Problems with coursera image blur matlab problem

Hey all,

I am near completing the introduction to matlab programming course on Coursera.

function[output] = blur(img,w) 
change = w; 
emptymatrix = zeros(length(img(:,1)),length(img(1,:))); 
for z = 1:length(img(1,:))
    for i = 1:length(img(:,1)) 
        if (i + change) <= length(img(:,1)) && (i - change) >= 1 && (z + change) <= length(img(1,:)) && (z - change) >= 1
         submatrix = img((i-change):(i+change),(z-change):(z+change)); 
         emptymatrix(i,z) = uint8(sum(submatrix(:))/numel(submatrix));
         output = uint8(emptymatrix); 
        elseif i == 1 && z == 1
         submatrix = img((i):(i+change),(z):(z+change)); 
         emptymatrix(i,z) = uint8(sum(submatrix(:))/numel(submatrix));
         output = uint8(emptymatrix); 
        elseif i == length(img(:,1)) && z == length(img(1,:)) 
         submatrix = img((i-change):(i),(z-change):(z)); 
         emptymatrix(i,z) = uint8(sum(submatrix(:))/numel(submatrix));
         output = uint8(emptymatrix); 
        elseif i == 1 && z == length(img(1,:)) 
         submatrix = img((i):(i+change),(z-change):(z)); 
         emptymatrix(i,z) = uint8(sum(submatrix(:))/numel(submatrix));
         output = uint8(emptymatrix);  
        elseif i == length(img(:,1)) && z == 1
         submatrix = img((i-change):(i),(z):(z+change)); 
         emptymatrix(i,z) = uint8(sum(submatrix(:))/numel(submatrix));
         output = uint8(emptymatrix); 
        elseif i == 1 && z-change >= 1 && z+change <= length(img(1,:)) 
         submatrix = img((i):(i+change),(z-change):(z+change)); 
         emptymatrix(i,z) = uint8(sum(submatrix(:))/numel(submatrix));
         output = uint8(emptymatrix); 
        elseif i == length(img(:,1)) && z-change >= 1 && z+change <= length(img(1,:)) 
         submatrix = img((i-change):(i),(z-change):(z+change)); 
         emptymatrix(i,z) = uint8(sum(submatrix(:))/numel(submatrix));
         output = uint8(emptymatrix);   
        elseif z == 1 && i-change >= 1 && i+change <= length(img(:,1))
         submatrix = img((i-change):(i+change),(z):(z+change)); 
         emptymatrix(i,z) = uint8(sum(submatrix(:))/numel(submatrix));
         output = uint8(emptymatrix);    
        elseif z == length(img(1,:)) && i-change >= 1 && i+change <= length(img(:,1))
         submatrix = img((i-change):(i+change),(z-change):(z)); 
         emptymatrix(i,z) = uint8(sum(submatrix(:))/numel(submatrix));
         output = uint8(emptymatrix); 
        end
    end
end
output = uint8(output); 
end

I have attached a screenshot of the problem below and of the output the online program is giving me.

Could anyone give me a helping hand as to where I am going wrong?

Answer 1

[Walter Roberson]

I would suggest to you that you could save a lot of code by using max() and min() on the coordinates, like

max(1, column-w):min(column+w, number_of_columns)

That will get you a block of values that you can take the mean of.

[Emre Yavuz]

@ Walter Roberson, thank you for your suggestion although I am not sure as to where to implement that. Do you think that will solve the 'incorrect answer' statement the program is bringing up?

[ey21]

@Walter Robertson, the 2nd part does not seem to be accepting my code (screenshot) whilst the first part does

[Walter Roberson]

emptymatrix = zeros(length(img(:,1)),length(img(1,:))); 

That creates emptymatrix as a double()

emptymatrix(i,z) = uint8(sum(submatrix(:))/numel(submatrix));

The right hand side is uint8() because of the uint8() that you use. The left hand side is double(), so the uint8() value will be converted to double. Have you considered the possibility of initializing emptymatrix as uint8 to avoid that, and to avoid needing to uint8() the right hand side?

output = uint8(emptymatrix); 

Why are you doing that every iteration inside both for loops?

output = uint8(output); 

Under what circumstances would output not already be uint8 after the loops?

[Emre Yavuz]

@ Walter Roberson, I have edited my code below but it is still not accepting...

function[output] = blur(img,w) 
change = w; 
emptymatrix = uint8(zeros(size(img)));  
for z = 1:length(img(1,:))
    for i = 1:length(img(:,1)) 
        if (i + change) <= length(img(:,1)) && (i - change) >= 1 && (z + change) <= length(img(1,:)) && (z - change) >= 1
         submatrix1 = img((i-change):(i+change),(z-change):(z+change)); 
         emptymatrix(i,z) = (sum(submatrix1(:))/numel(submatrix1)); 
        elseif i == 1 && z == 1
         submatrix2 = img((i):(i+change),(z):(z+change)); 
         emptymatrix(i,z) = (sum(submatrix2(:))/numel(submatrix2));
        elseif i == length(img(:,1)) && z == length(img(1,:)) 
         submatrix3 = img((i-change):(i),(z-change):(z)); 
         emptymatrix(i,z) = (sum(submatrix3(:))/numel(submatrix3));
        elseif i == 1 && z == length(img(1,:)) 
         submatrix4 = img((i):(i+change),(z-change):(z)); 
         emptymatrix(i,z) = (sum(submatrix4(:))/numel(submatrix4));  
        elseif i == length(img(:,1)) && z == 1
         submatrix5 = img((i-change):(i),(z):(z+change)); 
         emptymatrix(i,z) = (sum(submatrix5(:))/numel(submatrix5));
        elseif i == 1 && z-change >= 1 && z+change <= length(img(1,:)) 
         submatrix6 = img((i):(i+change),(z-change):(z+change)); 
         emptymatrix(i,z) = (sum(submatrix6(:))/numel(submatrix6));
        elseif i == length(img(:,1)) && z-change >= 1 && z+change <= length(img(1,:)) 
         submatrix7 = img((i-change):(i),(z-change):(z+change)); 
         emptymatrix(i,z) = (sum(submatrix7(:))/numel(submatrix7)); 
        elseif z == 1 && i-change >= 1 && i+change <= length(img(:,1))
         submatrix8 = img((i-change):(i+change),(z):(z+change)); 
         emptymatrix(i,z) = (sum(submatrix8(:))/numel(submatrix8)); 
        elseif z == length(img(1,:)) && i-change >= 1 && i+change <= length(img(:,1))
         submatrix9 = img((i-change):(i+change),(z-change):(z)); 
         emptymatrix(i,z) = (sum(submatrix9(:))/numel(submatrix9));
        end
    end
end
output = emptymatrix; 
end 

I believe I have covered every index in the empty matrix but somehow when I print the emptymatrix at the end, there are some zeros appearing where there should be non-zero numbers.

For example when my input matrix is this:

 225   225   225   225   225
   225     0     0     0   225
   225     0     0     0   225
   225     0     0     0   225
   225   225   225   225   225

[output] = blur(img,2)

Gives me:

  5×5 uint8 matrix

   125     0   135     0   125
     0     0     0     0     0
   135     0   144     0   135
     0     0     0     0     0
   125     0   135     0   125

But I believe it should give me:

  5×5 uint8 matrix

   125   113   135   113   125
   113    98   124    98   113
   135   124   144   124   135
   113    98   124    98   113
   125   113   135   113   125

[Walter Roberson]

You should clean up your code:

[rows, cols, ~] = size(img);

Now use rows instead of length(img(:,1)) and cols instead of length(img(1,:))

And I really do recommend using max() and min(), the code because very easy.

[Emre Yavuz]

@walter roberson, finally managed to solve it using this:

function[output] = blur(img,w) 
[rows, cols, ~] = size(img); 
emptymatrix = uint8(zeros(size(img)));  
for z = 1:cols
    for i = 1:rows
        maxrow = min(i + w,rows); 
        minrow = max(i - w,1); 
        maxcol = min(z + w,cols); 
        mincol = max(z - w,1); 
        if minrow >= 1 && maxrow <= rows && mincol >= 1 && maxcol <= cols
        submatrix1 = img((minrow):(maxrow),(mincol):(maxcol)); 
        emptymatrix(i,z) = (sum(submatrix1(:))/numel(submatrix1)); 
        end
    end
end
output = emptymatrix; 
end 

Thanks so much for your help!

[Walter Roberson]

Could you give an example of a situation in which the if came out false?

[ey21]

@ Walter Roberson, is this a trick question? As in, the if line does not need to be there?

[Walter Roberson]

I can't think of a reason to have it there.

[Emre Yavuz]

@ Walter Roberson, agreed!

Answer 2

[Muhammad Qaisar Ali]

function output = blur(img,w);
    
    [r,c]=size(img);
    output=ones(r,c);
    for ri=1:r
        for ci=1:c
            % checking for indicies,and making limits for sub matrix.
            if ri-w<1
                sub_mat_all_row_indicies=1:(ri+w);
            elseif ri+w>r 
                sub_mat_all_row_indicies=(ri-w):r;
            else
                sub_mat_all_row_indicies=(ri-w):(ri+w);
            end
            
            if ci-w<1
                sub_mat_all_col_indicies=1:(ci+w);
            elseif ci+w>c 
                sub_mat_all_col_indicies=(ci-w):c;
            else            
                sub_mat_all_col_indicies=(ci-w):(ci+w);
            end
            sub_mat=img(sub_mat_all_row_indicies,sub_mat_all_col_indicies);        %make sub matrix/window.
            % bluring or averaging.
            output(ri,ci)=mean(sub_mat(:));
        end
    end
    output=uint8(output);       %converting to grayscle. 0 to 255.
end