songxiao-tt/nonrigidimageregistrationmarkovrandomfields

Algorithm using a MRF for (medical) image registration.

NonRigidImageRegistrationMarkovRandomFields

Algorithm using a MRF for image registration.

How to use the algorithm.

Read the first image. For a Dicom file which will become Image I.

infoI = dicominfo('thorac.dcm');
I=dicomread(infoI); % image that will be transformed.

Read the second image. For a Dicom file which will become Image J.

infoJ = dicominfo('thorac2.dcm');
J=dicomread(infoJ); % image that will model the transformation.

Running the algorithm. The variable IT will become the registered image, the 3rd parameter is the number of iterations that will be performed, we recommend between 25 and 30 iterations.

IT = MRF(I,J,30);

Then, in order to calculate the Sum Absolute Difference between the 2 original images, we run:

IJSAD = SAD(I,J);

To calculate the Sum Absolute Difference between the registered image and the target images, we run:

ITJSAD = SAD(IT,J);

To plot the 3 different images. The image I, the transformed image and the target image J.

subplot(3,1,1)
imshow(I,[]);
title("Image I");

subplot(3,1,2)
imshow(IT, []) ;
title("Transformed Image");

subplot(3,1,3)
imshow(J,[]);
title("Image J");

More images used can be found in the bin.zip file.