Possible Fix for CA / Color Fringing?

[Tom Roper]

When experimenting with the VividRGB preset created by Steven Dempsey, I happened upon an observation when using the modded preset below...what appears to completely eliminate objectionable color fringing, aka chromatic aberration. It seems wholly disappeared from my observations, marginal vestiges at the extremes remain at most. Anyone want to try this out? You have to use the presets below, exactly, doesn't work for gamma Cine1. If someone can duplicate the observation, we can begin deviating from the preset to narrow down the contributing variables.

Gamma: Normal
Color Matrix: Normal
Color Gain: 25
Color Phase: 0
Knee: Auto
Black: Middle
Master Ped: -5
Setup Level: 0
HDF: Mid
H/V Detail: 0
Sharpness: 0
NR1: Off
NR2: Off
Coring: 0
Red Gain: -2
Green Gain: -2
Blue Gain: -3
RG Matrix: 0
RB Matrix: 0
GR Matrix: 0
GB Matrix: 10
BR Matrix: 0
BG Matrix: -13

[Ken Ross]

Tom, any idea how a preset could erase a lens deficiency?

[Tom Roper]

According to what Chris Hurd and A.J. DeLange have been saying all along, it's not a lens phenomenon, agree or disagree as you choose. But if it was, why would this fix it? So we seem to be in agreement about that.

Anyone else try this yet?

[Piotr Wozniacki]

One important factor is that with lower contrast the fringing is less visible. And not engaging cine (gamma or colours) simply diminish the contrast, hence the result...IMHO:)

[Ken Ross]

Originally Posted by Tom Roper

According to what Chris Hurd and A.J. DeLange have been saying all along, it's not a lens phenomenon, agree or disagree as you choose. But if it was, why would this fix it? So we seem to be in agreement about that.

Anyone else try this yet?

I guess the thing I don't understand is if it's not CA, why does it tend to show up on the edges of the frame where you would expect CA to show? If it's electronic, wouldn't that manifest itself in many parts of the frame?

[Tom Roper]

Originally Posted by Piotr Wozniacki

One important factor is that with lower contrast the fringing is less visible. And not engaging cine (gamma or colours) simply diminish the contrast, hence the result...IMHO:)

There's more contrast to the gamma Normal setting. Cine1 flattens the highlights. It's already a high contrast setting by virtue of the Master Pedastal -5 setting. That's why someone needs to try this so we can stop guessing.

[Michael Mann]

Tom, I'll give it a try today in bright sunlight.

[Michael Mann]

Meanwhile I tested the preset:
Frankly, I don't see significantly less fringing than with any other preset I have used so far. The fringing seems to have less contrast but it is still visible.

[Alex Leith]

Today I did some tests to see if I could find out if changing the preset settings helped to make any difference to the CA / green-magena colour fringing.

I found an area high contrast (a tall white ornament against a dark wall) and positioned it towards the edge of the frame on a wide lens.

What I found surprised me. With the custom presets off there was very little fringing at all, but as I started switching between different custom presets it became very apparent that the presets make a large difference to how obvious the fringing appears. VIVIDRGB setting showed a little more fringing than with presets off, and Wolfgang's new preset showed a huge amount of fringing (which is a shame because I wanted to use that preset!)

Anyway, my test wasn't particularly scientific. I wasn't recording, just mucking around. And I didn't have long to try it out. So I started making changes to the presets to see if I could understand what was going on, but I'm not really certain so far. Cine Gamma and colour Matrix seem to make the fringing look worse. And running the colour gain higher also makes it more obvious. But I need to get a bit more time experimenting with this to see if I can work it out.

[Tom Roper]

So it sounds like we're just making what's already there more or less apparent.

By increasing color saturation we're enhancing the fringing as well.

I still find some support in the observation about the Cine1 gamma showing more fringing. It seems noticeably less as the contrast setting swings to more, not what you'd think.

[Alex Leith]

With a little more time and an only-very-slightly-more-scientific approach (I wrote down what settings I was using!) I've found that there is very little in the custom presets and gamma settings that effects the actual amount of Chromatic Aberration / Green-Magenta Fringing in the image. However, the fringing may be hidden in the image better by some presets, depending on the nature of the scene you're shooting at that moment.

The only two controls that apparently effects the amoung of fringing in the image are HDF (Horizontal Detail Frequency) - low shows less fringing than mid, which shows less fringing than high; And DHV (Horizontal Vertical Detail Balance) - lower numbers show less fringing than higher numbers.

So, my conclusion is that you could significantly reduce fringing by setting HDF to LOW and DHV to -9, but obviously there is a significantly detremental effect to the quality of the image!

The attached images are a 1:1 resolution (click to see them at full size) showing the lower-right-hand area of the HDV frame, with the lens zoomed to widest angle.

[Ken Ross]

Originally Posted by Alex Leith

With a little more time and an only-very-slightly-more-scientific approach (I wrote down what settings I was using!) I've found that there is very little in the custom settings that effects the actual amount of Chromatic Aberration / Green-Magenta Fringing in the image. However, the fringing may be hidden in the image better by some presets, depending on the nature of the scene you're shooting at that moment.

The only two controls that apparently effects the amoung of fringing in the image are HDF (Horizontal Detail Frequency) - low shows less fringing than mid, which shows less fringing than high; And DHV (Horizontal Vertical Detail Balance) - lower numbers show less fringing than higher numbers.

So, my conclusion is that you could significantly reduce fringing by setting HDF to LOW and DHV to -9, but obviously there is a significantly detremental effect to the quality of the image!

The attached images are a 1:1 resolution (click to see them at full size) showing the lower-right-hand area of the HDV frame, with the lens zoomed to widest angle.

So it's pretty apparent that this is indeed a lens issue and not an electronic issue.

[Tom Roper]

Great work Alex!

Thanks for posting those. I'd really like to see them in full rez 1920 x 1080 to observe what other differences beside CA change with DHV and HDF, resolution wise. I'm assuming it was the server that reduced the size of your images to 640x360.

[Alex Leith]

Originally Posted by Tom Roper

Great work Alex!

Thanks for posting those. I'd really like to see them in full rez 1920 x 1080 to observe what other differences beside CA change with DHV and HDF, resolution wise. I'm assuming it was the server that reduced the size of your images to 640x360.

It's a full-size crop from an HD frame. It was just filmed against a blank wall, so there wasn't anything to see (apart from my home-made CA tester chart).

If I get a chance I'll shoot a frame later on today with something in it to see how it looks with those changes to DHV and HDF.

[Alex Leith]

Here are some of the full rez frames (just for reference). Not a lot going on. Just some peeling wallpaper in a room that's due to be decorated ;-D