基于经验模式分解法的光学相干层析成像去噪研究

Study of de-noising based on empirical mode decomposition method in optical coherence tomography

针对光学相干层析(OCT,optical coherence tomography)成像中存在的散斑噪声和扫描噪声,提出了采用经验模式分解(EMD,empirical mode decomposition)算法同时减小这两种噪声的思想。EMD是一种时频分析法,较傅立叶谱法能准确地确定时变非平稳的这两种噪声随时间变化的频率特性,从而获得更好的滤波效果。结果表明,通过合理设计EMD滤波参数,即可有效地同时减小散斑噪声和扫描噪声,信号的信噪比(SNR)提高(不考虑扫描噪声时,SNR达7dB左右,考虑到扫描噪声时,SNR提高达3dB左右),扫描噪声的条纹对比度降低60%以上,改善了成像质量,同时图像细节得到保留。与小波去噪法相比,本文方法具有滤波器设计简单、去噪效果明显及能同时有效地去除两种噪声的优点。

Optical coherence tomography （OCT） has been shown to have advantages for obtaining images of biological tissues noninvasively and has great potential for applications in medical diagnosis. In this work,a method based on the empirical mode decomposition （EMD） method is adapted to improve OCT image quality by reducing the effects of noises. EMD is a time-frequency analysis, So it can more accurately analyze the time-varying non stationary noise signal compared with the domain transform method. The experimental results show that EMD method can effectively suppress speckle noise,which improves signal-to-noise （SNR） to 7 dB and 3 dB for absence and presence of the scanner noise. It can also effectively remove the stripe noise due to instability of the scanning reference mirror in OCT as long as the EMD filter parameter is set properly. The fringe contrast can be reduced by more than 60 %. The quality of OCT images is improved. Compared with the wavelet de-noising method,EMD has the advantages of simple operation steps, obvious de-noising effect, and the ability of effectively removing two kinds of noises.