基于矢量相似性的多元滤波方法研究

为了避免彩色图像滤波时排序的难题，在探究矢量相似性的基础上提出一种针对彩色图像的多元滤波方法。首先，在RGB彩色空间中，利用彩色矢量间距离和角度定义其相似性测度，以刻画与人类视觉感知相一致的彩色相似性；其次，以上述彩色相似性为准则设计并构建一种彩色多元滤波方法，并就其相关参数对滤波性能的影响进行了深入分析和研究；最后，为了验证提出方法的有效性，将其应用于标准彩色影像和彩色遥感影像滤波。提出的方法不仅有效解决了传统滤波方法的排序难题，而且克服了因滤波使图像模糊、边缘不清等问题；另外，将提出的多元滤波与传统方法进行实验对比，结果表明其不仅能有效滤除多种类型噪声，而且较好地保持了原始图像信息，使图像信息保真清晰，其视觉效果优于传统，客观评价指标也有较大幅度的改善和提升。

Research on multivariate filtering method based on vector similarity

In order to avoid the problem of sorting in color image filtering, this paper defined and proposed a multivariate filter method for color image based on the study of vector similarity. Firstly, in RGB color space, it defined the similarity measure between the color vectors by the distance and the angle to describe the color vector similarity consistent with the human visual perception. Secondly, it designed and constructed the color multivariate filter using the color similarity criterion, and analyzed and studied in depth the influence of the related parameters on the filtering performance. Finally, in order to verify the effectiveness of the proposed method, it was applied to the actual remote sensing image filtering. It can be seen from the experiment that the proposed method not only effectively solves the problem of sorting of the traditional filtering methods, but also overcomes the problems of blurring and unclear edges due to the filtering. In addition, compared with the results of other traditional methods, the results show that the proposed method not only can effectively filter all kinds of noise, but also maintains the original image information better, making the image information clear and fidelity. The visual effect of this filtering method is superior to the traditional, its objective evaluation index also has greatly improved.