image mock-up

In message , fritz_2077
writes
Hi,
I have a requirement to mock up an image based on some MTF data that
has been given to me.

It sounds a bit like a homework or project problem. Though most
university terms have just ended...

In essence what I need to do is to take some imagery from an optical
system that I know the MTF curve for, and change the imagery such that
it looks similar to that produced by a modified version of the
initital system. The latter MTF curve will be of a lower performance
than the first so in practical terms I need to degrade the image.

Can anybody recommend a way to do this? I am relatively new to optics
so the principles are not 100% understood.

My initial idea was to perform a FFT on the image and then adjust the
magnitude values in some way before applying the inverse FFT. I'm not
entirely sure how I would go about doing the adjustment to the
magnitudes but any advice would be gratefully accepted.

Thanks in advance.

You will probably get more complete answers on sci.image.processing

ps. Also if anyone know of any software that could help with this I'd
appreciate knowing what it is and where it ca be obtained from.

There are various ones around. Most accept an image and a point spread
function in real space rather than the MTF. The standard way to do this
sort of thing is with an FFT based convolution code. Some thought is
needed about the edge discontinuities and other implementation details.

Regards,
--
Martin Brown

In message , fritz_2077
writes
Hi,
I have a requirement to mock up an image based on some MTF data that
has been given to me.

It sounds a bit like a homework or project problem. Though most
university terms have just ended...

In essence what I need to do is to take some imagery from an optical
system that I know the MTF curve for, and change the imagery such that
it looks similar to that produced by a modified version of the
initital system. The latter MTF curve will be of a lower performance
than the first so in practical terms I need to degrade the image.

Can anybody recommend a way to do this? I am relatively new to optics
so the principles are not 100% understood.

My initial idea was to perform a FFT on the image and then adjust the
magnitude values in some way before applying the inverse FFT. I'm not
entirely sure how I would go about doing the adjustment to the
magnitudes but any advice would be gratefully accepted.

Thanks in advance.

You will probably get more complete answers on sci.image.processing

ps. Also if anyone know of any software that could help with this I'd
appreciate knowing what it is and where it ca be obtained from.

There are various ones around. Most accept an image and a point spread
function in real space rather than the MTF. The standard way to do this
sort of thing is with an FFT based convolution code. Some thought is
needed about the edge discontinuities and other implementation details.

Regards,
--
Martin Brown

if u can FFT MTF and (TFF) then methinks u hav more skills than ur post
implies

try www.adobe.com

Artie

"fritz_2077" wrote in message
om...
Hi,
I have a requirement to mock up an image based on some MTF data that
has been given to me.

In essence what I need to do is to take some imagery from an optical
system that I know the MTF curve for, and change the imagery such that
it looks similar to that produced by a modified version of the
initital system. The latter MTF curve will be of a lower performance
than the first so in practical terms I need to degrade the image.

Can anybody recommend a way to do this? I am relatively new to optics
so the principles are not 100% understood.

My initial idea was to perform a FFT on the image and then adjust the
magnitude values in some way before applying the inverse FFT. I'm not
entirely sure how I would go about doing the adjustment to the
magnitudes but any advice would be gratefully accepted.

Thanks in advance.

Fritz

ps. Also if anyone know of any software that could help with this I'd
appreciate knowing what it is and where it ca be obtained from.

"fritz_2077" wrote in message
om...
Hi,
I have a requirement to mock up an image based on some MTF data that
has been given to me.

In essence what I need to do is to take some imagery from an optical
system that I know the MTF curve for, and change the imagery such that
it looks similar to that produced by a modified version of the
initital system. The latter MTF curve will be of a lower performance
than the first so in practical terms I need to degrade the image.

Can anybody recommend a way to do this? I am relatively new to optics
so the principles are not 100% understood.

My initial idea was to perform a FFT on the image and then adjust the
magnitude values in some way before applying the inverse FFT. I'm not
entirely sure how I would go about doing the adjustment to the
magnitudes but any advice would be gratefully accepted.

Thanks in advance.

Fritz

ps. Also if anyone know of any software that could help with this I'd
appreciate knowing what it is and where it ca be obtained from.

I don't have a specific answer, but oyu may be able to find it more easily
knowing that the process is called modulation. You are modulating an image
according to the properties of the optical system. The reverse process,
demodulation, is probably more common.

"fritz_2077" wrote in message
om...
Hi,
I have a requirement to mock up an image based on some MTF data that
has been given to me.

In essence what I need to do is to take some imagery from an optical
system that I know the MTF curve for, and change the imagery such that
it looks similar to that produced by a modified version of the
initital system. The latter MTF curve will be of a lower performance
than the first so in practical terms I need to degrade the image.

Can anybody recommend a way to do this? I am relatively new to optics
so the principles are not 100% understood.

My initial idea was to perform a FFT on the image and then adjust the
magnitude values in some way before applying the inverse FFT. I'm not
entirely sure how I would go about doing the adjustment to the
magnitudes but any advice would be gratefully accepted.

Thanks in advance.

Fritz

ps. Also if anyone know of any software that could help with this I'd
appreciate knowing what it is and where it ca be obtained from.

Take a look at the excellent MTF tutorial he
http://www.normankoren.com/Tutorials/MTF.html . It's not very
mathematical, but it contains a wealth of general information that might
give you some hints about how to proceed with your project.

I'd be tempted to try a heuristic approach first, if only to get a ball-park
idea: Take one of those perfect varable spatial-frequency lp charts and
using Photoshop or similar, make a test page of strips at various referenced
contrast levels, print it, then eyeball it for the degraded MTF you're
looking for. Load your image, then adjust its contrast to the eyeballed
reference level and Voila! you've got your result. Maybe. Hopefully.

An FFT approach is outlined on the web site indicated above and is as
follows:
----------------------------------------------------------------------------
------------------

The response of a component or system to a signal in time or space can be
calculated by the following procedure.

1. Convert the signal into frequency domain using a mathematical operation
known as the Fourier transform, which is fast and easy to perform on modern
computers using the FFT ( Fast Fourier Transform) algorithm. The result of
the transform is called the frequency components or FFT of the signal.
Images differ from time functions like sound in that they are two
dimensional. Film has the same MTF in any direction, but not lenses.

2) Multiply the frequency components of the signal by the frequency response
(or MTF) of the component or system.

3) Inverse transform the signal back into time or spatial domain.

Doing this in time or spatial domain requires a cumbersome mathematical
operation called convolution. If you try it, you'll know how the word
"convoluted" originated. And you'll know for sure why frequency domain is
widely appreciated.

--Excerpted from Norman Koren's Photography Page,
http://www.normankoren.com/Tutorials/MTF.html.

----------------------------------------------------------------

(I believe the quoted passage between the lines above constitutes "fair use"
and apologize profusely if such is not the case. The work and text were
produced by Norman Koren, NOT me.)

In the dim, ancient past, in an age before the Great Neuron Depletion, I
used Mathcad's FFT routines for such things. These days,
"Fuh...Fuh....Tea!", is something I diffusely exclaim when I at last
recognize the warm, welcoming beverage sitting in a cup on my tray at the
rest home.

:-)

(fritz_2077) wrote in message
. . .
My initial idea was to perform a FFT on the image and then adjust the
magnitude values in some way before applying the inverse FFT.
. . .

The way you are proposing to do it involves running 2 dimensional
FFT's on the image data. Most photo editors are not going to be able
to do that so you would need special software.

An alternate way would be to divide the new MTF by the original MTF to
get the MTF that represents the change from the old system to the new.
Then take the Fourier transform (1 dimensional) of this MTF to find
the convolution function (blur function) that corresponds to this MTF.
Then you only have to convolve an old image with this convolution
kernel to get a simulated image from the new system. If the old system
and new system MTF's differ only at high frequencies then the
convolution kernel may be small enough to type into a Photoshop custom
filter. In that case you would not need any any special software to
operate on the images. If the MTF's also differ at low frequencies
then the convolution kernel will be large and you will need special
software anyway.

Robert

(fritz_2077) wrote:

My initial idea was to perform a FFT on the image and then adjust the
magnitude values in some way before applying the inverse FFT.

Convolve the point spread function with your image.

www.google.com: MTF PSF OTF
www.google.com: image convolution

(fritz_2077) writes:

My initial idea was to perform a FFT on the image and then adjust the
magnitude values in some way before applying the inverse FFT. I'm not
entirely sure how I would go about doing the adjustment to the
magnitudes but any advice would be gratefully accepted.

You've got a workable idea. The MTF you have has values that range
between 0 and 1 (approximately). All you need to do is, for every FFT
coefficient, figure out what frequency it represents and multiply it by
the value of the MTF at that frequency. When you're done, take the
inverse FFT.

Sounds simple, and it is, but actually making it work requires attention
to details:

- You need to understand what the output format of your FFT software
looks like: where zero frequency is, and what frequency is represented
by an arbitrary location in the array. There are both positive and
negative frequencies present, too.

- FFT coefficients are complex numbers, which you'll be multiplying
by real numbers (MTF value)

- you'll probably have to represent the MTF by a bunch of data points,
and interpolate between those points to get a value for an arbitrary
frequency

- you'll need to map FFT frequencies, which are most easily calculated
in terms of cycles/pixel, into whatever the MTF uses as its domain
(probably cycles/mm) based on some geometric data from the imaging
system.

I'd suggest just playing around with FFTs first. Figure out how to do a
very simple low-pass filter (zero all the coefficients located at a
distance greater than some constant from the origin) before tackling
the simulation of a MTF.

ps. Also if anyone know of any software that could help with this I'd
appreciate knowing what it is and where it ca be obtained from.

A library called fftw is probably the best source for FFT code if you're
writing your own C or C++ code for this.

Dave

HUH??????

"Dave Martindale" wrote in message
...
(fritz_2077) writes:

My initial idea was to perform a FFT on the image and then adjust the
magnitude values in some way before applying the inverse FFT. I'm not
entirely sure how I would go about doing the adjustment to the
magnitudes but any advice would be gratefully accepted.

You've got a workable idea. The MTF you have has values that range
between 0 and 1 (approximately). All you need to do is, for every FFT
coefficient, figure out what frequency it represents and multiply it by
the value of the MTF at that frequency. When you're done, take the
inverse FFT.

Sounds simple, and it is, but actually making it work requires attention
to details:

- You need to understand what the output format of your FFT software
looks like: where zero frequency is, and what frequency is represented
by an arbitrary location in the array. There are both positive and
negative frequencies present, too.

- FFT coefficients are complex numbers, which you'll be multiplying
by real numbers (MTF value)

- you'll probably have to represent the MTF by a bunch of data points,
and interpolate between those points to get a value for an arbitrary
frequency

- you'll need to map FFT frequencies, which are most easily calculated
in terms of cycles/pixel, into whatever the MTF uses as its domain
(probably cycles/mm) based on some geometric data from the imaging
system.

I'd suggest just playing around with FFTs first. Figure out how to do a
very simple low-pass filter (zero all the coefficients located at a
distance greater than some constant from the origin) before tackling
the simulation of a MTF.

ps. Also if anyone know of any software that could help with this I'd
appreciate knowing what it is and where it ca be obtained from.

A library called fftw is probably the best source for FFT code if you're
writing your own C or C++ code for this.

Dave

HUH??????

"Dave Martindale" wrote in message
...
(fritz_2077) writes:

My initial idea was to perform a FFT on the image and then adjust the
magnitude values in some way before applying the inverse FFT. I'm not
entirely sure how I would go about doing the adjustment to the
magnitudes but any advice would be gratefully accepted.

You've got a workable idea. The MTF you have has values that range
between 0 and 1 (approximately). All you need to do is, for every FFT
coefficient, figure out what frequency it represents and multiply it by
the value of the MTF at that frequency. When you're done, take the
inverse FFT.

Sounds simple, and it is, but actually making it work requires attention
to details:

- You need to understand what the output format of your FFT software
looks like: where zero frequency is, and what frequency is represented
by an arbitrary location in the array. There are both positive and
negative frequencies present, too.

- FFT coefficients are complex numbers, which you'll be multiplying
by real numbers (MTF value)

- you'll probably have to represent the MTF by a bunch of data points,
and interpolate between those points to get a value for an arbitrary
frequency

- you'll need to map FFT frequencies, which are most easily calculated
in terms of cycles/pixel, into whatever the MTF uses as its domain
(probably cycles/mm) based on some geometric data from the imaging
system.

I'd suggest just playing around with FFTs first. Figure out how to do a
very simple low-pass filter (zero all the coefficients located at a
distance greater than some constant from the origin) before tackling
the simulation of a MTF.

ps. Also if anyone know of any software that could help with this I'd
appreciate knowing what it is and where it ca be obtained from.

A library called fftw is probably the best source for FFT code if you're
writing your own C or C++ code for this.

Dave