高速摄影下气液泡状流双阈值小波去噪方法

Double-threshold wavelet denoising of gas-liquid bubbly flow in the high-speed photography

根据气液泡状流三维特征参数高精度测量需求,针对高速摄影法采集气液两相流图像带来的噪声特性,提出双阈值小波去噪方法实现气液两相流多气泡图像去噪。基于硬阈值与软阈值的理论模型,采取双阈值小波去噪的有效方法,达到分离目标区域,保持良好边缘特性并获得良好去噪效果的目的,克服了传统的小波硬阈值过度"扼杀"图像信息的缺陷,优化了软阈值边缘特性的不足,同时解决了半软阈值算法复杂较难实现的根本问题。实验结果表明,本文方法原图像去噪后信噪比(SNR)可提高11%,熵值提高5.3%,均方根误差(RMSE)降低了6.9%,能有效地消除图像背景噪声,在不失真的情况下获取较为平滑的气泡图像,并保持气泡边缘特性,提高了后续流动特征的提取精度。

According to three-dimensional （3D） characteristic parameters of gas-liquid bubbly flow and high precision measurement requirements, double threshold wavelet denoising method is proposed for such noise characteristics to achieve gas-liquid multiple bubbles image denoising. Based on the theoretical models of hard threshold and soft threshold, we use double-threshold wavelet denoising method to divide target area effectively,maintain better edge characteristics and achieve a better result. The method can solve the problems that the hard-threshold wavelet denoising excessively strangles the real original information, the bubble edge characteristics cannot be optimized using soft-threshold effectively, and Semi-soft threshold algorithm is too complex to use in the practical applications. The experiment results show that the signal-to-noise ratio （SNR） of image aker denoising can be increased by 11 %, and the entropy is increased by 5.3 %. Compared with the general wavelet threshold denoising algorithm, root mean square error （RMSE） is reduced by 6.9%. The method can eliminate background noise effectively,smooth the images without image distortion, and maintain the edge feature of bubbles simultaneously. It improves the subsequent extraction accuracy of flow characteristic effectively.