Search the CONFOCAL archive at
http://listserv.acsu.buffalo.edu/cgi-bin/wa?S1=confocal
Hi Sarah,
You are correct that both the image and PSF are varied throughout the
iterations.
In 3D deconvolution you are typically working with a volume of data and
a spatially invariant PSF, such that it does not vary over the volume
being processed. This is necessary because of the use of Fourier
Transforms to efficiently calculate the convolutions required for
processing. In blind deconvolution, the PSF is also modified, however,
it is still invariant over the volume being processed.
The interesting thing is that you don't have to process your entire
volume in one deconvolution operation. You can have the software
subvolume the dataset into smaller chunks and have the blind
deconvolution determine a different PSF for each subvolume, thus
allowing some form of spatial variation over the image. The subvoluming
can be in either the XY plane or in Z depending upon how your software
is setup.
As you suggest, having a spatially varying PSF in Z may allow spherical
aberrations to be compensated for. Subvoluming is an approximation to
the true imaging model if you have a spatially varying PSF. There has
been work done in proper spatially varying PSF deconvolution, often in
2D for astronomical imagery such as that from the Hubble. However, the
computing requirements are significant, even compared to current 3D
deconvolution!
Subvoluming is also beneficial when you dataset is larger than that
which can be processed by your computer system, though you do need to be
careful of any potential blending artifacts that may occur at the
boundaries.
In our experience subvoluming of confocal data works extremely well.
Please contact me directly if you would like any further information.
Best regards,
David
---
David Biggs
Senior Research Scientist
AutoQuant Imaging, Inc.
www.aqi.com
518-283-8392
________________________________
From: Confocal Microscopy List [mailto:[log in to unmask]] On
Behalf Of Locknar, Sarah A
Sent: Monday, March 20, 2006 4:00 PM
To: [log in to unmask]
Subject: Blind deconvolution
Hi everyone-
I have a question about the blind deconvolution method. My
understanding is that both the image and PSF/OTF are adjusted in this
iterative method. My question is whether the PSF/OTF is optimized for
the stack as a whole, or as a function of the z-slice? For instance, as
spherical aberrations increase in deeper optical slices, will the
algorithm adjust the PSF in these sections to compensate, or will it
treat the entire data set as a whole to give an "average" PSF for the
entire volume? I'm teaching a course and am not totally clear on this
point.
Thanks-
Sarah
------------------------------------------------------------------------
---------------
Sarah Locknar, Ph.D.
Director, Neuroscience COBRE Imaging / Physiology Core
College of Medicine, University of Vermont
E015 Given Building
89 Beaumont Ave.
Burlington, VT 05405
802-656-0413
802-656-8704 (fax)
------------------------------------------------------------------------
---------------
|