I am trying different sizes of the kern and different filters.

I created a form to show the images and different filtering and various
results of finding the object.
I try it out with finding patterns equal to different numbers and
combinations of numbers as well as the actual kern without filter.

It takes time to do the manual checking to see what settings give best
results and are accurate.


2007/3/14, Oleg Kobchenko <[EMAIL PROTECTED]>:

The listed image formats should not produce any
difference since they all are lossless compressions.
You can verify creating images in different formats,
loading in J and then subtracting the values.

Regarding image recognition, although the simple difference minimum
method may work, it is not very stable and depends on
accuracy of defining the kernel and the chosen threshold,
which is different for every instance.

It would be nice to see how other methods fair,
such as cross-correlation, as listed by Devon McCormick

http://www.jsoftware.com/pipermail/general/2007-March/029123.html

You can go iteratively, tweaking a working algorithm
with different ideas to see what's improving.
For example, I like the idea of normalization (first link)

http://www.owlnet.rice.edu/~elec431/projects96/pictomaniacs/previous.html

You can apply it to the method you use now to see if it will
eliminate the mismatch.


--- Bj�rn Helgason <[EMAIL PROTECTED]> wrote:

> When extracting an image to work with from a Dicom file I have three
options
>
> dcm2pnm  -  Convert  DICOM images to PPM/PGM, PNG, TIFF or BMP
>
>          +obt  --write-24-bit-bmp
>                  write 24-bit truecolor BMP
>
>          +ot   --write-tiff
>                  write 8-bit (monochrome) or 24-bit (color) TIFF
>
>          +on   --write-png
>                  write 8-bit (monochrome) or 24-bit (color) PNG
>
>
> I have up to now used bmp and am reasonably happy with the results.
>
> I am wondering if the other forms TIFF or PNG would give me better
results.
> Would I need different technique to find the object I am looking for in
the
> image?
>
> The method Oleg sent in earlier finds the object most of the time in the
BMP
> images.
> Actually only found one image so far where it missed the object and gave
> another fairly similar looking location.





____________________________________________________________________________________
We won't tell. Get more on shows you hate to love
(and love to hate): Yahoo! TV's Guilty Pleasures list.
http://tv.yahoo.com/collections/265
----------------------------------------------------------------------
For information about J forums see http://www.jsoftware.com/forums.htm




--
Björn Helgason, Verkfræðingur
Fugl&Fiskur ehf, Þerneyjarsund 23, Box 127
801 Grímsnes ,t-póst: [EMAIL PROTECTED]
Skype: gosiminn, gsm: +3546985532
Landslags og skrúðgarðagerð, gröfuþjónusta
http://groups.google.com/group/J-Programming


Tæknikunnátta höndlar hið flókna, sköpunargáfa er meistari einfaldleikans

góður kennari getur stigið á tær án þess að glansinn fari af skónum
         /|_      .-----------------------------------.
        ,'  .\  /  | Með léttri lund verður        |
    ,--'    _,'   | Dagurinn í dag                     |
   /       /       | Enn betri en gærdagurinn  |
  (   -.  |        `-----------------------------------'
  |     ) |        (\_ _/)
 (`-.  '--.)       (='.'=)
  `. )----'        (")_(")
----------------------------------------------------------------------
For information about J forums see http://www.jsoftware.com/forums.htm

Reply via email to