基于序列相关性的超声图像自适应去噪

Correlation of sequence-based ultrasound image adaptive denoising

针对医学超声图像易受电子噪声与斑点噪声影响的问题,研究超声图像序列在不同变化速率下帧间相关性,提出一种基于序列相关性自适应加权复合去噪策略。通过测试不同速率下人体组织超声影像序列图像间运动剧烈程度的相关性,利用超声图像间像素点差值直方图获取互相关信息,根据直方图阈值波动范围调整运动剧烈指标,通过基于方差特性分析帧相关复合,快速实现不同相关性图像间的自适应加权复合。实验表明,基于超声图像序列相关性去噪方法能很好地压制电子噪声,平滑图像斑点区,保持超声图像生物组织结构的细节信息,优于常规去噪方法。

In order to solve the problem that the medical ultrasound images were degraded by electronics and speckle noise, this paper investigated the correlation between ultrasound image sequences under different speed, and proposed an adaptive compounding denoising approach according to the correlation of ultrasound sequences. Firstly, the approach tested correlation of motion intensity from sequence in vivo ultrasound images, and obtained the cross-correlation （CC） parameters from discre- pancy histogram between neighbor frames. Secondly, it modulated motion intensity by the fluctuation range of histogram threshold. Finally, it analyzed characteristics of correlation in compounding procedure based on variance, weighted compoun- ding adaptively and quickly under different correlation parameters. The experimental results indicate that the approach of cor- relation of sequencey-based ultrasound image adaptive denoising can suppress electronic noise, smooth the speckle regions, keep the detail resolution in biological structure, and have better performance than the general denoising methods.