Noise and the Matrix

So is the coring setting the threshold of spacial frequency? Trying to understand what the criteria is for sharpening there.

I've been thinking about this a lot lately, and I can certainly see that I wonder the order in which these things are applied. Here is why I say that:

If the circuitry works like this (and by circuitry I mean digital algorithms as well)

IMAGE->SHARPENING->COLOR CURVE

Then I can see why the noise characteristics are changing at different IRE levels as to what some have noticed. Mainly because a non-linear curve will in some cases squish the noise delta smaller, and in other cases expand it. In this case, the sharpening is happening BEFORE the curve, so it won't filter out those spatial frequencies that get created after the curve is applied.

However, if you were to do

IMAGE->COLOR CURVE->SHARPENING

In this case, you'd apply the curves already to the image, which will expand and contract the noise differences, which then, the coring setting would be able to ensure that any noise with a specific level would not get sharpened. This would keep noise lower in the final result.
 
Mike@AF said:
I thought about buying one of those Du Monde charts, but I see they're about $700+. Any recommendations of alternatives that are much, much less money? I'd like to do some of these tests on my own to see what I come up with.
Click to expand...

The CDM (Chroma Du Monde) Cam-Align charts are so expensive because each chart is painstakingly hand-checked.

You can pick up one of their most basic Front-Box Cam-Align series charts (batch calibrated/checked) for about $280. One with flesh tones starts at $335. You could pick up a pocket chart for as little as $105 bucks.

They're running a decent deal now where you can get a pocket White 'n Warn "chart", a Front-Box 12+4 a stand and little portable bags for about $540 and free shipping (called the DSC Field Kit)

I don't work for DSC Labs - but am about to become a customer and so have been familiarizing myself with the product lines.
 
Mike@AF said:
I thought about buying one of those Du Monde charts, but I see they're about $700+. Any recommendations of alternatives that are much, much less money? I'd like to do some of these tests on my own to see what I come up with.
Click to expand...

If you're a student, you can get their Imaging Technology book, complete with 5 test charts for $70. I remember seeing this book at B&H a couple of years ago, but now I think you have to get it through your school.

http://dsclabs.com/student_edu.htm

Maybe they'll do a DVXUser group buy.
 
I had a try on what Coring does to the image and using a macro lens in 1:1 on the Marshall Coring does take away alot of the coarse noise found. It's hard to judge but when I turn it up to +7 much of the hard noise goes away and left is a smoother image. I even did it with Detail turned down to -7 so it works regardless of what Detail setting you have.

Had another go at the Matrix settings and Fluo/Norm 2 has much less chroma noise than Norm 1 or Cinelike. But Fluo is adding magenta to compensate for fluorescent lights.

So my updated settings are:
Film camera
1080p25
Gamma Cinelike V (B-press is also pretty good)
Matrix Fluo or Norm 2 (Fluo adds magenta so red will look less correct)
Detail -7
Coring +7 (It has to have something negative about it or?)

Everything else at neutral. I've done these tests in 320/1250 ISO and for what it is 1250 ISO looks good.
 
Those settings will give you a very noise free image, but not too contrasty for you?

I'm still debating between the ideal numbers for detail but it looks like I'm stuck between -5 and -7, still doing tests with the coring and haven't going the magic setting yet.
 
The downside of coring is that if you set it too high, it starts to remove fine detail that you want, and things can start to look smoothed-over and plastic-y.
 
I'd say a good test on coring would be faces. The 7D for instance really smooths out faces and makes them look pretty plastic on medium shots. That's what I will be testing on mine, as I really want some of that face detail back.
 
As far as faces go, don't forget the SKIN DTL function -- it actually works on the AF100! Prior Panasonic users know that the skin tone function really didn't much of anything at all on the DVX/HVX etc. But on the AF100 it's quite effective. Crank up the detail to +7, and your subject's skin will look like it's grainy and gnarly. Then turn on the SKIN DTL function and bam -- they'll look like an airbrushed magazine cover model.
 
Petros Kolyvas said:
The CDM (Chroma Du Monde) Cam-Align charts are so expensive because each chart is painstakingly hand-checked.

You can pick up one of their most basic Front-Box Cam-Align series charts (batch calibrated/checked) for about $280. One with flesh tones starts at $335. You could pick up a pocket chart for as little as $105 bucks.

They're running a decent deal now where you can get a pocket White 'n Warn "chart", a Front-Box 12+4 a stand and little portable bags for about $540 and free shipping (called the DSC Field Kit)

I don't work for DSC Labs - but am about to become a customer and so have been familiarizing myself with the product lines.
Click to expand...

RED and DSC came out with a nice one as well:

http://www.filmtools.com/dsc-labs-red-cambook.html

Not bad at all for $425.00

cinemasupplies_2139_123464073
 
I compared my AF to a Canon 5D (which is a camera I'm very pleased with) in noise and sure there are about as much noise from the 5D as the AF101. Somehow I always end up with Cinelike/Fluo noisewise since it makes the image with the least noise. But Norm 2 looks even a bit smoother in blacks and doesn't have the slight magenta tint.

During this latest test I ran in blue/green/red to see if the colurs had more or less noise than any other colour but they didn't to any extent. But I can tell you Cine D is the noisiest of them all. Blown up in 1:1 they look like big chunks of pixels.

After more testing Detail between -7 and -5 looks about the same in noise but -5 has a smidgen more detail. Also V Detail can be cranked high since it's impact on noise is small. It also has a little impact on sharpness.

So:
Cinelike
Norm2
Detail -5
V Detail +7
Coring +5

Makes it look just as good as the 5D when it comes to noise. Actually a bit better since I start to see the CMOS pattern in the 5D images.
 
Last edited:
editman said:
I compared my AF to a Canon 5D (which is a camera I'm very pleased with) in noise and sure there are about as much noise from the 5D as the AF101. Somehow I always end up with Cinelike/Fluo noisewise since it makes the image with the least noise. But Norm 2 looks even a bit smoother in blacks and doesn't have the slight magenta tint.

After more testing Detail between -7 and -5 looks about the same in noise but -5 has a smidgen more detail. Also V Detail can be cranked high since it's impact on noise is small. It also has a little impact on sharpness.

So:
Cinelike
Norm2
Detail -5
V Detail +7
Coring +5

Makes it look just as good as the 5D when it comes to noise. Actually a bit better since I start to see the CMOS pattern in the 5D images.
Click to expand...

I tried cranking the Vdetail and wasn't too pleased.

If you have a monitor like me, with 1:1 pixel, grab a lens, focus on lines far away, crank the Vdetail up and watch as you nudge the camera up and down the level of aliasing. Lowering the Vdetail lowered the aliasing from what i could see.
 
That seams to be a very probably side effect of cranking V Detail up. I did test it in 1:1 mode but couldn't see it. Thanks for pointing it out.

Having done crazy no-lit tests I finally turned on some lights and yeah it's a super clean image. Can't wait to put it to use :)
 
Last edited:
mcgeedigital said:
RED and DSC came out with a nice one as well:

http://www.filmtools.com/dsc-labs-red-cambook.html

Not bad at all for $425.00
Click to expand...

Sumfun said:
If you're a student, you can get their Imaging Technology book, complete with 5 test charts for $70. I remember seeing this book at B&H a couple of years ago, but now I think you have to get it through your school.

http://dsclabs.com/student_edu.htm

Maybe they'll do a DVXUser group buy.
Click to expand...

Barry_Green said:
Check out the Camette at B&H: under $100.
http://www.bhphotovideo.com/c/product/465713-REG/DSC_Labs_PCE_PCE_Pocket_Camette_Test.html
Has a color & gray chart on the front, and a camwhite on the back.
Click to expand...

I just spent a couple hours I think looking at all the DSC Labs options. None of them work for what I want without have to spend about $700.

I'm not a student so the student book is not an option. It also only has a 5 step grayscale and the resolution chart only goes up to 800.

The Camette is only a 5 step grayscale. I want to test the dynamic range of the AF100's various settings and feel I really need to have more steps. They also have only a backfocus option and not a resolution option. I'd prefer to have resolution so I can test the detail and coring settings.

I'd be happy with the Front Box Standard, except that it doesn't have resolution, and on the back is backfocus instead of resolution. Now they do have the Front Box CineZone with CamWhite on the back, but the resolution only goes up to 800.

And the CamBook 5 and CamBook 6 would actually be perfect, but they only go up to 800 on the multiburst resolution chart as well.

So that leaves the CamAligns which are expensive. The CamAlign ColorBar/GrayScale has almost everything and resolution goes up to 1200. The only thing missing is the CamWhite card. It costs $690 plus shipping at B&H.

Is there any other options out there that will give the same features at a lower price? I'm looking for 11 step grayscale, 6 primary colors, and resolution of 1200 or so. A skin tone card like the CamBelles would be nice too and a CamWhite-like white card would be a bonus.
 
Last edited:
Ah, but you CAN get more steps out of the Cambook, I do it all the time. The key is to use a light with barn doors so you can light up the top half of the chart directly, and flag the light off the bottom half. Use your light meter to balance the ratio, but yeah, you could easily double the testable dynamic range with that.

Frankly, frontlit charts aren't of all that much value for testing dynamic range anyway. If you want an effective dynamic range test, you want the Ambi backlit chart, either the 14-step or the 17-step. That thing's perfect, it's immensely accurately calibrated, it is not affected by reflections or anything. It's ideal. Frontlit -- meh. I tried to make do with a Stouffer wedge step transmission chart, as it's soooo much cheaper, but ... blah. It's designed for testing scanners, not video camera dynamic range; it sort of works but it's not really accurate at all. The Ambi is the best for testing dynamic range.

If you want to use a front-lit, like I said, you can make it happen but it takes a little bit of creative lighting, it takes a lightmeter, and you also should ideally be conducting that test in a place where there are ZERO reflections possible. A black-painted studio would be best, and you should be operating the camera from behind a sheet of duvetyne. Yes I'm making it sound ridiculous and obsessive, but hey, if you want accuracy, you gotta do what it takes. Reflections off a glossy chart can seriously impact your readings, so you need there to be no reflections.
 
Barry_Green said:
Ah, but you CAN get more steps out of the Cambook, I do it all the time. The key is to use a light with barn doors so you can light up the top half of the chart directly, and flag the light off the bottom half. Use your light meter to balance the ratio, but yeah, you could easily double the testable dynamic range with that.

Frankly, frontlit charts aren't of all that much value for testing dynamic range anyway. If you want an effective dynamic range test, you want the Ambi backlit chart, either the 14-step or the 17-step. That thing's perfect, it's immensely accurately calibrated, it is not affected by reflections or anything. It's ideal. Frontlit -- meh. I tried to make do with a Stouffer wedge step transmission chart, as it's soooo much cheaper, but ... blah. It's designed for testing scanners, not video camera dynamic range; it sort of works but it's not really accurate at all. The Ambi is the best for testing dynamic range.

If you want to use a front-lit, like I said, you can make it happen but it takes a little bit of creative lighting, it takes a lightmeter, and you also should ideally be conducting that test in a place where there are ZERO reflections possible. A black-painted studio would be best, and you should be operating the camera from behind a sheet of duvetyne. Yes I'm making it sound ridiculous and obsessive, but hey, if you want accuracy, you gotta do what it takes. Reflections off a glossy chart can seriously impact your readings, so you need there to be no reflections.
Click to expand...

Good tip on using a light with barn doors, so that could solve the dynamic range aspect well enough for me. However, the CamBook is still stuck at 800 resolution, right?

I agree with you on the accuracy part. I don't really like the fact that they're glossy. I watched the video on DSC Labs' website and they said it's because it gives you a real pure black. The problem is that you'll rarely get a real true black in an actual shoot because wardrobe is typically not glossy, and anything that is glossy is going to have something reflecting off it that makes it not true black anyway. I'd rather have the matte finish so it's more like at actual shooting situation.

I suppose I can go with the CamBook 5 or 6 and barndoor the light. Do you think 800 might be enough resolution to test the detail and coring settings?
 
You must log in or register to reply here.
Back
Top