Error : "Assignment has more non-singleton rhs dimensions than non-singleton subscripts"

I am a newbie to Matlab so forgive me if this is a redundant query.Posting here because I still can't find out how to rectify the error. I am trying to convert the decimals to their respective state vectors, on all the planes of an image and store it a matrix. The piece of code is given below.

    img1 = imread('lena.jpg');
a = zeros(256,256,3);
for z = 1:3
    for x=1:256
        for y=1:256
            m = dec2vec(img1(x,y,z),8);
            a(x,y,z)= m;
        end
    end
end

The error am receiving is Assignment has more non-singleton rhs dimensions than non-singleton subscripts

   Error in test1 (line 7)
            a(x,y,z)= m;

Kindly help me out because am unable to figure out where I am going wrong.Thanks a lot.

Answer 1

[Guillaume]

From the name, I'd guess that dec2vec returns a vector. a(x, y, z) is a scalar. You can't obviously squeeze a vector into a single element, hence the error you see. Possibly, you want:

a = cell(256, 256, 3)  %how about giving that variable a better name, such as bitpattern?
for ...
  for ...
     for ...
        m = ...
        a{x, y, z} = m
...

To store your vectors in a cell array, or:

a = zeros(256, 256, 3, 8);
for ...
  ...
          a(x, y, z, :) = m;
 ...

to store your vector in a 4D matrix.

Note that if all you want to do is convert the pixel intensities into a bit pattern, you don't need loops:

img1 = imread('lena.jpg');
bitpattern = reshape(num2cell(dec2bin(img1, 8) - '0', 2), size(img1)) %to get it as a cell array
bitpattern = reshape(permute(dec2bin(img1, 8) - '0', [1 3 4 2]), [size(img1), 8]) %to get it as a 4D array

[Ananya Anurag]

Sir, I tried to store it in a 4-D vector as you said(since that is what I actually wanted),I get the error

Subscripted assignment dimension mismatch.

        Error in test1 (line 7)
                a(x,y,z,:)= m;

[Guillaume]

dec2vec is not a matlab function, so I don't know what it returns. The size of the fourth dimension of a must match the size of the array returned by dec2vec.

I assumed it was 8, obviously it is not.

[Ananya Anurag]

Sir, dec2vec converts a given decimal to its respective state vector. For example

        >> q = dec2vec(5,3)

 q =

     0
     0
     0
     0
     0
     1
     0
     0

what am trying to accomplish is that I am taking an image(256x256x3) and trying to convert each given pixel(in decimals) to its representative vector form and store it in a matrix to perform further operations on it.

[Guillaume]

I have no idea what the "state vector" of a decimal, and I certainly have no idea how you get [0 0 0 0 0 1 0 0] out of (5,3).

As a I said, whatever is returned by dec2vec must have the same numbers of elements as the 4th dimension of a. And it must always be the same numbers of elements regardless of the pixel value. So, you need to change the 8 in

a = zeros(256, 256, 3, 8)

into that number of elements.

Without seeing the code for dec2vec, I can't help you any further.

[Guillaume]

If dec2vec is the same as this, then you need to declare a as:

a = zeros([size(img1),  256]); i.e. 256, 256, 3, 256

and your code will work.

Of course, since your image is basically telling you which index of the 3rd dimension is to be set to one for the correspond (row, column, page) coordinate, you can do this much easier with plain indexing. The 4 lines of code below are all that's needed to obtain the same result:

   img = imread('lena.jpg');
   statevector = zeros([size(img), 256]);
   [row, col, page] = ndgrid(1:size(img, 1), 1:size(img, 2), 1:size(img, 3));
   statevector(sub2ind(size(statevector), row, col, page, img+1)) = 1;