Novel Moire reduction technique

[JohnBarlow]

I was out and about the other day and saw some fellow filming with a very odd thing in his camera hotshoe.

 

It was a small motor with a small eccentric weight which gently vibrated the camera at some set speed. He explained that is was a cheap (<£10) way of ridding moire from his footage.

 

I thought he was maybe quite mad, but later found that Pentax are using their in-body stabilisation to perform the same function.

 

How weird is that?

[richg101]

this method is also used on dof adaptors. wobbling the ground glass to mimmic the grain changing every frame.

 

loving the idea to use it on a camera to reduce moire.  great idea.  was it just a motor with an offset weight?

[JohnBarlow]

Yep, it looked like a cheap chinese motor probably 6v with a pot to control speed, the eccentric looked like the earth brass pin from a UK mains plug but cut down.

 

I guess someone could make nice looking ones for cheap to sell

[Nikkor]

I wouldn't use that on my cameras, if the frequency of the vibration coincides with the fundamental frequency of some structure inside you will break it.

[maxotics]

this method is also used on dof adaptors. wobbling the ground glass to mimmic the grain changing every frame.

 

Hi richg, having built a crude version of a DOF adapter myself, I can tell you that the reason the glass is vibrated is to stop the same grain from showing in the frame.  The ground glass has very fine grain to diffract the light enough to be focused on.  If you don't vibrate, you see the grain in the same place.  Vibrating eliminates seeing the ground glass graininess,though you still get "graininess" because the ground/glass/vibration is adding softness to the image.

 

I have studied the moire problem in some depth.  I have tried many kinds of software filters, bent plastic, and other goofy ideas.  I have also thought of the vibration idea, but feel it is no different that not pulling perfect focus or using a heavy soft filter.

 

I've created many raw (DNG) images of moire-inducing charts.  My sense is that much of the aliasing effects have a similar interference pattern to them.  I believe software can eliminate much of it, indeed, believe the camera manufacturers have it as part of their firmware when converting to H.264.  Wish I could find some open-source versions.  

 

I want to point out, to anyone starting with this that the de-bayering algo greatly influence the severity of chroma problems in this.  Amaze and LMMSE work well fro me.  

 

There are tons of chroma smoothing, etc., algorithms, but they all work on a center pixel or block of pixels.  An algo to deal with moire needs to work on multi pixel lines of these interference patters.    It needs to scan a line and say, this 8 pixels of RGRGRGRGRG are too bright, and interpolate them to match the line above and below.  The logic needs to think about the problem in lines.

 

All that said, this must be a VERY difficult problem.  The sony VG cameras, which aren't cheap, suffer from moire and you would think if Sony could fix it, they would. 

 

The filters from Mosaic may be the only solution.  However, they need to be taken out before photos are taken, or one needs to live with the fact that the image must be degraded enough to prevent a lot of visible moire in video, which can be seen in high resolution photo mode.