MATLAB vectorize

问题:

I was wondering if anyone could help me vectorize this piece of code.

fr_bw is a matrix.

 for i=1:height
   for j=1:width      
                [min_w, min_w_index] = min(w(i,j,:));  
                mean(i,j,min_w_index) = double(fr_bw(i,j));
                sd(i,j,min_w_index) = sd_init;
     end
end

回答1:

I can't help you with this sif (match == 0) stuff -- if it's supposed to be if (match == 0) you're not changing match so it could be brought outside the loop.

Otherwise, how about this:

[min_w, min_w_index] = min(w, [], 3);
r = repmat((1:height)',1,width);
c = repmat(1:width,height,1);
ind = sub2ind(size(w),r(:),c(:),min_w_index(:));
w_mean(ind) = double(fr_bw);
w_sd(ind) = repmat(sd_init,height,width);

(Please note that mean is a built-in function so I renamed your variables to w_mean and w_sd.)

The sub2ind call gives you linear indices that correspond to subscripts. (Direct subscripts won't work; z([a1 a2 a3],[b1 b2 b3],[c1 c2 c3]) refers to 27 elements in the z array with subscripts that are the cartesian product of the specified subscripts, rather than z(a1,b1,c1) and z(a2,b2,c2) and z(a3,b3,c3) that you might expect.)

Here's an illustration of this technique:

>> height = 6; width = 4;
>> w = randi(1000,height,width,2)

w(:,:,1) =

   426   599    69   719
   313   471   320   969
   162   696   531   532
   179   700   655   326
   423   639   408   106
    95    34   820   611


w(:,:,2) =

   779   441   638   696
   424   528   958    68
    91   458   241   255
   267   876   677   225
   154   519   290   668
   282   944   672   845

>> [min_w, min_w_index] = min(w, [], 3);
>> min_w_index

min_w_index =

     1     2     1     2
     1     1     1     2
     2     2     2     2
     1     1     1     2
     2     2     2     1
     1     1     2     1

>> z = zeros(height,width,2);
>> r = repmat((1:height)',1,width);
>> c = repmat(1:width,height,1);
>> ind = sub2ind(size(w),r(:),c(:),min_w_index(:));
>> z(ind) = 1

z(:,:,1) =

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


z(:,:,2) =

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

回答2:

A few comments on your code:

1. Did you mean if rather than sif?

2. The code isn't replicable at the moment, since you haven't provided examples of the variables w, fr_bw and sd_init. This makes it tricky to give an exact answer.

3. It looks like you are assigning things to a variable named mean. This will shadow the mean function, and probably cause you grief.

4. I'm just guessing, but I don't think double does what you think it does. You don't need to convert individual elements of a numeric matrix to type double; they are already the correct type. (On the other hand, if fr_bw is a different type, say integers, then you should create a new variable dbl_fr_bw = double(fr_bw); before the loops.

You might need to adjust the dimension over which you calculate the minimums, but the first line of the loop can be replaced with

[min_w, min_w_index] = min(w, [], 3)

The second line with

mean_values(:, :, min_w_index) = double(fr_bw)

Not sure about the third line, since I don't know what sd_init is/does.