基于改进光流场模型的脑部多模医学图像配准

 基于光流场模型的配准其常亮假设的光流场约束要求待配准的源图像和目标图像具有一致的灰度,因而只适用于单模态图像之间的配准,为此使用基于排序的精确直方图规定化对脑部MR-PD图像进行模态变换,完成与MR-T2图像之间的灰度映射.由于此配准方法主要用来寻找时间序列图像中的细小形变,当待配准的两图像差异较大时就不能取得满意的配准效果,为此使用能反映图像结构的标记点构造附加的外力对光流场模型加以改进,以获得更理想的配准参数.实验证明,本文方法能够实现具有较大差异的脑部MR多模序列图像之间的准确配准.

Registration of Multimodal Brain Medical Images Based on Improved Optical Flow Model

Registration method based on optical flow model is very suitable to time-sequence images,but the assumption of constant brightness at a point makes it only be used to make registration for signal-modal images.So an exact histogram specification method is proposed to transform MR-PD model to MR-T2 modal before registration based on optical flow model.If the difference between two images is large the registration result could not be ideal because this registration method is mainly used to correct the light deformation between time-sequence images,so information which reflects the structural feature of image is taken to improve optical flow model,that is,image landmarks are utilized to produce an additional external-force for optical flow model to get ideal registration parameters.Experimental results demonstrate that the improved method can realize accurate registration of multi-modality time-sequence MR images between which there is large difference.