image interpolation using "Sinc"

Share icon

Fri, 02/25/2011 - 07:50 am

Hello,

I can't achieve an image interpolation using the sinc function :
ImageInterpolate  /DEST=$str_Interpolated_spectrum  /FUNC=sinc resample $str_spectre_source

By the way, would you know which sinc fnuction it is ? (sin(x)/x or sin(pi*x)/(pi*x)). In signal processing, we usually use the latter.

Thank you
Hello Nasser,

In general it is helpful if one provides more details than "I can't achieve an image interpolation using the sinc function". In particular it would be useful to know what you are trying to accomplish and what is the nature of your data.

Your command line does not specify ANY transformation. See the /TRNS flag for more information. I suspect that you misunderstood the Resample keyword: It is intended to allow you to correct various aberrations as should be apparent from the choices in the /TRNS flag. Since your command line does not include this flag it is difficult to determine what you want to accomplish.


A.G.
WaveMetrics, Inc.

February 25, 2011 at 10:23 am - Permalink

Hello A.G.,

Thank you for your answer.

My data consists of a 2D spectrum, in other words an image containing bivariate gaussian peaks.
Since the signal was digitized using rectangular pulses, I think it would be nice to interpolate it using the Sinc function. I've used the "Resample" keyword because my aim is to use the "Sinc" function. As you wrote, I may have misunderstood...

Best Regards

February 28, 2011 at 05:54 am - Permalink

Hello Nasser,

As I indicated before, the Resample keyword is intended for correcting optical aberrations so it is not what you want to do.

If you know the form of the PSF for your data set you could try some form of deconvolution. Note that for 1D data we have a demo experiment for spectral deconvolution (see File Menu->Example Experiment->Analysis). I mention the 1D case because depending on how you collected your data it may be possible to deconvolve it on a 1D basis.

A deconvolution of 2D images using the Wiener filter is discussed in the Image Processing Tutorial (File Menu->Example Experiments->Tutorials->Image Processing Tutorial (search for deconvolution).

I hope this helps,

A.G.
WaveMetrics, Inc.

March 1, 2011 at 01:48 pm - Permalink

Hello A.G.,

Quote:
the Resample keyword is intended for correcting optical aberrations so it is not what you want to do.


I suppose it is not reserved for optical signals, but also other 2D signals...?

Quote:
If you know the form of the PSF for your data set you could try some form of deconvolution.


Unfortunately, I don't know the PSF. Our measurement device analyses chemical compounds. The measurements consist of 2D data. One dimension is made of chromatograms, the other one is the result of a kind of spectrometry. Tghat's why I called the resulting image a "spectrum".
Considering that, I think it's quite difficult to determine the PSF...It's a pity, because the Wiener filter seems to be a basic and interesting function in signal processing, even if I've never used it.

As I told you, I was interested in "re-build" the original signal using the SINC function; could you tell me how to do this please ? I think it's the function to employ, because this is the Fourier transform of the boxcar function...(please excuse me if I'm talking rubbish!)

Thanks a lot

Beyt regards,

Nasser




Best Regards

March 2, 2011 at 01:18 am - Permalink

Quote:
I suppose it is not reserved for optical signals, but also other 2D signals...?


There is nothing special about optical signals in this connections other than the fact that the two available functions are scaleShift and radialPoly; both of which are used for correcting images and would not be suitable for your application.

Quote:

Unfortunately, I don't know the PSF. Our measurement device analyses chemical compounds. The measurements consist of 2D data. One dimension is made of chromatograms, the other one is the result of a kind of spectrometry. Tghat's why I called the resulting image a "spectrum".
Considering that, I think it's quite difficult to determine the PSF...It's a pity, because the Wiener filter seems to be a basic and interesting function in signal processing, even if I've never used it.

As I told you, I was interested in "re-build" the original signal using the SINC function; could you tell me how to do this please ? I think it's the function to employ, because this is the Fourier transform of the boxcar function...(please excuse me if I'm talking rubbish!)


It sounds to me as if you should consider one or possibly two one-dimensional de-convolutions along the axes of your data. Note that a deconvolution is difficult when you know the PSF. When you don't know the PSF you need to look for "blind deconvolution", which is even more difficult and more susceptible to error.

There have been threads on deconvolution both here and on the IGOR mailing list. If they are not immediately helpful you may want to try and post a question that describes your data (specifying the generating instrument/method) and what you want to accomplish (e.g., deconvolution, interpolation, etc.) as it is more than likely that a similar problem has been encountered and handled by others. If you are unable to find a satisfactory solution or if you want to solve the problem from scratch, contact support@wavemetrics.com with example data and a precise definition of the problem you are trying to solve.

A.G.
WaveMetrics, Inc.

March 2, 2011 at 10:14 am - Permalink

OK, thank you, I'll send an e-mail to the address you gave me.

Thank you very much.

Best Regards,

Nasser

March 3, 2011 at 08:36 am - Permalink