How to process each non-overlapped block of an image after it's sub-division?

Question

Mohsin Shah 2018-3-14

0
链接

此问题的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/388148-how-to-process-each-non-overlapped-block-of-an-image-after-it-s-sub-division

评论： Image Analyst 2018-3-14

My function sub-divides an image into non-overlapping blocks (each block is a cell matrix). I wonder how to process each block and then recombine them to make a processed image? Let's say I want to apply fast Walsh Hadamard transform (fwht2) on each block. Here is my function:

if true
  function Blocks = imageBlock(I)
[m,n] = size(I);
Blocks = cell(m/8,n/8);
counti = 0;
for i = 1:8:m-7
 counti = counti + 1;
 countj = 0;
 for j = 1:8:n-7
      countj = countj + 1;
      Blocks{counti,countj} = I(i:i+7,j:j+7);
 end
end
end

0 个评论
显示 -2更早的评论隐藏 -2更早的评论

请先登录，再进行评论。

请先登录，再回答此问题。

Answer 1

Walter Roberson 2018-3-14

1
链接

此回答的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/388148-how-to-process-each-non-overlapped-block-of-an-image-after-it-s-sub-division#answer_309926

cellfun() to process them. cell2mat() to recombine.

Have you considered using blockproc() instead of the steps you are taking now?

1 个评论
显示 -1更早的评论隐藏 -1更早的评论

Mohsin Shah 2018-3-14

在 MATLAB Online 中打开

Yes, I did consider blockproc() but I am making a mistake in using blockproc().

if true
 I  = imread('lena.jpg');
fh = @fwht2d;
Ip = blcokproc(I, [8, 8], fh);
end 
where fwht2d is a user-defined function. This code gives the following error: "BLOCKPROC encountered an error while evaluating the user-supplied function handle, FUN.
The cause of the error was:
Conversion to double from struct is not possible."

请先登录，再进行评论。

Answer 2

Image Analyst 2018-3-14

2
链接

此回答的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/388148-how-to-process-each-non-overlapped-block-of-an-image-after-it-s-sub-division#answer_309927

Use blockproc() and give it your function. It can be whatever you want. See attached examples.

4 个评论
显示 2更早的评论隐藏 2更早的评论

Mohsin Shah 2018-3-14

编辑：Mohsin Shah 2018-3-14

It worked. Thank you, Walter Roberson. I wonder which answer to accept and which to vote. Both answers (from Image Analyst and Walter Roberson) have solved my problem.