Non-uniform multi-resolution analysis approach in medical image registration

When using conditional multi-resolution analysis of uniform space sub sampling, the important edge information will be lost in image registration based on intensity, which will result many local maxima and wrong registration. In order to solve the above problem, the non-uniform multi-resolution analysis method is proposed, which is based on sub sampling method of intensity probability distribution and gradient information. The non-uniform multi-resolution analysis is applied in the rigid registration of brain images. The results of test show that the new method can decrease the number of maxima and improve registration speed.