超声图像序列运动估计相似度函数研究

Investigation of Similarity Functions for Ultrasound Image Sequence Motion Estimation

研究超声图像序列运动估计的相似度函数性能问题,超声图像易受斑点噪声恶化,造成图像质量差。传统的相似度函数的选择影响运动估计精度和应用效果。为评估不同相似度函数性能,通过对像素强度和超声散斑特性的两类相似度函数在仿真的超声图像序列上用图像块匹配方法进行分析,并采用运动矢量场角度误差和终块误差法量化评估其在超声图像序列上进行运动估计的优劣。仿真结果表明,基于超声散斑特性的相似度函数能更准确地对具有超声图像特性的仿真图像序列进行运动估计,结果对不同超声研究领域中运动估计及相似度函数的选择具有重要的参考价值。

Similarity function is one of the key technologies for ultrasound image motion estimation. Because of the speckle noises of the ultrasound images, the selection of similarity function has direct effect on the accuracy of motion estimation and applications. The intensity-based similarity function and ultrasound feature-based similarity functions were used to track the motion of simulated image pairs. The motion estimation was carried out by fixed block matching method. Average angular error and end block error were used to indicate the performance of different simi- larity functions under simulated situation. Simulation results show that ultrasound feature-based similarity functions can capture the motion more accurately for simulated uhrasound image sequences, compared with the intensity-based similarity measures. And thus it provides a reference for real case.