzy image. |
|
| |
We're talking about the same thing - I mean that the frequency of the data signal is below the sampling frequency. |
|
| |
Also, of course computer vision can identify a wheel - but can it identify that it's causing the wagon wheel effect quite so easily? You'd need to test consecutive frame captures, identifying the wheel as an artefact and it's rate of spin. |
|
| |
My suggestion was based on just this: Identifying the wheel and blurring it a bit. (Didn't say it explicitly, I grant you, but it's what I was thinking - honest!) |
|
| |
Ultimately, isn't our eye going to be the sampling factor? So the wagon wheel effect is always going to be in play... |
|
| |
//Ultimately, isn't our eye going to be the
sampling factor?// No, because your eye
doesn't have a frame rate. You do
sometimes see aliasing effects under
artificial light, but only if bad lighting is
acting as a strobe. |
|
| |
In many technologies, there are trade-offs between making images look good in motion, and making them look good in freeze-frame. Increasing the proportion of each frame time that the shutter is open will cause objects whose motion would exceed Nyquist to appear as a blur. This will generally improve the appearance of the objects when the film is projected at speed, but will naturally mean that the objects will be blurred on every frame. |
|
| |
You're quite right about the tradeoffs. The basic idea, with clarification by [jinbish], is to selectively isolate rotating portions of an image, and blur just that portion, while retaining crisp outlines of the rest of the frame. |
|
| |