基于十字模板的数字减影图像配准快速算法

提出一种基于十字模板的数字减影图像配准快速算法.采用块匹配法消除几何失真带来的运动伪影.选择不受血管区域影响的差值直方图能量尺度作为相似性尺度.通过计算图像梯度信息,选取边缘区域像素作为控制点,既改善减影图像质量,又提高运算速度.用十字模板替代传统块匹配法中常用的矩形模板,可减少直方图统计中的像素数量,缩短计算时间.对比矩形模板和十字模板的差值直方图能量搜索结果,发现参数选取适当时,改变模板形状不影响控制点运动位移量的正确确定.对十字模板中像素进行采样,可进一步压缩像素数量,提高运算速度.实验结果表明,改变模板形状后,图像配准速度显著加快,且减影图像质量保持良好.

A fast image registration technique in digital subtraction angiography with cross template

A new cross template was applied to the problem of image registration in digital subtraction angiography. Block matching was used to eliminate artifacts caused by motion displacement. The energy of the histogram of diffe-rences was used as a measure of similarity. The quality of the subtraction image was improved and computation time was reduced by selecting edge pixels as triangular control points. Computation time was significantly reduced by replacing the traditional rectangular template with this cross template. Comparison of rectangular templates and the new cross template shows that the determination of motion displacement at control points is robust regardless of the shape of the template if the template parameters are appropriately chosen. Computation speed was further increased by sampling pixels in the cross template. Our results show significant improvement in calculation speed using the cross template while the quality of the resulting subtraction image remains high.