Unexpected res test results at DVinfo.net

Bob Grant · February 6th, 2008, 05:57 PM

We ran some tests with an unusual test chart, well unusual for testing video cameras. All the results were in line with Adam Wilts conclusions. Except for one small thing.
This res chart includes 4K plates. As expected they just appear as grey areas at 50i. In 25p at one point when the camera was moved ever so slightly I can see aliased remnants of the 4K patterns. Would seem to indicate the lens is capable of very high resolution and the OLPF is set high.
We all want all the res we can get however I think what I saw might call for a heads up, in some scenarios we might need to wrangle that much resolution to avoid aliasing and/or line twitter.
I'll see if I can get permission to post a frame grab if anyone is interested.

Steven Thomas · February 6th, 2008, 09:21 PM

I'm looking forward to seeing it Bob.

Tom Roper · February 7th, 2008, 12:14 PM

I LOVE looking at resolution charts! But the eye's perception is the best judge of the overall resolution and picture quality.

My judgment is that the process goes awry when you try and attach a number to resolution chart testing.

It's just not possible to visually judge when the contrast drops to 50%, or even agree if that's the right number to apply to what you see.

Inbetween 10% and 90%, you can observe resolved lines, but one gives you a TV Lines resolved number of 600, the other 1000.

Even if you try to be more objective with the process, using resolution testing software like Imatest, and using a MTF50 as a baseline, the number you get depends on what part of the picture you look at, and changes with everything including zoom range, focal length, distance to target, aperture and the amount of in-cam sharpening used. All have an effect on the number of TV lines. In the case of in-cam sharpening, the difference is huge! Imatest software attempts to normalize measurement results by using a 2 pixel sharpening radius as the standard. You have to be a dedicated pixel peeper to discern that visually.

What number are you shooting for? A peak number obtained in a limited area of the picture? Or a broader overall number that is averaged over many different parts of the picture, zooms, focal lengths, distance to target and aperture?

Of more concern, if someone is attaching a TV lines of resolution number to a photographed chart, HOW did they come up with it? They usually don't go into any detail to describe their methods. Even Adam Wilts gives only vague mention of his methods.

My experience with the Imatest software tells me I can come up with many different numbers to describe the same image, just depending on how I define it. And they are all valid numbers for the same image...600,700,800,1000!

So I don't preach reliance on attaching numbers to resolution charts. It's more valid to put cameras side by side and obtain results, and let the eye be the judge for perceived differences for images shot under identical test conditions for lighting, lens length, aperture etc.

February 6th, 2008, 05:57 PM	#1
Bob Grant Trustee Join Date: Nov 2005 Location: Sydney Australia Posts: 1,570	Unexpected res test results We ran some tests with an unusual test chart, well unusual for testing video cameras. All the results were in line with Adam Wilts conclusions. Except for one small thing. This res chart includes 4K plates. As expected they just appear as grey areas at 50i. In 25p at one point when the camera was moved ever so slightly I can see aliased remnants of the 4K patterns. Would seem to indicate the lens is capable of very high resolution and the OLPF is set high. We all want all the res we can get however I think what I saw might call for a heads up, in some scenarios we might need to wrangle that much resolution to avoid aliasing and/or line twitter. I'll see if I can get permission to post a frame grab if anyone is interested.

February 6th, 2008, 09:21 PM	#2
Steven Thomas Trustee Join Date: Sep 2005 Location: Gilbert, AZ Posts: 1,896	I'm looking forward to seeing it Bob.

February 7th, 2008, 12:14 PM	#3
Tom Roper Trustee Join Date: Jun 2004 Location: Denver, Colorado Posts: 1,891	I LOVE looking at resolution charts! But the eye's perception is the best judge of the overall resolution and picture quality. My judgment is that the process goes awry when you try and attach a number to resolution chart testing. It's just not possible to visually judge when the contrast drops to 50%, or even agree if that's the right number to apply to what you see. Inbetween 10% and 90%, you can observe resolved lines, but one gives you a TV Lines resolved number of 600, the other 1000. Even if you try to be more objective with the process, using resolution testing software like Imatest, and using a MTF50 as a baseline, the number you get depends on what part of the picture you look at, and changes with everything including zoom range, focal length, distance to target, aperture and the amount of in-cam sharpening used. All have an effect on the number of TV lines. In the case of in-cam sharpening, the difference is huge! Imatest software attempts to normalize measurement results by using a 2 pixel sharpening radius as the standard. You have to be a dedicated pixel peeper to discern that visually. What number are you shooting for? A peak number obtained in a limited area of the picture? Or a broader overall number that is averaged over many different parts of the picture, zooms, focal lengths, distance to target and aperture? Of more concern, if someone is attaching a TV lines of resolution number to a photographed chart, HOW did they come up with it? They usually don't go into any detail to describe their methods. Even Adam Wilts gives only vague mention of his methods. My experience with the Imatest software tells me I can come up with many different numbers to describe the same image, just depending on how I define it. And they are all valid numbers for the same image...600,700,800,1000! So I don't preach reliance on attaching numbers to resolution charts. It's more valid to put cameras side by side and obtain results, and let the eye be the judge for perceived differences for images shot under identical test conditions for lighting, lens length, aperture etc.