东南大学学报(自然科学版)
東南大學學報(自然科學版)
동남대학학보(자연과학판)
Journal of Southeast University (Natural Science Edition)
2015年
5期
851-855
,共5页
李碧草%张俊峰%杨冠羽%舒华忠
李碧草%張俊峰%楊冠羽%舒華忠
리벽초%장준봉%양관우%서화충
Arimoto 熵%Demons 算法%结构图像表示%图像配准
Arimoto 熵%Demons 算法%結構圖像錶示%圖像配準
Arimoto 적%Demons 산법%결구도상표시%도상배준
Arimoto entropy%Demons algorithm%structural image representation%image registration
为了提高多模态医学图像的配准精度,利用参考图像和浮动图像的结构信息,提出了一种基于结构图像表示和微分同胚 Demons 算法的多模态医学图像配准算法。该算法由结构图像表示和图像配准组成。在结构图像表示阶段,采用 Arimoto 熵图像来描述参考图像和浮动图像的结构信息,首先计算2幅图像中所有像素点周围指定大小邻域的熵值,然后把计算的熵值作为熵图像中对应点的灰度值以生成2幅熵图像。在图像配准阶段,使用微分同胚 Demons 配准算法对2幅熵图像进行配准。基于 Brainweb 数据库中磁共振数据的测试结果表明:与微分同胚 Demons 算法和基于香农熵的 Demons 算法相比,利用 Arimoto 熵表示图像的结构信息可以进一步提高配准精度。
為瞭提高多模態醫學圖像的配準精度,利用參攷圖像和浮動圖像的結構信息,提齣瞭一種基于結構圖像錶示和微分同胚 Demons 算法的多模態醫學圖像配準算法。該算法由結構圖像錶示和圖像配準組成。在結構圖像錶示階段,採用 Arimoto 熵圖像來描述參攷圖像和浮動圖像的結構信息,首先計算2幅圖像中所有像素點週圍指定大小鄰域的熵值,然後把計算的熵值作為熵圖像中對應點的灰度值以生成2幅熵圖像。在圖像配準階段,使用微分同胚 Demons 配準算法對2幅熵圖像進行配準。基于 Brainweb 數據庫中磁共振數據的測試結果錶明:與微分同胚 Demons 算法和基于香農熵的 Demons 算法相比,利用 Arimoto 熵錶示圖像的結構信息可以進一步提高配準精度。
위료제고다모태의학도상적배준정도,이용삼고도상화부동도상적결구신식,제출료일충기우결구도상표시화미분동배 Demons 산법적다모태의학도상배준산법。해산법유결구도상표시화도상배준조성。재결구도상표시계단,채용 Arimoto 적도상래묘술삼고도상화부동도상적결구신식,수선계산2폭도상중소유상소점주위지정대소린역적적치,연후파계산적적치작위적도상중대응점적회도치이생성2폭적도상。재도상배준계단,사용미분동배 Demons 배준산법대2폭적도상진행배준。기우 Brainweb 수거고중자공진수거적측시결과표명:여미분동배 Demons 산법화기우향농적적 Demons 산법상비,이용 Arimoto 적표시도상적결구신식가이진일보제고배준정도。
In order to improve the registration accuracy of multi-modal medical images,a multi-mo-dal medical image registration algorithm based on structural image representation and diffeomorphic Demons algorithm is proposed by using the structural information of the reference image and the float image.This algorithm contains structural image representation and image registration.In the process of structural image representation,the Arimoto entropy is used to describe the structural information of the reference image and the float image.First,the entropic values of the neighbourhoods with the designed size of all pixel points in the two images are calculated.Then,these entropic values are re-garded as the intensity values of the corresponding points in the entropy image and two entropy ima-ges can be obtained.In the process of image registration,the diffeomorphic Demons registration ap-proach is used to register these two entropy images.The experimental results of the MRI (magnetic resonance imaging)data in Brainweb database show that applying Arimoto entropy to represent the structural information can further improve the registration accuracy compared with the diffeomorphic Demons algorithm and the Demons method based on Shannon entropy.