针对医学图像配准对准确性高、鲁棒性强和速度快的要求,本文提出一种新的基于区域联合Rényi熵的多模配准算法.该算法将区域信息融入到联合Rényi熵中,并使用最小生成树来估计区域联Rényi熵.这样,不仅改善了传统配准方法由于忽略像素空间信息造成的配准鲁棒性的降低,而且避免了使用直方图估计高维熵遇到的"维数灾难"问题.实验结果表明在图像含有噪声、灰度不均匀和初始误配范围较大的情况下,该算法在达到良好配准精度的同时,具有鲁棒性强、速度快的优点.作为一种一般性的配准算法,基于区域联合Rényi熵的配准方法还可以应用到图像配准以外的更广阔的领域,如图像检索、对象识别等.%For medical image registration of high-accuracy, robustness and speed requirements, this paper proposes a new multi-modality image registration algorithm based on regional joint Renyi entropy. This algorithm incorporates regional information into the joint Renyi entropy, and then estimates joint Renyi entropy directly by minimum spanning tree. Thus, the new algorithm improves the losing spatial information of the traditional registration methods, and also avoids the "dimension disaster" problem encountered when using high-dimensional joint histogram to estimate the entropy. Experimental results show that in the images with noise, non-uniform intensity and large scope of the initial misalignment case, the algorithm achieves better robustness and higher speed while maintaining good registration accuracy. As a general registration algorithm,it can be used in wider areas, such as image retrieval, object recognition.
展开▼