中国数字人数据集断层图像自动配准算法

在采集数字人数据时，由于相机在拍摄过程中存在的各种干扰，使得到的各断层图像的投影参数不能保证一致，通过对相机拍摄模型和干扰因素的分析，提出了一种基于二维射影变换（pmjective transformation）的断层序列图像自动配准算法。该算法先通过层间继承的方法寻找每张照片上的定位杆区域，用形态学操作来确定该层定位杆圆心坐标；然后根据定位杆的基准坐标，得到该层的射影变换算子T；最后对原始图像进行图像变换，得到投影参数一致的照片数据集。本算法对数字人采集的原始数据预处理有重要意义，并能为后续的二次数据库开发、图像分割、重建和数字解剖学等提供精确的三维数据集。

An Auto-registration Algorithm for Chinese Digital Human Anatomy Image Database

Because of the existence of many different types of distortion in the process of taking photos when generating the true-color 2D slides from cryosections of the human cadavers for Chinese digital human image database, the uniformity of projection parameters for each slide can not be assured. Through analysis of the model for camera shoot and elements of distortion, this paper proposed an auto-registration algorithm based on 2D projective transformation. First, seek the approximate region of locating rod for each slides using inherit method and obtain the center coordinate of locating rod using morphology operation. Then, according to the base coordinate of locating rod, gain the coefficients of projective transformation. At last, generate ideal slides with uniformity of projection parameters using image conversion. This algorithm is of importance for the preprocessing of original digital human database and is expected to provide a precise 3D database for secondary development of database, image segment, reconstruction and digital anatomy.