Transpose does not support N-D arrays error

Hi, I am performing image morphing using interp2, however i keep on getting the error :

Error using .'

TRANSPOSE does not support N-D arrays. Use

PAGETRANSPOSE/PAGECTRANSPOSE to transpose pages or

PERMUTE to reorder dimensions of N-D arrays.

Any idea how should I modify my code so that I can get the dimension right ?

A = imread('1.png');
B = imread('3.jpeg');

[height,width,channels] = size(A);
[X,Y] = meshgrid(1:width, 1:height);
number_of_frames = 10;

for i=1:number_of_frames
    interp_x = (i - 1) / (number_of_frames - 1);
    interp_y = (number_of_frames - i) / (number_of_frames - 1);
    X_interp = interp_x * X + (1-interp_x) * Y;
    Y_interp = interp_y * Y + (1-interp_y) * X;
   
    
    results = interp2(A,X_interp,Y_interp,'linear',0);
    imshow(results)
    imwrite(result, sprintf('morph_%03d.jpg', i));
    
    pause(0.1);
end

[Stephen23]

Please show us the complete error message. This means all of the red text.

The code you show does not incude the line where the error occurs.

[Christopher Gan]

Error using .'

TRANSPOSE does not support N-D arrays. Use PAGETRANSPOSE/PAGECTRANSPOSE to transpose pages or PERMUTE to reorder

dimensions of N-D arrays.

Error in interp2 (line 105)

V = V.';

Error in imagemorph (line 25)

result = interp2(image1, X_interp, Y_interp, 'linear', 0);

Hi, the errors are the lines above.

Answer 1

[DGM]

There are a couple problems. First, the error is because you're feeding a MxNx3 array to interp2(). You're also feeding an integer-class array to interp2().

A = imread('peppers.png');
%B = imread('3.jpeg');

A = im2double(A); % images fed to interp2() must be floating-point

[height,width,channels,~] = size(A);
[X,Y] = meshgrid(1:width, 1:height);

number_of_frames = 10;
results = zeros(height,width,channels); % preallocate
for i=1:number_of_frames
    interp_x = (i - 1) / (number_of_frames - 1);
    interp_y = (number_of_frames - i) / (number_of_frames - 1);
    X_interp = interp_x * X + (1-interp_x) * Y;
    Y_interp = interp_y * Y + (1-interp_y) * X;
    
    % one way is to just process it one channel at a time
    for c = 1:channels
        results(:,:,c) = interp2(A(:,:,c),X_interp,Y_interp,'linear',0);
    end
    
    imshow(results)
    %imwrite(result, sprintf('morph_%03d.jpg', i));
    
    pause(0.1);
end

I don't know what the rest is actually supposed to do, but it does something.

[Christopher Gan]

Thanks for the help ! it somehow works, any idea on how i could make the transition quicker?

[DGM]

If you want the frames to display quicker, you can reduce the pause, but if that's not enough, then it's probably a good idea to rearrange things:

A = imread('peppers.png');
A = im2double(A); % images fed to interp2() must be floating-point

[height,width,channels,~] = size(A);
[X,Y] = meshgrid(1:width, 1:height);

% generate all the frames
frames = 10;
thisframe = zeros(height,width,channels);
allframes = cell(frames,1);
for f = 1:frames
    interp_x = (f - 1) / (frames - 1);
    interp_y = (frames - f) / (frames - 1);
    X_interp = interp_x * X + (1-interp_x) * Y;
    Y_interp = interp_y * Y + (1-interp_y) * X;
    
    % one way is to just process it one channel at a time
    for c = 1:channels
        thisframe(:,:,c) = interp2(A(:,:,c),X_interp,Y_interp,'linear',0);
    end
    allframes{f} = thisframe;
    
    % if you want to write outputs, they can be done here
    %imwrite(thisframe, sprintf('morph_%03d.jpg', i));
end

% addressing a cell array may be slightly faster than a giant 4D array
% for faster display, keep everything else out of the display loop
% this includes all the axes setup that imshow does when called
hi = imshow(zeros(height,width,3)); % set up the axes with a dummy image
for f = 1:frames
    hi.CData = allframes{f}; % then update the image object's content
    pause(0.01);
end

At least on my machine, with this image, these changes reduced the execution time of the loop contents (excluding the pause) by about a factor of 500.

[Christopher Gan]

Yeah it does display a lot of quicker! thank you so much man! :D

Answer 2

[Steven Lord]

I'd probably adapt the "Interpolate Multiple Sets of Values on Same Grid" example from the griddedInterpolant documentation page. This uses the functionality introduced in release R2021a to interpolate multiple data sets simultaneously. See the Version History section of the documentation page linked above for more information.

A = imread('peppers.png');
A = im2double(A);
whos

  Name          Size                 Bytes  Class     Attributes

  A           384x512x3            4718592  double              
  cmdout        1x33                    66  char                

Let's interpolate A to a coarser grid.

[x, y] = ndgrid(1:384, 1:512);
G = griddedInterpolant(x, y, A);
[xx, yy] = ndgrid(1:20:384, 1:20:512);
B = G(xx, yy);

Now display the two images.

subplot(1, 2, 1)
imshow(A)
title('Original')
subplot(1, 2, 2)
imshow(B)
title('Interpolated')