Applying Movmedian Within Cell Array

Hello,

I have a cell array (2 x 6) called "output", each cell in row #1 {1 -> 6, 2} contains a 1024 x 1024 x 100 matrix. I want to apply movmedian to each cells in row #1. I would like to apply this function in dimension = 3 with window size = 5.

 output = cellfun(@movmedian(5,3), output,'uniform', 0);

This is the code that I have come up with so far, however, it produces an "unbalenced or unexpected parenthesis or bracket" error. I am unsure what is causing this error. I am also somewhat unsure how to instruct matlab to perform this operation only on row 1 of the cell array, please help!

Thank you for your time!!

[Jan]

I still do not understand the notation "each cell in row #1 {1 -> 6, 2}". Please do not invent new terms without explaining it. The "row 1" of the cell C is "C(1, :)" [EDITED, was C(:,1)]. You can assume, that all readers are familiar with the Matlab syntax.

[itend]

Got it. Thank you for the advice!

[Image Analyst]

You say C is an array of cells with 2 rows of 6 columns ("2x6").

C(:,1) is a column vector of the 2 cells in column #

C(1,:) is a row vector of the 6 cells in row #1.

C(expression, 2) is a column vector extracted from only the rows identified in "expression" and only from column #2. So it would have either one cell or two cells since there are only two rows.

Not sure what you mean, but it's clear that reading this FAQ item might help: http://matlab.wikia.com/wiki/FAQ#What_is_a_cell_array.3F

[itend]

Thank you for your response.

I was interested in C(1,:), the row vector of 6 cells in row 1.

Answer 1

[Jan]

output(:, 1) = cellfun(@(c) movmedian(c, 5, 3), output(:, 1), 'uniform', 0);

Remember, that cellfun is very nice, but less efficient than a simply loop in general. Try this:

for k = 1:size(output, 1)
  output{k, 1} = movmedian(output{k, 1}, 5, 3);
end

I prefer to keep it simple stupid, most of all if it takes less runtime in addition.

[itend]

Thank you for your help!

There seems, however, to be an issue... after running your code, it doesn't seem to affect "output". I was expecting the matrices to be changed to 1024 x 1024 x 19. I have attached a pic of my matlab screen for reference.

Also, I apologize if I wasn't clear in my initial post. I am interested in applying movmedian to output(1,:).

What do you think?

[Jan]

@itend: There is no reason for apologies. All question concerning Matlab are welcome in the forum and questions for clarifications belong to the standard for solving problems. The communication about a powerful language like Matlab is not trivial, we know this. :-)

Please prefer to post text as text, not as screenshot. When I open the image, I loose the text I'm currently typing.

The output of movmedian has the same size than the input. I assume that you do not want the "moving", but the "block-wise" median over 5 elements:

for k = 1:size(output, 1)
  C            = output{k, 1};
  C            = reshape(C, [size(C, 1), size(C, 2), 5, size(C, 3) / 5]);
  output{k, 1} = squeeze(median(C, 3));
end

[itend]

Ok! Thank you for your kindness.

For ease of vocabulary, each of the (1024 x 1024) matrices represents an image frame. The 3D array represents 100 image frames taken over time. I thought that the movmedian function would replace a given pixel in each 1024 x 1024 matrix with the median of that pixel taken over 5 frames. Am I incorrect? If so, what is another approach to solving this problem?

Thanks again!!!

[Image Analyst]

Two questions:

1. Why are you using a cell array to store all the frames? I see no need for that.
2. Why aren't you using medfilt2() to do a median filter on your images?

[itend]

1) The idea is that I use a cell array to store each image, then perform operations such as background subtraction to all entries within the cell array. How would you recommend storing/processing the images?

2) I actually figured out how to modify the parameters of movmedian to solve the issue I was having.

Thanks!