Dividing the image into blocks

I have a image that I want to divide into blocks until each pixel in block have same value. For example in the link below I have an image

http://www.sendspace.com/file/vp3njo

The image consists of waters ,leaves and green parts. I want to keep dividing into blocks until it contains the same value.

Please help.

[Jurgen]

You mean each pixel _exactly _the same? there might not even be such a block in your image. Plus what do you mean with dive into?

[Jan]

I do not see any picture on the linked page, but a lot of advertisments.

[kash]

Assuming the value below taken from image

im =   64    65    66    69    69
       64    66    68    70    72
       64    64    66    67    73
       64    62    64    62    67
       63    60    62    63    66
       64    62    60    63    67
       67    63    61    61    65
       67    68    67    63    63

In this pixel value 64 has 4 values, so it must be in a block, 63 must be block, 64 must be block, last two 67s must be a block and so on. In 4th and 5th columns 2 values of 69s must be grouped, and so on i.e same pixel value must be in a block

[Image Analyst]

I didn't see a picture immediately either. When I clicked on the download link, my company threw up a page that says they've forbidden us to go to that web site. Please pick a site where we can see the image immediately, like tinypic.com or dozens of others.

What do you mean "grouped" or "in a block"? I think of a block as rectangular - what if your contiguous set of gray levels is not rectangular? Can you have blob-shaped "blocks"? If so are you talking about connected components labeling like done with bwlabel and bwconncomp?

Answer 1

[Image Analyst]

kash, I know you've been doing image processing for over a year now and I know you know about bwlabel() and bwconncomp(), so tell me why doing something like

labeled67blobs = bwlabel(grayImage == 67); % Similar for any other number.

is not doing what you want?

[kash]

The image is http://imgur.com/upT3y

i saw tour demo at

http://www.mathworks.in/matlabcentral/answers/46705-how-images-are-divided-into-sub-images

like that i want to divide as in peppers image,

want to divide into sub image such that wach subimage has same pixel value

[kash]

labeled67blobs = bwlabel(grayImage == 67)

does not give answer

Answer 2

[Salma Hassan]

i=imread('image_1_0.png');

i=imresize(i,[256 256]);

   a=64; b=64; %window size

x=size(i,1)/a; y=size(i,2)/b; 

m=a*ones(1,x); n=b*ones(1,y);
C =mat2cell(i,m,n);

C = C(:);

%# display tiles i subplots

      for j=1:a*b
          block_image  = mean2(C{j});
          if block_image > 10 
            imshow( C{j} )

          end
      end

[Walter Roberson]

That will only display the last image for which the mean is large enough. imshow() will be called for the others, but will delete the axes and queue the new image for rendering, but the rendering will not happen until figure() or pause() or uiwait() or waitfor() or drawnow() is called, or until MATLAB returns to the keyboard, and since none of those happen inside the loop, each previous request to render will be zapped by the axes deletion that imshow() does, leaving only the final request to be rendered.

Using image() instead of imshow() would not delete the axes, but since hold is not on, image() would replace the content of the axes, and the visual result would be pretty much the same, that only the last of the requests would update the screen.

[Image Analyst]

When the poster says "I want to divide into blocks until each pixel in block have same value." he's basically describing either quadtree decomposition, done by function qtdecomp(), or superpixels, done by the function superpixels(). The code in this answer just divides into rectangular blocks of all the same size, and each block does not necessarily have all the same value.

[djeffal khaled]

what is the n variable?

[suha athab]

Your code is very useful for me, but I wonder, if I can display only the block with maximum mean value, what will I do? Please answer me.

[Image Analyst]

It is not necessary that you do ONLY that. You can do whatever you want. You can keep track of the means for all block and display only the single block that has the highest mean if you want. See my attached blockproc demos.