how to specify positions of the blocks in an image

i have divided my 64x64 image into overlapping blocks of 2x2. and the block elements are stored in another array. the array contains the first block elements in first row. second block elements in the second row and so on. now can anyone please help me in finding the the starting row and col position of that block?

i used the function im2col to divide my image into blocks. thank you!

Answer 1

[David Young]

Suppose you have

B = im2col(A, [2 2]);

and the size of A is [64 64].

Then B(1,k) comes from A(r,c) where

 c = floor((k-1)/63) + 1
 r = k - (c-1)*63

More generally, if the size of A is [mm nn], and you start from

B = im2col(A, [m n]);

then B(1,k) comes from A(r,c) where

 c = floor((k-1)/(mm-m+1)) + 1
 r = k - (c-1)*(mm-m+1)

[Prajakta Shende]

p=imread('D:\image forgery\db\1aa.jpg');

% imshow(p);

% p=[1,2,3,4;5,6,7,8;9,10,11,12;13,14,15,16];

p=imresize(p,[64 64]);

[m,n]=size(p);

in=input('Enter block size');

disp('Beginning block division using im2col');

c=im2col(p,[in in],'sliding');

c=c';

disp('Block division complete')

[row col]=size(c);

% y=0;

% for i=1:row-in

% for j=1:col-in

% x=0;

% y=y+1;

% pos(y,1)=i;

% pos(y,2)=j;

% end

% end

xyz=0;

for i=1:row

max(c(i,:));

min(c(i,:));

mean(c(i,:));

var(c(i,:));

xyz=xyz+1;

% output(xyz,1)=pos(i);

% output(xyz,2)=pos(j);

%

output(xyz,1)=max(c(i,:));

output(xyz,2)=min(c(i,:));

output(xyz,3)=mean(c(i,:));

output(xyz,4)=var(c(i,:));

end

toc

there is some big problem in the min, max part of the code itself. even if i run it for 16x16 block size it gives toooo many values. which is wrong.

for 16x16 with a blck size of 2 (say) the no of rows would be 9 each containing 4 columns.

and the mean,max,min,var of these rows should be only 9.

but it gives many many values.

can u please tell me the necessary change to be made?

[David Young]

I am sorry, but I don't understand what you're trying to do. It looks like a new question. The answer I gave is to your original question as I understood it, in the simplest form I could.

Perhaps you need to explain what you are trying to achieve as your final result.

Answer 2

[Walter Roberson]

Why should the min or max of the rows be only 9? You are applying the statistics to c(i,:) where c is the output of im2col() applied to p, where p is the resize of the image that was read in. im2col does not return indices, it returns array contents -- so your c is going to have whatever content the image had. Unless the image itself was confined to values in the range 0 to 9, the max() could easily exceed 9.

[Prajakta Shende]

this s respective to a scenario where i have 16x16 sized image. and if the block size is 2 then the resultant im2col output would have 9 rows. since they are overlapped blocks.

hence for 16 though this program should give me an output of 9 rows. it does not!

[Walter Roberson]

Seems to me that there should be 15 blocks in each direction when you using overlapping 2 x 2 blocks on a 16x16 image. For any top-left image index (I,J) then (I+1,J) and (I,J+1) and (I+1,J+1) are also valid top-left indices (overlapping by one pixel of the 2x2 block); by induction, top left indices can range from 1 to 16-1 on each side, which gives 15x15 = 225 blocks.

[David Young]

Walter is correct:

>> size(im2col(zeros(16), [2 2]))

ans =

4 225