Welcome to our community

Be a part of something great, join today!

Log in
  • Hey all, just changed over the backend after 15 years I figured time to give it a bit of an update, its probably gonna be a bit weird for most of you and i am sure there is a few bugs to work out but it should kinda work the same as before... hopefully :)

MTF Curves

D

Dave Gray

Active member
Joined
Nov 14, 2008
Messages
39
Reaction score
0
Points
0
Hey guys,

Basically I'm doing digital cinema camera's as part of my university dissertation and have a question about MTF on RED and I guess digital sensors in general.

Do red sensors have an MTF curve to show the sharpness of the sensor or does it not apply in the same way as it does to film?

Cheers
Dave G
 
Strictly speaking, MTF is the property of a system, not a single element of that system.
 
Yes, the way to measure a relative MTF response would be to take a system through to the point of an identical result. So a given lens mounted on a given 35mm film camera shooting to a given film stock, then developed by a given lab and contact printed by a given lab to a given 35mm film stock. Compare that to the same given lens mounted to a RED and then taken through a specific given post process and scanned out by a given device to the same given 35mm negative to them print to the same 35mm print stock. Then the two prints can be compared to measure the relative MTF curves of each. Note that the camera and its sensor is just one step in the process and the "weakest link in the chain" principle applies.

All of that understood, there are ways to get a fairly good measurement indication of MTF. IMATEST test charts and software are fairly standard in the industry. Shoot the chart and run the file through the software to measure relative resolving power.
 
Cheers for the response guys, very informative.

I was just wondering if it was similar to say Kodak, who publish MTF curves for all their film stocks on the website.
 
Gravy said:
Do red sensors have an MTF curve to show the sharpness of the sensor
Click to expand...

Yes, but the effects of the demosiac algorithm and post processing are tremendous and not as easy to calculate in the form of MTF. The OLPF, too, has its own MTF that contributes to the system, and perhaps even the compression will affect system MTF in some circumstances.

All Bayer cameras have the same sensor MTF relative to their pixel pitch (or Nyquist): the only differentiator is the fill factor and OLPF. You might research the MTF curves given for high end DSLR cameras since they are similar to the RED ONE.

In any case, to calculate the MTF for just the sensor, the general formula is (sin(pi*a*f))/(pi*a*f), where a is pixel width in microns and f is the spatial frequency in cycles/mm. The Nyquist limit for the sensor is the spatial frequency of f=1/(2*p).

To calculate MTF at Nyquist for the simple case of 100% fill factor (a=p), it comes to (sin(pi*a/(2*a)))/(pi*a/(2*a)). Which is the same as (sin(pi/2))/((pi/2). The end result is 64% MTF at Nyquist. You can see you don't even need the pixel pitch to calculate MTF for a given Nyquist. Keep in mind that Nyquist on the RED ONE at 4K is a much higher frequency than 2K and 1080p cameras. So RED ONE has an even higher MTF at the Nyquist of the lower-resolution cameras.
 
I'd be curious to see your test results. My practical experience was that, oddly enough, the 1080P image from the Genesis appears "sharper" than the 4K RGB image generated from 4K RAW REDCODE from the RED ONE.

For example, I had to double the strength of diffusion filters that I used on the Genesis versus the RED ONE when shooting close-ups.

But I also had more problems with moire / aliasing on the Genesis, and very little to none on the RED. I don't know what's going on, but I suspect that the Genesis uses a milder OLPF as a way of maximizing sharpness in a 1080P recording, but at the expense of pushing the limits of aliasing, whereas the RED opted for a heavier OLPF as a way of avoiding aliasing in the three color channels once debayered (plus it is probably harder to design an OLPF for a mosaic Bayer pattern compared to the simpler RGB stripe pattern used by the Genesis).

Perhaps the line resolution limitations of 1080P recording makes aliasing more likely with high resolution sensors, although the 12MP sensor of the Genesis is really optimized for separation into 1080P RGB, so I don't know what the story is...

But it's an interesting issue, what "sharpness" means.

So when comparing the MTF of different cameras as a way of analyzing ability to resolve fine detail, one should somehow also quantify aliasing relative to sharpness.

It seems to me to be a creative judgement call on the part of engineers and designers in terms of how much aliasing to allow relative to how much fine detail they want to record.

One could argue that the smoother, less edge-sharp look from the RED is more like 35mm film than the more crisp, edgy look that the Genesis, F23, and F35 generates.
 
The aliasing you get from no or a weak OLPF add false sharpness that is not coming from properly resolved detail. The aliasing tricks the eye into thinking it's seeing detail when it's seeing false edge information.

The second pieces is that perceived sharpness doesn't necessarily equal resolution. Having enough pixels to properly and smoothly resolve detail in my opinion is better than having too few pixels and a harsh transitions that lead to false representations of detail.
 
I never got why some people touted the f23 and Genesis for their sharper images. I have always hated the sharper HD images as too clinical and electronic.

But Deanan, maybe this is a conversation to have to the side, but you are saying some of that extreme sharpness is a by product of the aliasing. Thats fascinating.

I've noticed that it has also caused a few DP's to question the back focus on the R1 because they are so used to seeing the "HD Pop."

I would not go so far as to call the Genesis's images "weak", but I do in fact like the way RED has chosen to approach imaging.
 
The issue is that often sharpness is equated with resolution and detail. In an image without aliasing, this can be the case, but as soon as you allow in excessive aliasing, sharpness can be a product of the aliasing and hence an aggressive, unwanted artifact. Similarly, sharpness can be detail, or it can be edge enhancement.

It could be that good images are sharp, but not all sharp images are good :-)

Graeme
 
Sure, but perceived sharpness, as opposed to true detail, is still a valid issue to consider when choosing a system. Perceptions mean a lot in filmmaking. Look at the issue in film as to whether the grain pattern is increasing the sensation of detail and sharpness that isn't there actually.

Aliasing itself isn't detail, but fine detail can cause aliasing, so the existence of aliasing suggests an excess of high frequency detail.

My point is that we have to get beyond simplistic notions of what constitutes a "sharp" image with high resolution of detail. With these new digital systems, it seems that there are a lot of "flavors" and looks because different systems have taken different design approaches and have different priorities based on the tastes of the engineers and certain physical limitations of the system. Simply comparing pixel resolution or simply looking at MTF isn't going to describe the whole picture accurately.

Like I said, it seems to me that there has to be a way of describing MTF and detail in relationship to aliasing in the system (and to some extent, aliasing is inherent to most digital images.)
 
If you plot an MTF curve where you point the system at a source of detail so high as it should provoke aliasing, you can see the MTF bottom out, then peak up again, giving you a measure of the false detail the aliasing is adding. If you stop your MTF curve too soon, or don't have a high enough detail in your source, then the end result will be misleading.
 
MTF Curves?

Good read>>>

IMATEST - Do it yourself>>>

"Imatest™ is a suite of programs for measuring the sharpness and image quality of lenses, digital cameras, digitized film images, and prints using inexpensive, widely available targets".
 
I like to use the term "functional resolution."
 
Deanan said:
The aliasing you get from no or a weak OLPF add false sharpness that is not coming from properly resolved detail. The aliasing tricks the eye into thinking it's seeing detail when it's seeing false edge informatio.n
Click to expand...

You can see this really easily in 3D as well. If each pixel is only sampled once (0x AA) The image is razor sharp due to the aliasing. If you sample each pixel 16 times (3x AA) you'll get a much smoother, much 'softer' looking image.

If you use a reyes renderer like renderman you almost by definition get an extremely over sampled image. It's a way you used to be able to really easily spot Renderman footage since it would be notably silkier than a ray tracer.

The real trouble with aliasing is it's just about impossible to remove in post. The irony being that in order to to convert an aliased image into an antialiased image you have to sample down. So if the Genesis hypothetically had no OLPF and was completely aliased at 1080p you would need to cut its resolution down to 540p just to get 1x AA. The sharper image is actually softer when sampled correctly. :D

Moral of the lesson being I would rather use a sharpening process to increase the appearance of sharpness than than have an aliased image.
 
You must log in or register to reply here.
Back
Top