Blurry X-rays become clear with new software - tech - 21 December 2007 - New
Scientist
ACCESSED: Mon Mar 16 2009 15:59:13 GMT-0500 (EST)
PAGE URL: http://www.newscientist.com/article/dn13128-blurry-xrays-become-
clear-with-new-software.html
...
They took 15 images made with low doses of X-ray radiation and applied super-
resolution to turn them into a single picture with four times better resolution
than any
of the originals. To produce the same quality image would normally mean using a
third
more radiation than the combined dose of all 15 low-resolution exposures.
Combined
images
...
Super resolution involves taking several low-resolution images by shifting a
camera
slightly each time. Each resulting image is subtly different. These images are
then
automatically aligned by comparing them all, two at a time.
By analysing the way the same features are blurred differently in each
low-quality
image, it is possible to mathematically reverse the blurring effect and make a
higher-
resolution image.
"In practice, most algorithms for alignment are not accurate enough, or are too
computationally intensive," says Sina Farsiu, a member of the team at Duke.
"That's why
they have not yet widely been used in digital cameras."
Farsiu and colleagues devised more efficient algorithms that perform the two
main steps
involved in super-resolution in one go. Instead of comparing images two at a
time, all of
the low-resolution frames are compared at once to create the final image.
The result is more accurate and can process even large X-ray images without
requiring
excessive computing power.
Pier Liugi Dragotti and Loïc Baboulaz at Imperial College London, UK, are also
working
the same problem. Their approach is to model the way a particular camera's lens
projects an image onto its sensor. This helps identify the features in the
pictures, so that
they can be lined up more accurately.
"It is possible to overcome the hardware limitation of the camera and retrieve
some
essential information about the original scene," says Dragotti. For example, he
says, this
makes it possible to resolve the edges of an image as if they had not been
blurred. A
video (top right) shows the results.
To use the technique means first analysing a particular camera's lens, but this
is possible
with a few calibration images. Dragotti and Baboulaz are also working on a way
to skip
the calibration process.
"I think this approach has the potential to be very fast," says Farsiu.
"Ultimately, if we
are use super-resolution in consumer devices, we need to speed it up
dramatically."
Farsiu and colleagues work was presented at the Asilomar Conference on Signals,
Systems, and Computers, in Pacific Grove, California, US in November.
Dragotti and Baboulaz' work was presented at the IEEE International Conference
on
Image Processing, in San Antonio, Texas, US in September.
...
===================================================
-- Absum! --
tOM Trottier, +1 613 860-6633
469 Ancaster Ave, Ottawa, ON K2B 5B6 Canada
http://Information.Architecture.Abacurial.com
Est-ce c'est necessaire d'imprimer ce courriel?
Do you really need to print this email?
--
_________________________________________________________________
Options: http://lists.thomasclausen.net/mailman/listinfo/olympus
Archives: http://lists.thomasclausen.net/mailman/private/olympus/
Themed Olympus Photo Exhibition: http://www.tope.nl/
|