Re: Remove speckles in bilevel (bw) image



Laurent S. wrote:
Thanks for the 'heads-up'. Turkel offers a wealth of
material in .../notes/...

Yes, thanks for the hint.

....
I do not understand exactly what Turkel is proposing here. Is the group 'a' the pixels in the 'X shaped' subarea

aa aa
aa aa
aa
aa aa
aa aa

And is group 'b' similarly the '+ shaped' subarea?

Thats what i think.

If so, we
presumably are required to take the median value of group 'a'
and also that of of group 'b'. But then what's to do?

And then the median from this two values plus the center value
becomes the new value, but i am also not really shure.

This way, the result for bw_04 would be:
http://freenet-homepage.de/JDierks/tmp/bw_04_medhybrid.png

Because it erases very much, here is a version with a center weight
of 3 for Group A and B:
http://freenet-homepage.de/JDierks/tmp/bw_04_medhybrid3Center.png

....
Sounds like the features we want.

Maybe i misinterprated the paper, because it is somewhat unclear,
but i made also a very easy function, that looks in the 8 directions
from the center pixel and counts how many pixels have the same
value as the center and if so, one more pixel in this direction is taken
into account.
If the amount is at least 3, including the center pixel, the value stays,
otherwise it changes (so this function makes only sense for 2 level
images):
http://freenet-homepage.de/JDierks/tmp/bw_04_least3.png And 2 iterations of it:
http://freenet-homepage.de/JDierks/tmp/bw_04_least3_2it.png

For comparison here are the results for a 3x3 Median with a center
weight of 5, that should come close to the gaussian07 weighted
median(?):
http://freenet-homepage.de/JDierks/tmp/bw_04_med9Cen5.png
And 2 iterations:
http://freenet-homepage.de/JDierks/tmp/bw_04_med9Cen5_2it.png

It erases more values from the tips of the letters.

Is this (once understood) trivial to program within MATLAB? As ImageAnalyst's recipe is!

Cant tell you really, because i dont use matlab. But it should be
not very complicated.

Jens
.



Relevant Pages

  • Re: "temperature" layer
    ... weight of each pixel and how it affects its neighbors. ... LockBits for speed and put your legend color in the place of the monochrome ... Answer those GDI+ questions with the GDI+ FAQ ...
    (microsoft.public.dotnet.framework.drawing)
  • Optimization Question
    ... Copy work area 1 to work area 2 offset one pixel right. ... +6.16611296E+002 Iterations per second. ... odd code, even data ...
    (comp.lang.asm.x86)
  • Re: Antialiasing the Mandelbrot Set
    ... Divide the pixel to render in 3x3 sub-pixels. ... of iterations and write that one as the pixel on the screen. ... create the images.) ...
    (sci.fractals)
  • Re: Antialiasing the Mandelbrot Set
    ... Divide the pixel to render in 3x3 sub-pixels. ... of iterations and write that one as the pixel on the screen. ... > By playing with gradients I can get interesting images but I end up ... > images with absolutly no artifacts and I would like to produce high ...
    (sci.fractals)
  • Re: CENTER OF GRAVITY METHOD FOR IMAGE SMOOTHING
    ... To make the matter simple: I have a 100x 100 pixel set ... The reason is that the human eye gives more "weight" to one channel and less to ... You should check "color models" and how to switch from RBG to YUV or CMYK, ... It should be noted that most color models are designed to be as simpel as possible. ...
    (sci.image.processing)