doubt regarding bwconncomp example.

This is a small doubt.....I'm stuck in the second example given for bwconncomp by Mathworks. This example removes the group of pixels having largest pixels connected (in an 8 connected neighbourhood):

BW = imread('text.png');
imshow(BW);
CC = bwconncomp(BW);
numPixels = cellfun(@numel,CC.PixelIdxList);
[biggest,idx] = max(numPixels);
BW(CC.PixelIdxList{idx}) = 0;
figure, imshow(BW);

I want to change the code to erase groups of pixels which have between 20 to 70 connected pixels (in their 8 connected neighbourhood). How can I do this?

If I do:

idx = numPixels < 70;

I get an error saying : " Maximum variable size allowed by the program is exceeded."

[Meghana Dinesh]

I also tried using bwlabeln. But how can I remove (i.e. make = 0) those pixels which repeat between 20 to 70 times?

[Walter Roberson]

What is size(numPixels) ?

[Meghana Dinesh]

As an example:

c is a 4X6X3 logical matrix.

   c(:,:,1) =

     0     0     0     0     0     1
     1     0     0     1     0     0
     0     1     1     1     0     0
     0     0     0     0     0     0

   c(:,:,2) =

     0     0     0     0     0     1
     1     0     0     1     0     0
     0     1     1     1     0     0
     0     0     0     0     0     0

   c(:,:,3) =

     0     0     0     0     0     1
     1     0     0     1     0     0
     0     1     1     1     0     0
     0     0     0     0     0     0

   CC = bwconncomp(c);
   numPixels = cellfun(@numel,CC.PixelIdxList);

numPixels has a value [15,3]

and

size(numPixels) = 1 2 ;

Ultimately, my size of c will be around 3000 X 3000 X 3. In this, I want to remove the points (i.e., make their values = NaN) which have less than say 100 number of connected points.

[Walter Roberson]

With those sizes there is no reason why you should get the reported error message for idx = numPixels < 70 . If you had instead received an error message about being out of memory then that would be more understandable. "Maximum variable size allowed by the program is exceeded" is an error message that you would get on 32 bit MATLAB systems if you attempted to create an array that was larger than 2 gigabytes, and is not the error message given if you fill up your available memory.

If you do not need the actual number of pixels, just to know that the number is in the right range, then recall that you can use

   inrange = @(x, a, b) a <= x & x <= b;

   numPixelsInRange = cellfun( @(C) inrange(numel(C), 20, 70), CC.PixelIdxList);

this should save a small amount of memory at the expense of time.

[Meghana Dinesh]

This is what I am doing:

I am first converting my point cloud (in this example, I have created a small 4 X 6 point cloud) into a logical matrix. (all NaNs = 0 while all other finite values = 1).

Here is a screenshot.

I don't seem to face any memory issues as of now.

[Image Analyst]

numPixels<10 produces a binary variable, true or false. It does not produce two numbers. There are not two numbers to try to stick into biggest, and idx like you're trying to do.

[Walter Roberson]

It looks like you are trying to find a maximum, but it is not clear what you are trying to find a maximum of.

idx = 20 <= numPixels & numPixels <= 70;
valididx = find(idx);
validsizes = numPixels(idx);

if isempty(valididx)
  lastvalid = [];
else
  lastvalid = valididx(end);
end

[sizelargest, idxlargest] = max(validsizes);
ccidxlargest = valididx(idxlargest);

lastvalid would be for the case where you want to find the index (into CC) of the last cluster that is in the size range -- the maximum being taken over the indices

ccidxlargest would be for the case whee you want to find the index (into CC) of the largest cluster that is left after you isolate down to that range of sizes -- the maximum being taken over the remaining sizes.

[Meghana Dinesh]

Maybe my approach is wrong.

This is my goal: I have a pointCloud (a 3D matrix including Inf, NaN and finite values). I want to remove the groups of all those points ("remove" i.e., make them = NaN) which have lesser than N (say, 100) points connected to it (in a 26-connectivity neighbourhood).

In order to do this, I am performing the steps mentioned above (converting all finite values to logical and applying bwconncomp).

Are my steps right?

Answer 1

[Image Analyst]

Use bwareafilt instead.

[Meghana Dinesh]

My data is in 3D.

[Image Analyst]

If you want to " want to remove the groups of all those points ("remove" i.e., make them = NaN) which have lesser than N (say, 100) points connected to it (in a 26-connectivity neighbourhood)." then the function you want to use is definitely 100% bwareaopen(). It's meant for exactly that. No need to get PixelIdxList at all.

Answer 2

[Walter Roberson]

N = 100;

a = imread('text.png');
c = a;
c(~isfinite(c)) = 0;
BW = logical(c);
CC = bwconncomp(BW);
numPixels = cellfun(@numel,CC.PixelIdxList);

idx = numPixels < N;
unwantedCC = CC(idx);
unwantedpixels = vertcat(unwantedCC.PixelIdxList);

b = a;
b(unwantedpixels) = NaN;

[Meghana Dinesh]

I am still getting the same error.

RimPointCloud_filt_lst_sq is a 3D matrix.

[Walter Roberson]

unwantedpixels = vertcat(unwantedCC.PixelIdxList(idx));

[Meghana Dinesh]

The line

 unwantedCC = CC(idx);

gives me the error mentioned above.

So when I use:

 unwantedpixels = vertcat(unwantedCC.PixelIdxList(idx));

I get unwantedCC as undefined.

[Walter Roberson]

Sorry, it is the early hours of the morning here.

Leave out the unwantedCC definition. Use

unwantedpixels = vertcat(CC.PixelIdxList{idx});

[Meghana Dinesh]

:D Nevermind! Anyway, thanks for your help so far.

I am still not able to execute since the variable unwantedpixels is a cell. So the line:

 b(unwantedpixels) = NaN;

gives an error saying: Function 'subsindex' is not defined for values of class 'cell'.

[Walter Roberson]

unwantedpixels = cell2mat(CC.PixelIdxList(idx));