Re: Positive image interpolation

aruzinsky wrote:
"Positivity of course is from physics: sampled images are basically energy
maps, so any interpolation methods returning negative values does not
properly model the image formation process."

Potentially negative values should be the least of your modelling
concerns. Natural images do not follow a purely interpolative model
because
a light sensor's output does not represent the intensity at the sensor's
center. All light sensors integrate intensities across their areas.
Thus, natural images are more accurately modelled by convolution followed
by decimation. See:

http://www.interpolatethis.com/phpBB2/viewtopic.php?t=489&sid=f3b65e6745a427745268b655e84af084

"The specific problem that I am considering is the calculation of the
Fourier coeffcients for frequencies close to the Nyquist rate. It is well
known that in this case the FFT gives very crude approximations."

Please, refresh my memory. What is wrong with the calculation of the
Fourier coeffcients for negative f(x)? How would this differ for f(x) +
c, where c is some small positive constant such that f(x) + c is positive
for all x?

My point is not that it is not possible to computer the Fourier
coefficients for negative f(x).
I would like to understand exactly what makes f(x) negative, and use
this knowledge to
improve the interpolation process. See my reply to Dave, above in this
thread.
Thanks for the link to your blog, it's interesting stuff.

-Arrigo

.

Relevant Pages

  • Re: Best software for increasing resolution?
    ... interpolation' Photoshop action better than these. ... using 16-bit images you're already at the limit of PhotoShop's math-bandwidth. ... 16bit source data for storing images. ...
    (rec.photo.digital)
  • Re: Blackman interpolation instead of linear or cubic
    ... interpolation with Blackman interpolation, ... what we would like to see when filtering or reconstructing images. ... Fourier sense by taking the discrete fourier transform, ... The Blackman function is not meant as a filter in itself. ...
    (comp.graphics.algorithms)
  • Re: Scientifically Based Presharpening for Enlargement
    ... Your starting images should be uncompressed and unsharpened. ... interpolation, which is familiar to many people, and doesn't ... You might address why the sharpening should be a preprocessing step. ... the sharpening is a linear filter, ...
    (sci.image.processing)
  • Re: Convolution-based image interpolation
    ... It is quite common to interpolate images by FIRs. ... interpolation by non-simple integer ratios is needed, ... even be told the name of a single bell-shaped self-overlapping function ...
    (comp.dsp)
  • Re: Positive image interpolation
    ... "Positivity of course is from physics: sampled images are basically energy ... so any interpolation methods returning negative values does not ... properly model the image formation process." ...
    (sci.image.processing)