Image-Filtering Techniques for meteorological radar

The images collected by meteorological satellites and radars are often infected by noise due to the environment, the data acquisition devices and the transmission system used. These images carry information about weather, which must be cleared up. In general, the filtering techniques applied to scrambled images aim at giving them a better visual aspect. In the case of meteorological radar images, such techniques are usefully employed to separate the atmospheric echoes from the ground clutter. Most radars equipping the meteorological stations are non coherent and then especially designed to estimate the amount of precipitations falling in the observed region. In this paper, the images collected by 5.6 GHz non coherent radar in the region of Setif (Algeria), are processed so as to keep the rainfall echoes unchanged and eliminate the ground ones. For this purpose, two filtering techniques are utilized, which respectively consist in computing the textural features of these images and masking the ground echoes. The textural approach is implemented by using an analysis window of 7times7 pixels. The radar echoes are then filtered by using either energy or homogeneity. The masking method is based on the selection of clear sky images among a sequence of radar data collected during a sufficiently long period of time. The template of the relief of the region under study is built by averaging all the clear sky images. To get only the rainfall echoes, the resulting mask is applied to each of the original radar images and the filtering is improved by removing the residual clutter with the help of median filter and shape detection. All these filters are found to be accurate enough for meteorological uses.