This proposed method calculates the centroids of two registering images by applying the moments for\r
acquiring the original displacement parameters, and then uses modified K-means clustering to classify\r
the image coordinates. Before clustering, the medical image coordinates is centralized, the two-row\r
coordinate matrix is created to construct the 2-D sample set to be partitioned into two classes, the slope\r
of a straight line fitted to the two classes is computed, and the rotation angle is computed by solving the\r
arc tangent of the slope. The edges are detected by the edge convolution kernel and the binary images\r
covering the characteristic points are extracted. Experimental results from aligning experiments reveal\r
that, this approach has lower computation costs and a higher registration precision.
