Kindly review my program for block matching ? How do I overcome its constraints?

Hello all! Below is the code for block matching of two imags al and br(which is a sub-image of al). I divided al into 4 blocks, matched with br and computed SSd. Though the result was quite accurate, I have few questions:

• In this case, my images blocks are divided exactly, what if my block sizes are different
• This is not the most accurate algorithm. Is there any other way to implement "block matching algorithm"?

clear all;
close all;

al = imread('testimage.jpg');
br = uint8(zeros(125, 125, 3));
br(1:126, 1:126,:) = al(80:205, 100:225 , :)

imtool(al); imtool(br);
al= rgb2gray(al);
al = im2double(al);

br= rgb2gray(br);
br = im2double(br);

          c(1:126 , 1:126, :) = al(1:126 , 1:126, :)
          d(1:126, 1:126,:) =  al(100:225, 100:225,:)
          e(1:126, 1:126,:) = al(100:225, 1:126,:)
          f(1:126, 1:126,:)  = al(1:126,100:225, :)

          imtool(c); imtool(d); imtool(e); imtool(f)
       ssdc= 0; ssde=0; ssdd=0; ssdf=0;  
       diffc = 0; diffd = 0; diffe=0; difff=0;
       for( k = 1:1:126)
           for(l = 1:1:126)
               diffc = c(k,l) - br(k,l);
               ssdc = ssdc+ (diffc*diffc);
                diffd = d(k,l) - br(k,l);
               ssdd = ssdd+ (diffd*diffd);
                diffe = e(k,l) - br(k,l);
               ssde = ssde+ (diffe*diffe);
                difff = f(k,l) - br(k,l);
               ssdf = ssdf+ (difff*difff);
           end
       end
    y = [ssdc ssdd  ssde ssdf]  ;
    min(y)

<<

>> The sub-block is the image d, which is a part of image a. I got the minimum cost at d. Kindly help me solve the above problem. Thank you!!

Answer 1

[Image Analyst]

You can use normalized cross correlation. See demo attached below the image it creates.