Compute fun handler for blockproc

Hello, i want to compute a function handler for blockproc() like

fun = @(block_struct) ...
 FD(block_struct.data) * ones(size(block_struct.data));
A = blockproc(I,[16 16],fun)

to process a imagine I and then calculate Box Counting in every blocks. My question is how to compute handler (fun) to do that ? Thanks in advance.

Answer 1

[Walter Roberson]

Make FD a function that accepts one variable that is an array of data, and have it do the box counting on that array.

Just watch out for incomplete blocks: along the bottom edge and the right edge you will get blocks that are not 16 x 16 unless the size of the image happens to be an exact multiple of 16 in each direction.

[Constantin Alin]

FD return a VALUE (D=) for entire image. What i want is to divide image to apply in each block FD. Result of blockproc is a Matrix A with these values calculated Per block.

I=512x512 (2^9)

Algorithm Reference :

https://www.mathworks.com/matlabcentral/fileexchange/58148-hausdorff--box-counting--fractal-dimension-with-multi-resolution-calculation

[Walter Roberson]

fun = @(block_struct) BoxCountfracDim(block_struct.data) * ones(size(block_struct.data));
A = blockproc(I, [16 16], fun);

BoxCountfracDim will be invoked a number of times, each time with an image that is 16 x 16 (or smaller at the edges).

[Constantin Alin]

The code returns a matrix of the same size as the original image (512x512) and returns the same fractal dimension everywhere. A is a matrix 512x512 double.

[Walter Roberson]

fun = @(block_struct) BoxCountfracDim(block_struct.data);
out = blockproc(img, [16 16], fun, 'trim', false);
imagesc(out);

Answer 2

[Constantin Alin]

Thank you so much for your answers. It's work, but in some region of Matrix i have NaN values. Why ?

[Walter Roberson]

BoxCountfracDim produces NaN if all of the entries in the input it is passed are zero. It produces NaN+NaN*i if the sum of the entries is negative.

[Constantin Alin]

Thanks, i understand Now.