Pages: [1]   Go Down

Author Topic: MTF is not always interpreted correctly.  (Read 823 times)

joofa

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 544
Logged
Joofa
http://www.djjoofa.com
Download Photoshop and After Effects plugins

Doug Gray

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 1381
Re: MTF is not always interpreted correctly.
« Reply #1 on: February 24, 2018, 12:58:49 AM »

https://www.dpreview.com/forums/post/60804614

I prefer using a set of superimposed, equal amplitude, sine waves that are arranged in phase offsets such that they have the largest possible rms magnitude relative to their max-min luminance (measured linearly).This provides low noise estimates of MTF but only at the specific points as defined by the sine wavelengths. It has several advantages. As long as the sine waves are below Nyquist there is no variation at all from aliasing and bars will alias. There are issues of calibration since printing is intrinsically a low pass function but there are ways to measure and cancel that.

A side benefit is chromatic aberration measurement from the red/blue displacements relative to green.
Logged

joofa

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 544
Re: MTF is not always interpreted correctly.
« Reply #2 on: February 26, 2018, 05:26:42 PM »

I prefer using a set of superimposed, equal amplitude, sine waves that are arranged in phase offsets such that they have the largest possible rms magnitude relative to their max-min luminance (measured linearly).This provides low noise estimates of MTF but only at the specific points as defined by the sine wavelengths. It has several advantages. As long as the sine waves are below Nyquist there is no variation at all from aliasing and bars will alias. There are issues of calibration since printing is intrinsically a low pass function but there are ways to measure and cancel that.

A side benefit is chromatic aberration measurement from the red/blue displacements relative to green.

There is this method to get MTF directly from bars without the CTF / MTF hassle:
https://www.osapublishing.org/ao/abstract.cfm?uri=ao-34-34-8050

Logged
Joofa
http://www.djjoofa.com
Download Photoshop and After Effects plugins

Tim Lookingbill

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 2429
Re: MTF is not always interpreted correctly.
« Reply #3 on: February 26, 2018, 08:23:52 PM »

Is there any practical application to this topic? If so, can you show us in the form of a photograph of a scene?

As an aside I wonder where post processing affects the results of a lens's MTF.

Below is a demonstration of a well known post processing visual effect using Adobe Camera Raw's linear Exposure adjust vs compressed highlights using Brightness in PV2010. Of course PV2012 moved the compression function to Exposure and placed the Linear function in the Whites slider only if the Highlights slider is reduced significantly.

But still the image below shows a significant increase in sharpness and local contrast.

How does MTF of a lens play a part in increasing or decreasing this desired linear effect?

Logged

Doug Gray

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 1381
Re: MTF is not always interpreted correctly.
« Reply #4 on: February 26, 2018, 09:41:43 PM »

There is this method to get MTF directly from bars without the CTF / MTF hassle:
https://www.osapublishing.org/ao/abstract.cfm?uri=ao-34-34-8050

Interesting. Unfortunately is been some years since I had easy access to journals.

The following sequence of points, sampled at Nyquist and derived from a maximal length gf extension field, has the interesting property that it produces 7, nicely spaced frequencies each of the same amplitude. And they have the maximum RMS magnitude relative to peak to peak magnitude. The maximum freq. component is at 7/15 of sample freq. The individual components are 1/15, 2/15, ... 7/15 excluding the DC, fixed offset.

1 1 0 0 0 1 0 0 1 1 0 1 0 1 1 .... repeated

What I like about it is that MTF computations are straightforward and not sensitive to aliasing at all. It does require a good, linear printer and is part of the reason I'm fairly picky about RelCol profile accuracy.

I used the 7 points of MTF from this in a formula to determine imager chip/lens registration flatness for QA purposes in a camera company I was once involved with.
Logged
Pages: [1]   Go Up