What is the error, attempt to grow an array along ambiguous dimension?

Hello, what is the error attempt to grow an array along ambiguous dimension? And, how would I be able to fix that. The code that gives the error is.

    for j = 1:rep
    canv2 = canv;
    del = j*inc;
    canv2(1+del:row+del,1+del:col+del,:) = cm;
    filter = find(canv2 > 0);
    canv3 = bg; 
    canv3(filter) = canv2(filter); 
    image(canv3)
    pause(0.05)
end

[Matt J]

Incidentally, there's no apparent reason to use the FIND command in this situation. It would be less computation to just do

 filter=(canv2>0);
 canv3(filter) = canv2(filter);

assuming canv3 and canv2 are the same size.

[Billy]

The error apparently involves line of canv3(filter)=canv2(filter); and I tried changing that line of code and the error still came up. I'll try checking the size.

[Jan]

@Billy: Please do not only mention, that you have changed the code, but show us the new code also. We do not have crystal balls (except for Walter, of course).

Answer 1

[Jan]

When canv3 is an array with more than 1 non-singelton dimension (e.g. 2D matrix or 3D), using a linear or logical index for growing cannot be interpreted uniquely anymore:

x = [1,2; 3,4];
x([true, true, true, true, true, true]) = 7;

Does this create the same error message? If so, Matlab cannot decide, if the new array should be [2x3] or [3x2].

1. Avoid using the name "filter" for variables, because it is a built-in function
2. Let canv3 grow explictly at first:

index = (canv2 > 0);  % not "filter"
s = size(canv2);
if any(size(canv3) < s)
  canv3(s(1), s(2), s(3)) = 0;  % Grow
end
canv3(index) = canv2(index);

[Billy]

What this code is supposed to do is translate my image, but I am still getting the error even though I allow canv3 to grow explicitly. Bg is the back ground image and obj is the color array of the object i am moving around. Here's what I have:

 canv = cat(3,pg,pg,pg);
image(canv)
pause(1)
x=900
y=100
obj=canv(1+y : orows+y, 1+x : ocols+x, :);
    switch gs
        % get indices of where obj is on canv
        case 0
            index= find(canv > 0);
        case 255
            index= find(canv < 255);
        otherwise
            disp('can only handle white and black backgrounds')
    end
compImage = bg;
    compImage(index) = canv(index); 
  [row col pg] = size(cm);
inc = 2;
rep = 10;    
for j = 1:rep
    canv2 = canv;
    del = j*inc;
    canv2(1+del:row+del,1+del:col+del,:)=cm;
    index = (canv2 > 0);
    s = size(canv2);
    if any(size(canv3) < s)
  canv3(s(1), s(2), s(3)) = 0;
    end
    canv3 = bg; 
   canv3(index) = canv2(index); 
    image(canv3)
    pause(0.05)
end

[Jan]

Where do you get the error now? Please post a copy of the complete message.

[Billy]

I get the error where canv3(index)=canv2(index).

 Attempt to grow array along ambiguous dimension.

 Error in imageTestTranslate (line 54)
    canv3(index) = canv2(index);

[Matt J]

You cannot use linear indexing or other kinds of 1-dimensional indexing to grow an array that is not a vector.

Answer 2

[Matt J]

Copy/paste the error messages, so we can see what lines generate them and also the output of WHOS so we can see the sizes of all your variables.

My guess would be that it occurs in this line

 canv3(filter) = canv2(filter);

and that it occurs because "filter" contains indices greater than numel(canv3).

[Billy]

Yes, that appears to be true. How could I make the number of elements the same?

numel(filter)
ans =

     3959856

 numel(canv3)

 ans =

     3932160

Answer 3

[Matt J]

If the whole point of all this is to translate an image by an integer amount, just use circshift

 canv3=circshift(canv2,shiftvector) + bg;

or the function below if you want non-circulant shifting.

    function [B,src_indices,dest_indices]=noncircshift(A,offsets)
    %Like circshift, but shifts are not circulant. Missing data are filled with
    %zeros.
    %
    %  [B,src_indices,dest_indices]=noncirchift(A,offsets)
    %
    %B is the resulting array and the other outputs are such that
    %
    %  B(dest_indices{:})=A(src_indices{:})

    siz=size(A);
    N=length(siz);

    if length(offsets)<N
       offsets(N)=0; 
    end

    B=zeros(siz);

    indices=cell(3,N);

    for ii=1:N

        for ss=[1,3]

         idx=(1:siz(ii))+(ss-2)*offsets(ii);

          idx(idx<1)=[];
          idx(idx>siz(ii))=[];

         indices{ss,ii}=idx;

        end
    end

    src_indices=indices(1,:);
    dest_indices=indices(3,:);

    B(dest_indices{:})=A(src_indices{:});