which similarity measure to use on image blocks as i m getting bad results with correlation and root mean square error?the explanation is listed below

i m working on gray scale images. and i've divided it into overlapping blocks.

then i've used the root mean square error method in order to get the similarity between the image blocks.but it does not produces good result.

like

block 34=  [158 158
            162 129]

 block 33=   [ 158 158
               162 162]

when i calculated there rms value it comes out to be 16.500000

the above two blocks are approximately similar, the only difference in the last row's 2 elementi.e 129 and 162 rest values of both the blocks are same..

when i applied the condition as rms>0 and rms <2 the above two blocks are totally eliminated but these two blocks are identical(approximately).

plz help me.

can u tell me which measure should i use so that the pronblem can be solved.