| Seeking specific graphic filter - Any help? [message #46736] |
Sun, 11 December 2005 23:58 |
Bernhard Schaffer
Messages: 1
Registered: December 2005
|
Junior Member |
|
|
Hi!
This is not directly an IDL question, although I want to implement the
filter in IDL later on (and may then ask again :c))
I just post it here, because I have the hope that among the audience some
"graphic experts" might be listening,
who can point me in the proper direction. I am looking for a good image
filter or image processing routine for my
image data.
The aim of the images is to identify phases. The image data I get has the
following "specialties" plus "additional
knowledge" to be used:
1.) - The "(physical) source" from which my image data is generated consists
of two or more phases with more
or less sharp interfaces to each other. So, despite a short translation
width of a few pixels, each point in the image
is either of the phases.
2.) - The images start up with "zero-counts" in each pixel. The, depending
on the physical source, individual pixels
get increased values. To clarify this: say the upper half of an images is
Phase A, and the lower one Phase B. The
image acquisition starts with 0 everywhere. Now, step-by-step, one (random)
pixel is chosen. If this pixel is within the
upper half, the pixel-value is increased by 1. If in the lower, nothing
happens. (The image is going to "map" Phase A)
After a while I will have an image with "no" values in the lower half, and
several bright pixels in the upper half. (But also
several "dark" pixels in between.)
3.) - The phases have a "concentration" too, on which the value of the
pixels will be further increased. Following the
picture from above, we would still increase "nothing" in the lower half, but
plus "A1" in the upper half, whereas "A1" stands for the
concentration of the phase A in this pixel. After a while we would therefore
create an image which is "dark" in the lower half, but
has several bright pixels in the upper half. However, some regions in the
upper half will hold "brighter" pixels than others.
4.) - Unfortunately, the image will also contain additional "noise"-pixels.
Say, to a certain extent, a random pixel will increase its value by 1,
independently of where on the map it sits.
All together we end up with an image which is "zero" in general, has some
individual "brighter (noise) pixels" all over, and has
more bright pixels in the upper half (more dense) as well as brighter pixels
there (more intensity). Within the upper half, there will be
some regions with in general "brighter" pixels. We, additionally, have the
knowledge, that the interface between "Phase A, where pixels should have
intensity, depending on the concentration of Phase A" and "Phase B, where no
pixels should show any intensity" is rather sharp.
The filter I am seeking should provide me with the following:
An image with a "smooth" intensity-variation, but outlining Phase A/Phase B
SHARPLY. The "smooth" variation within Phase A should
represent the "concentration" of Phase A.
Now the problem I get with "standard" Low-Pass filters is, that I get my
"smooth" variation in Phase A, but I will loose my "sharp" interface in
between the phases. However, I need to "segmentize" the image into the two
phases (currently done by thresholding, therefore I need the smooth
variation within Phase A).
I created three sample image to demonstrate the problem:
http://bejoscha.tavernmaker.de/tempstore/demo.htm
Any help is appreciated!
regards,
Bernhard Schaffer
|
|
|
|
comp.lang.idl-pvwave archive
Members
Search
Help
Login
Home
