Yang Fan, He Yan, Deng Zhen Sheng, Yan Ang
School of Biology & Engineering, Guizhou Medical University, Beijing Road No. 8, Guizhou Province, Guiyang, China.
School of Biology & Engineering, Guizhou Medical University, Beijing Road No. 8, Guizhou Province, Guiyang, China.
Comput Biol Med. 2016 Apr 1;71:108-14. doi: 10.1016/j.compbiomed.2016.01.026. Epub 2016 Feb 10.
The full bone structure of X-ray images cannot be captured in a single scan with Digital radiography (DR) system. The stitching method of X-ray images is very important for scoliosis or lower limb malformation diagnosing and pre-surgical planning. Based on the image registration technology, this paper proposes a new automated image stitching method for full-spine and lower limb X-ray images. The stitching method utilized down-sampling to decrease the size of image and reduce the amount of computation; improved phase correlation algorithm was adopted to find the overlapping region; correlation coefficient was used to evaluate the similarity of overlapping region; weighted blending is brought in to produce a panorama image. The performance of the proposed method was evaluated by 40 pairs of images from patients with scoliosis or lower limb malformation. The stitching method was fully automated without any user input required. The experimental results were compared with previous methods by analyzing the same database. It is demonstrated that the improved phase correlation has higher accuracy and shorter average stitching time than previous methods. It could tackle problems including image translation, rotation and small overlapping in image stitching.
数字放射成像（DR）系统无法在一次扫描中获取X射线图像的完整骨骼结构。X射线图像的拼接方法对于脊柱侧弯或下肢畸形的诊断以及术前规划非常重要。基于图像配准技术，本文提出了一种用于全脊柱和下肢X射线图像的新型自动图像拼接方法。该拼接方法利用下采样来减小图像尺寸并减少计算量；采用改进的相位相关算法来找到重叠区域；使用相关系数来评估重叠区域的相似度；引入加权融合以生成全景图像。通过对40对脊柱侧弯或下肢畸形患者的图像评估了该方法的性能。该拼接方法完全自动化，无需任何用户输入。通过分析相同数据库，将实验结果与先前方法进行了比较。结果表明，改进的相位相关方法比先前方法具有更高的准确性和更短的平均拼接时间。它可以解决图像拼接中的图像平移、旋转和小重叠等问题。
