矩匹配和变分方法相结合的 MODIS 条带去除模型

摘要

A novel noise-reduction model is proposed to remove the multi-type stripe noises by combining moment matching with variational-based method,based on analyzing the characteristics of three types of stripe noises in moderate resolution imaging spectroradiometer (MODIS)data.Firstly,moment matching method is employed for preprocessing the MODIS imagery.Secondly,the variational method is performed to remove the stripe noises left in the image.The energy functional consists of L 1 regularization term,image data fidelity term and image gradient fidelity term.Split Brgman iteration is adopted to achieve the optimization solution of the proposed energy functional.Experimental results indicate that the proposed model can reduce stripes in MODIS data effectively.Compared with destriping approaches including moment matching,low-pass filter and unidirec-tional variational destriping,the proposed model can preserve the details of the MODIS image and alleviate the block effects caused by regularization-based methods better.%针对中分辨率成像光谱仪（moderate resolution imaging spectroradiometer，MODIS）影像常见的3种条带噪声，提出了一种矩匹配和变分相结合的多类条带噪声去除模型。首先采用矩匹配方法对 MODIS 影像做预处理，然后基于变分方法去除条带噪声。能量泛函由 L 1正则化项、数据保真项和梯度保真项组成，基于 Split Bregman 迭代计算能量泛函的最优解。实验结果表明，此算法能够有效地去除 MODIS 影像中常见的条带噪声。与矩匹配、低通滤波及单向全变分方法相比，此模型更好地保留了图像的细节信息，有效抑制了正则化方法在条带去除过程中出现的分块现象。