Prev: Google Adsense Account Approval Tricks
Next: Problem with real time Multicore
From: asaf on 13 May 2010 10:06
Hi,
My name is Asaf and I am looking for en intelligent, one line without use of loops, solution to the following problem:

I have two matrixes:
First (A) is en image, i.e. 2D matrix of doubles.
Second (B) is logical, same size matrix, indicating "true" for bad pixels of the image.

I want to perform next neighbor correction ( 8 neighbor average) for the bad pixels.

Thanks in advance
Asaf
From: us on 13 May 2010 10:13
"asaf " <asafdidi(a)yahoo.com> wrote in message <hsh10c$f3u$1(a)fred.mathworks.com>...
> Hi,
> My name is Asaf and I am looking for en intelligent, one line without use of loops, solution to the following problem:
>
> I have two matrixes:
> First (A) is en image, i.e. 2D matrix of doubles.
> Second (B) is logical, same size matrix, indicating "true" for bad pixels of the image.
>
> I want to perform next neighbor correction ( 8 neighbor average) for the bad pixels.
>
> Thanks in advance
> Asaf

what have YOU done so far to solve YOUR particular problem...

us
From: ImageAnalyst on 13 May 2010 10:26
Asaf
First do a median filter. Then replace the bad pixels you identified
with the median image, something like this (untested)

medianFilteredImage = medfilt2(inputImageArray);
fixedImage = inputImageArray; % Initialize
fixedImage(badPixels) = medianFilteredImage (badPixels); % badPixels
is your 2D logical array.


From: Matt J on 13 May 2010 10:27
"asaf " <asafdidi(a)yahoo.com> wrote in message <hsh10c$f3u$1(a)fred.mathworks.com>...
> Hi,
> My name is Asaf and I am looking for en intelligent, one line without use of loops, solution to the following problem:
>
> I have two matrixes:
> First (A) is en image, i.e. 2D matrix of doubles.
> Second (B) is logical, same size matrix, indicating "true" for bad pixels of the image.
>
> I want to perform next neighbor correction ( 8 neighbor average) for the bad pixels.
===============

What if some or all 8 neighbors are also bad pixels? Also isn't taking the 8 neighbor median usually prefered to the 8 neighbor average?

In any case, you can generate a table of linear index coordinates of all bad pixels and their neighbors without for-loops as follows

increm=bsxfun(@plus,(-1:1).',(-1:1)*size(B,2));
Coordinates=bsxfun(@plus,find(B),increm.'); %each row is one bad pixel neighborhood

The remaining processing, I leave to you. In particular, the above gives invalid (negative) coordinates for pixels at the boundary of the image, where fewer than 8 neighbors are present. You'll have to decide how you want to process those.
From: Gilad Asseraf on 13 May 2010 10:32
Hi,

You can try this:
--------------------
you will get vector specifying all location of "true" values.
then apply for matrix A, a gaussian filter with 8 neighbors for each one of the locations found in the vector.
This should "land" the bad pixels to values of neighbors

suggested code (I didn't tried it):
----------------------------------------
v=find(B == 1); %(vector specifing the "true" locations)
h=fspecial('gaussian', [3,3],1); %(will create gaussian filter with 8 neighbors, with sigma=1)
imfilter(A(ind2sub(v(:))), h);


Gilad.


"asaf " <asafdidi(a)yahoo.com> wrote in message <hsh10c$f3u$1(a)fred.mathworks.com>...
> Hi,
> My name is Asaf and I am looking for en intelligent, one line without use of loops, solution to the following problem:
>
> I have two matrixes:
> First (A) is en image, i.e. 2D matrix of doubles.
> Second (B) is logical, same size matrix, indicating "true" for bad pixels of the image.
>
> I want to perform next neighbor correction ( 8 neighbor average) for the bad pixels.
>
> Thanks in advance
> Asaf
 |  Next  |  Last
Pages: 1 2
Prev: Google Adsense Account Approval Tricks
Next: Problem with real time Multicore