My task
1. I have a haar filter such as [1][-1]
2. I have an image such as
- - [a][b][c][d]
- - [e][f][g][h]
3. I want to perform a wavelet transform on the image and have the following result:
- - Step 1:
- - - [mean(a,b)][mean(c,d)][a-b][c-d]
- - - [mean(e,f)][mean(g,h)][e-f][g-h]
- - Rename: --- just for a clear presentation
- - - [a'][b'][c'][d']
- - - [e'][f'][g'][h']
- - Step 2:
- - - [mean(a',b')][a'-b'][c'][d']
- - - [mean(e',f')][e'-f'][g'][h']
4. The result of Step 2 is what I want.
Question
I found that there is a toolbox in Matlab that is about wavelet application.
I used this command [cA,cH,cV,cD] = dwt2(I,'haar'); The output cV is close but not what I want.
So I would like to ask that how can I perform my task in Matlab in a simply way which means that I can use some built in tools to perform this task.
My way
function haar_wavelet(input_im)
I = imread(input_im);
I = rgb2gray(I);
I = imresize(I,[512 512]);
[I_row,I_col]=size(I);
if(mod(I_col,2))
I_col=I_col-1;
end
haar_mask = make_haar(I_row,I_col);
I=double(I);
new_M = I(:,1:I_col);
j=2;
i=I_col;
while(I_col/j ~= 2)
i = [i I_col/j];
j = j*2;
end
for k = i
new_M(:,1:k) = my_haar_trans(new_M(:,1:k),haar_mask(:,1:k));
end
figure
imshow(new_M)
end
function [output_haar] = make_haar(row,col)
output_haar=ones(row,col)*2;
for i = 2:2:col
output_haar(:,i)=output_haar(:,i)*0;
end
output_haar=output_haar-1;
end
function [output_M] = my_haar_trans(target_M,mask)
temp_diff=target_M.*mask;
[diff_row,diff_col]=size(temp_diff);
diff_M=zeros(diff_row,diff_col/2);
for i = 1:diff_col/2
diff_M(:,i)=temp_diff(:,2*i-1)+temp_diff(:,2*i);
end
mean_M=zeros(diff_row,diff_col/2);
for i = 1:diff_col/2
mean_M(:,i)=target_M(:,2*i-1)+target_M(:,2*i);
end
output_M=[mean_M diff_M];
end
This is what I am doing, is there a easier way?
