Abstract: How to improve images registration's precision and efficiency while keeping a low error rate is an important topic worthy of study. In order to meet the need of clinical demand, a mixed programming registration strategy was proposed. In this method, the preregistration is finished by the combination of the extraction of the image's gravity center and the detail enhancing of the image which is generated by wavelet decomposition. On this basis, the process of registration is based on the gray-scale of the images,and it uses the Powell optimization algorithm and the traditional maximum mutual information as the method of measuring the similarity between the images. In addition, we proposed an improvement on the oncdimensional search within Powcll algorithm which is different from the 13rent algorithm, and this makes it more suitable for image registration without losing the accuracy and efficiency. The experiment shows that our registration strategy performs very well in avoiding false registration, and its accuracy reaches sub-pixel level, moreover, its efficiency meets the need of clinical demand.

Key words: Image registration, Center of gravity, Wavelet decomposition, Powell optimization algorithm, One-dimensional search