A Separable Filter for Directional Smoothing
A Separable Filter for Directional Smoothing
复制标题
DOI：
10.1109/lgrs.2004.828178
复制
发表时间：
2004-07-01
影响因子：
4.8
通讯作者：
Lakshmanan, V.
中科院分区：
文献类型：
--
作者：
Lakshmanan, V.
各向异性和方向过滤器可以在保持对象边界的同时平滑噪声图像。由于几何或功率的限制，来自遥感仪器的数据往往会丢失像素。在这种情况下，这些非各向同性过滤器的效率非常低，因为当存在丢失数据或需要执行逻辑操作时，不能使用转换方法。在这封信中引入了一个方向过滤器，它保留了处理丢失数据的能力，并且是可分离的，使其在计算上变得高效。我们在天气雷达数据上演示了方向滤波，它可以用来平滑锋面。由于这里介绍的过滤器可以针对比例、方向和纵横比进行参数化，因此该过滤器可以用于任何不能使用变换方法但需要计算效率的定向过滤应用中。
Anisotropic and directional filters can smooth noisy images while preserving object boundaries. Data from remote sensing instruments often have missing pixels due to geometric or power limitations. In such cases, these nonisotropic filters are very inefficient, because transform methods cannot be used when there is missing data or when logical operations need to be performed. A directional filter is introduced in this letter that retains the ability to handle missing data and is separable, making it computationally efficient. We demonstrate the directional filter on weather radar data where it can be used to smooth along fronts. Since the filter introduced here can be parameterized for scale, orientation, and aspect ratio, this filter can be used in any directional filtering application where transform methods cannot be used, but computational efficiency is desired.