OLPF-mod/fish-net/back-filter for Moire?

boulder said:
Way to ruin an exciting thread guys :(
Click to expand...

I apologize for getting all technical. I get frustrated sometime when incorrect information gets posted around and then people think they understand what is happening and think they understand how to fix it, and they'll be sadly mistaken when they get in the field and realize no, that's not how it works.

A UV filter does not remove moire, we all know this, it's not a question. I don't want someone thinking that will solve the problem even though a UV filter is good to protect from getting lens scratches!. :)

A filter that employs diffusion, like the Caprock, works in some cases and not others depending on the lens. I have tested this, I know (as has Barry). The purpose, just as with the OLPF is to spread the incoming light out so it is not sharp (blurry). The interesting part about the Caprock is each of the different sizes use different size and spacial frequency of circles of diffusion with clear areas around them. That is why they each work differently depending on the size of the lens which determines how the actual fine detail gets resolved at the sensor.

Since those of us trying to eliminate Moire and aliasing are tired of guessing which filter goes with which lens at which focusing distances, the goal is to put that diffusion in front of the sensor and after the lens so it can be tuned EXACTLY to the detail coming in, which is what we care about.

I've been using an UV filter on the front of my camera now since I got it, gee, why is there still Moire in my shots :happy:

Be sure to post a link here to the new thread about the results of the new OLPF! I cannot wait to see what you get!

And just because this horse hasn't been beaten enough, a pretty definitive product list for OLPF offerings, look at the offering from Epson:

http://www.epsontoyocom.co.jp/english/application/appli_opt.html

Would you look at that, they offer OLPF that have just IR cut filter OR one that has UV-IR cut filter. Gee I wonder why they off both and they are both called OLPF?

Edit: Oh, it doesn't get any better than this:

http://www.newton.dep.anl.gov/askasci/eng99/eng99312.htm

And I quote:

"An optical low pass filter is a fancy name for a thin transparent sheet that causes a special kind of blurring. The filter blurs the image, but only a pixel or two, and not more than that. This eliminates aliasing and moiré, yet hardly degrades the image. The amount of spreading is chosen specifically for the type of camera."

And this, which is I believe is what these OLPF are using:

"Some digital cameras use a birefringent plate to precisely blur the image. A thin sheet of quartz is used to split or deflect the light according to the polarization. One spot will be split into two spots by the birefringent plate. Then a second plate splits those spots again into a total of 4 spots, all of them a pixel or two apart. The sheet is placed in front of the image sensor. This makes the picture a little blurry and is enough to eliminate aliasing and moiré. Other digital cameras use a clear plastic plate that has dots molded into it to blur the image."

Again, this has nothing at all to do with color. Period. It differences in contrast (detail) between two pixels. But don't take my word for it, ask a scientist :)
 
Last edited:
But you still miss the point that the mathematical function of diffraction is identical of that of low passing the signal once you've plotted the FFT. Even the links some of the posters used to attempt to prove me wrong, explained exactly that.

Yes ok, I agree the physical effect is "blurring" but you are still cutting high frequency content to get that "blurring" effect.

And no an ordinary UV filter won't do the trick I never ever said that. You actually need a filter that cuts a small portion of the upper end of you visible spectrum too.

"An optical low pass filter is a fancy name for a thin transparent sheet thatcauses a special kind of blurring. The filter blurs the image, but only apixel or two, and not more than that. This eliminates aliasing and moiré,yet hardly degrades the image. The amount of spreading is chosenspecifically for the type of camera."
Click to expand...
Wow, I'm overwhelmed with the technical nature of that explanation. I guess this math stuff doesn't matter anymore.

Oh wait, you forgot to quote the next very important paragraph:

Electrical engineers use "low pass" filters too, but these let the low
electrical frequencies past, but block out the high frequencies. The optical
idea is similar. The highly detailed, "high frequency" image components are
blocked.
Click to expand...

Which completely proves what I've been saying all along. It's a frequency dependent LPF. Funny how you left that out. Again, the mathematical function translates to a low pass filter that is frequency dependent. Colors have frequencies, therefor it is also assumed that it is a color filter(for those of you who say it's not).

Anyway, I suppose using the term UV was wrong of me. The frequency bandwidth ORIGINATED in UV but is being mixed down into the baseband signal along with your frequencies of interest. So yeah, it's no longer UV wavelength per se.


It differences in contrast (detail) between two pixels. But don't take my word for it, ask a scientist :)
Click to expand...

Oh i did. i asked myself.

I said this about 4 pages back:

Detail is nothing more than the power level of a reconstructed pixel. In other words, detail is really just the power level contrast between two pixels.
Click to expand...

But you were too busy jumping all over me to actually read what I said.
 
Last edited:
Which completely proves what I've been saying all along. It's a frequency dependent LPF.
Anyway, I suppose using the term UV was wrong of me. The frequency bandwidth ORIGINATED in UV but is being mixed down into the baseband signal along with your frequencies of interest. So yeah, it's no longer UV wavelength per se.
Click to expand...
No, it doesn't actually because the frequency he is talking about is a spacial frequency of the projected image, it has NOTHING to do with color AGAIN. It has nothing to do with UV or any part of the color. Why can't you understand this simple concept?

It's the contrast that matters.

If you have very tiny thin perfectly red lines with black space between them, there is absolutely NO UV in that image in front of the lens. These lines are spaced closer than the pixels in the rows on the CMOS sensor. Even worse, we rotate those lines slightly. With no OLPF there will be a moire pattern and the red lines will look wrong with a pattern in them (as the rotation produces aliasing as well). There is NO UV LIGHT in this scene. To avoid this, we blur the image slightly to avoid picking up those patterns that don't exist. Blurring has nothing to do with UV and you don't shift light from one color to another. The frequency is relative to the fact that it goes from red to black to red to black SPATIALLY so quickly in the X and Y directions. Literaly in SPACE, which is from the image projected on to the sensor. The color has nothing at all to do with what is happening. It's the frequency that the projected image on to the CMOS sensor is relative to the frequency of the photosites on the cmos sensor. This has NOTHING to do with time. Things have a spacial frequency as well as a color frequency. You are filtering light next to each other by spreading it out. The color in the Moire pattern on the 7D has more to do with the Bayer pattern and how they line skip. The Panasonic GH1 has moire as well but does not have any color in the patterns.

Did you look at the damn OLPF for sale from Epson that I posted?!?! Did you see they come with AND WITHOUT a UV filter?

I think you are taking too much of your RF knowledge and trying to apply it to optical and that's not how these systems work. The frequency of the light (UV) has nothing to do with the sensor. It has nothing to do with nyquist theory or anything like that, whole different problem altogether. You're trying to think of the frequency of the light as being more than the sampling, but that's not what the active pixel bins are doing. They are not confined by nyquist at the pixel level because the photodiode is continious, in it's detection. It's only sampled periodically and reset but even at the highest framerates, the pixel is left open for so long, there is able time to count ANY frequency of light even close to visible spectrum. For nyquist to come into account, it would be about modulated color and subsequent measurements of the same pixel, which produces yet a third frequency as I mentioned earlier OR pixels next to each other and whatever spacial projection is happening over and between them in terms of constrast or detail if you will. The OLPF is not affecting this function, it's just spreading out all photons slightly so detail is not lost between the pixels on the sensor. That spreading of the light is NOT changing it's frequency and is not shifting light from one frequency to another, it just spreads out the direction of it. The frequency of the light does not matter. Pixels are achromatic on their own which is why you need an optical bandpass to only allow them to see red, green or blue. Since IR frequency can be so abundent, it has to be filtered additionally so it doesn't overwhelm the R, G and B photons.

You mention detail (constrast) is the difference in the values of two pixels next to each other. That is correct. That has nothing to do with color. The problem is if the image projected on those pixels has detail between were the pixels measure, like a tilted line, it creates a stair step in the recording image. No color is shifted to another color. You just blur the line a little and you loose sharpness but you spread the edges of that line across multiple pixels and make it bigger so it is not so fine. Exactly like how photoshop does antialiasing on a line. You reduce constrast (or detail) by blurring it to the point that nothing is between the lines. This is a low pass filter because if you took a row of pixels and plotted it out no matter what colors were there, you are causing sharp changed in contrast to happen more slowly. It has absolutely nothing to do with the actual colors involved per say, at all.

And for an experiment, go to your local hardware store and pick up two small pieces of screen. Now put them up next to each other and rotate them slightly while looking at maybe a white screen on the computer. Observe the moire pattern causes by the fact the holes in one screen do not line up with the holes on the other screen. Feel free to buy colored screen if you can find it and change the colors behind the screen. You will see no change in the moire other than how it's colored. Then, maybe, you will realize that moire is caused by physical differences in the periodicity of elements in a projected image and the capture device and absolutely devoid of any issue of color frequency.

Also, look at these images:

http://en.wikipedia.org/wiki/Moiré_pattern

You will see Moire is not caused by color issues, at all. Moire does not mention color because it's not a limitation of colors. It's a detail and sampling problem of detail.

I know I said this before, but seriously, I am done trying to help you understand concepts you do not grasp. My advice to you is stop trying to backwards engineer every graph you find on the internet and send an email to a company that produces OLPF, maybe they'll have more energy in trying to set you right. If you want to talk about it further, I'll be glad to answer in private, but this horse has been beat to absolute death.

http://xkcd.com/386/
 
Last edited:
Just to finalize the concepts here, I had written to Sunex (http://www.sunex.com/) yesterday asking them about the properties of the OLPF. This is the response from one of their optical engineers:

"Hi,
You are right … in this case OLPF has nothing to do with light frequencies, but rather spatial frequencies. OLPF uses the birefringent properties of crystals to spread light. When we use multiple layers of such crystals and divide multiple times, we create patterns, such as square, parallelogram, etc. In essence, this is how it works.

Regards, XXXXXXX"
 
jacked.gif


Some seem to be enjoying this tangent.

Unfortunately, their feet are not on the ground and they completely missed the purpose of this thread was a pragmatic one.
 
Last edited:
USLatin said:
So, bottom line, any real hope on any of these fronts Barry?
snip
Click to expand...

Not Barry here but have you considered dressing your actors in "Mao gray" (70's cultural rev style) and filming with only landscape backgrounds? :happy:

Gee..................... I was rather enjoying Svarts "analysis". Hope you guys didn't chase him away. :Drogar-Shock(DBG):

JohnG
 
You must log in or register to reply here.
Back
Top