How to fix error: "Index in position 1 is invalid. Array indices must be positive integers or logical values"

Hello everyone! I am making an automatic software in MATLAB (R2019b) which converts 2D images (e.g. depth maps, reflective maps etc.) into a 3D point cloud matrix. I have been investigating methods of denoising which finds the neighbours of a single point in a rxr kernel and stores their indices. This process is repeated for all points. I can't seem to find a reliable method of scanning the points on the edges as it will return an error that reads "Index in position 1 is invalid. Array indices must be positive integers or logical values." I thought padding the image in zeros would help but that seems to be only a temporary solution because once I increase the radius of the kernel, the same error pops up. If anyone can help solve this I would really appreciate it! I'll attach the code below:

clc; clear all; close all;

% Image and radius are going to be function variables but for testing purposes are set to constants
Image = [21:2:37; 90:98; 31:2:47; 71:79; 53:2:69; 41:49; 21:29; 81:89; 11:19]  %input matrix
Radius = 3; % function variable

r = Radius - 2; % box filter parameter

%Indices of neighbouring points in a moving kernel for point cloud
Image = padarray(Image,[r,r],0,'both');
n = 1; % enables for indice loop
[row,col] = size(Image);
 for i = 2: row -1
        for j = 2: col - 1
          LocationMat=zeros(row,col);
            for iKer = -r/2:r/2
                for jKer = -r/2:r/2
                    LocationMat(i + iKer,j + jKer) = 1 %kernel of rxr
                    Indices(n,:)=find(LocationMat);   %finds the indices of the box kernel
                end
            end
          n = n + 1;   
        end
        
 end

Answer 1

[Cris LaPierre]

You need to incorprate a method that restricts your indices to the resolution of the pixels of your image. I find min and max work well for this purpose. Perhaps something like this

LocationMat(min(row,max(1,i + iKer)),min(col,max(1,j + jKer)))

[Fezan Tabassum]

I implemented what you have suggested but I ran into the this problem in the picture below. The idea was the find the neighbours but it looks like it isn't working properly. Might have to redo the coding on the kernel.

[Cris LaPierre]

Sorry, I got excited fixing a problem you didn't have. The cause is likely the same reason you were getting the error the first time. Your index values are not integers. You can't use 1.5 as an index, for example.

When I=2 and j=2, iKer and jKer both start at -0.5.

[Fezan Tabassum]

so should i replace the -r/2 : r/2 with a different method? How about something like this:

clc; clear all; close all;
% Image and radius are going to be function variables but for testing purposes are set to constants
Image = [21:2:37; 90:98; 31:2:47; 71:79; 53:2:69; 41:49; 21:29; 81:89; 11:19]  %input matrix
Radius = 3; % function variable
r = Radius - 2; % box filter parameter
%Indices of neighbouring points in a moving kernel for point cloud
Image = padarray(Image,[r,r],0,'both');
n = 1; % enables for indice loop
[row,col] = size(Image);
 for i = 2: row -1
        for j = 2: col -1
          LocationMat=zeros(row,col);
          LocationMat(min(row,max(1,i-r:i+r)),min(col,max(1,j-r:j+r)))
          LocationMat(i-r:i+r,j-r:j+r)= 1
          Indices(n,:)=find(LocationMat);   %finds the indices of the box kernel
          n = n + 1;   
        end
        
 end

I replaced the kernel line to better suit the parameters but i still run into the same problem where I cant find the neighbours of the edge points. I can show a more in depth process if there is a way for us to message but if not, Ill post it here just encase.

 Image =

    21    23    25    27    29    31    33    35    37
    90    91    92    93    94    95    96    97    98
    31    33    35    37    39    41    43    45    47
    71    72    73    74    75    76    77    78    79
    53    55    57    59    61    63    65    67    69
    41    42    43    44    45    46    47    48    49
    21    22    23    24    25    26    27    28    29
    81    82    83    84    85    86    87    88    89
    11    12    13    14    15    16    17    18    19

In order for me to get the neighbours of the edge points, I have to pad out the "Image".

    
Image =

     0     0     0     0     0     0     0     0     0     0     0
     0    21    23    25    27    29    31    33    35    37     0
     0    90    91    92    93    94    95    96    97    98     0
     0    31    33    35    37    39    41    43    45    47     0
     0    71    72    73    74    75    76    77    78    79     0
     0    53    55    57    59    61    63    65    67    69     0
     0    41    42    43    44    45    46    47    48    49     0
     0    21    22    23    24    25    26    27    28    29     0
     0    81    82    83    84    85    86    87    88    89     0
     0    11    12    13    14    15    16    17    18    19     0
     0     0     0     0     0     0     0     0     0     0     0

This allows me to include the edge points but using padding causes problems in the denoiser.

     %this part is me changing the numbers manually in text
     Idealy =

     @     1     0     0     0     0     0     0     0     0     0
     1     1     0     0     0     0     0     0     0     0     0
     0     0     0     0     0     0     0     0     0     0     0
     0     0     0     0     0     0     0     0     0     0     0
     0     0     0     0     0     0     0     0     0     0     0
     0     0     0     0     0     0     0     0     0     0     0
     0     0     0     0     0     0     0     0     0     0     0
     0     0     0     0     0     0     0     0     0     0     0
     0     0     0     0     0     0     0     0     0     0     0
     0     0     0     0     0     0     0     0     0     0     0
     0     0     0     0     0     0     0     0     0     0     0

where the @ is the centre point and the 1's are the neighbours. I need a method for the code to notice if the kernel is going past the boundaries of the image and just return a zero or to replicate the matrix to get this values.

[Cris LaPierre]

I am not an expert on what it is you are trying to do, so take that into consideration with my comment. If you use the min/max method, i don't think you need to pad with zeros. The reason you are padding is to acocunt for your kernel.

When you use the min/max, it never extends beyond the edges.

If that ends up not being the problem, I suggest closing this question about the indexing error, and opening a new one about your issue denoising.