Abstract
Image registration is a crucial problem in medical image analysis. A new automatic registration algorithm is presented for complex multimodality medical image registration. The proposed method starts by finding edge elements locally, and then aggregates them globally to find straight lines segments. Based on the consistency of the characteristics such as orientation and length of the straight-line segments, the relations of these segments are first estimated. Afterwards, the distributions of scaling and translation differences are used to estimate the remaining transformation coefficients. An excellent result has been obtained by the proposed algorithm when it tested over the real medical images taken from the King Fahad Specialist Hospital-Dammam, Saudi Arabia.