论文部分内容阅读
为有效抑制超声多普勒血流信号声谱图中的背景噪声和多普勒斑点,提出了Matching Pursuit(MP)及单向衰减阈值脉冲耦合神经网络(MP-PCNN)模型。首先将分段的多普勒超声信号进行MP循环分解,分离噪声与信号,然后用单向衰减阈值PCNN模型计算声谱图在各个灰度等级上的点火时刻图并定位斑点,用中值滤波器抑制斑点。通过对各种信噪比的仿真超声多普勒血流信号处理,实验结果表明,MP-PCNN方法可有效地滤除声谱图中的噪声与斑点,并较好地保持边缘与细节信息,在主观及客观性能比较上优于同类降噪去斑方法。
In order to effectively suppress the background noise and doppler spots in the ultrasound Doppler flow signal spectroscopy, a Matching Pursuit (MP) and one-way decay threshold pulse coupled neural network (MP-PCNN) model is proposed. Firstly, the segmented Doppler ultrasound signal is decomposed by MP, and the noise and signal are separated. Then the PCNN model of one-way attenuation is used to calculate the ignition timing map of the spectrogram on each gray level and to locate the spots. Suppresses spots. The experimental results show that MP-PCNN method can effectively filter out the noise and speckle in the sonogram and keep the edge and detail information well, In the subjective and objective performance is better than the same noise reduction spot method.