论文部分内容阅读
为有效抑制激光主动成像中的散斑噪声以及减少运算耗时,提出一种基于信号子空间谱域约束的散斑去噪方法。使用同态变换将乘性噪声变为加性噪声,对含噪图像进行奇异值分解并估计信号子空间的维数,根据该维数对含噪图像的协方差矩阵进行特征值分解,利用噪声子空间的特征值估计噪声的方差。根据信号子空间的特征向量计算出滤波估计矩阵,并与含噪图像卷积,做同态逆变换,得到降噪后的图像。为了去除后向散射和背景辐射对实验结果的影响,搭建一套基于距离选通ICCD的激光主动照明系统,采集不同散斑强度的主动照明图像进行实验。结果表明该方法拥有比经典的Lee、Frost和Kuan算法更好的散斑噪声抑制效果,同时运算时间在毫秒量级,满足激光主动成像算法实时性的需求。
In order to effectively suppress the speckle noise in laser active imaging and to reduce the computational time, a new speckle denoising method based on spectral subspace spectrum constraints is proposed. The homomorphic transform is used to transform the multiplicative noise into additive noise. The noise image is decomposed and the dimension of signal subspace is estimated. The covariance matrix of the noisy image is decomposed according to the dimension. Eigenvalues of subspace estimate the variance of noise. According to the eigenvector of the signal subspace, the filter estimation matrix is calculated and then convoluted with the noisy image to do the homomorphic inverse transform to get the denoised image. In order to remove the influence of backscattering and background radiation on the experimental results, an active laser illumination system based on distance gating ICCD was set up and the active illumination images with different speckle intensities were collected for experiments. The results show that this method has better speckle noise suppression than the classical algorithms of Lee, Frost and Kuan, and the computation time is on the order of milliseconds to meet the real-time requirement of laser active imaging algorithm.