Automatic B-spline image registration using histogram-based landmark extraction

Recognition and correction of inhomogeneous displacement caused by patient´s movement has been recently discussed as an interesting topic in medical image processing. Considering consistency in general structure of the image during distortion, histogram could be employed as a fast implementation method in feature domain. Accordingly, attribute vectors could be defined for each pixel based on spatial features to find corresponding points in two images. Consequently a point-based and non-rigid transformation approach will be designed. A B-spline image registration has been applied to match those pairs with a defined smoothness factor. This algorithm is a step-by-step registration process controlled by this factor. The proposed algorithm has been applied to brain MR images. The normal mean square error value has been measured between registered and original images and the result shows a significant improvement in the proposed algorithm.