Project

Control point selection and matching for registration of CT/MRI images

Medical images are acquired using different modalities depending on the type of body part to be imaged. This results in various types of images like MRI, CT, PET, SPECT and X-Ray Images, each representing different features/aspects of the area being scanned. Study of these different types of images may be needed for clinical analysis by doctors. A general scenario is to study the images separately. It would be very useful if the information from different modalities could be presented in a single image, the process being called as image fusion. Analysis becomes easier if the 2D stack of images is reconstructed into a 3D image. All these require that the corresponding point in the different images to be matched. This process of finding corresponding points in the images is called as Image Registration. This aspect is the main focus. In this work, a structural method for selecting and matching the control points, for a point based image registration method has been proposed. This method involves representing the image as a graphand then matching the corresponding structures in the input images using the degree, weighted edge and angle between the edges as features for matching.