Accurate rainfall intensity measurements are important to many applications like weather monitoring and forecasting, economics, urban design, agriculture and so on. A weather radar is one of the instruments to measure rainfall by reflected microwave signal from raindrops. However, the rainfall rate estimated by weather radar has various errors as it is an indirect measurement of rain. One type of errors relates to the measurement of other objects than raindrops and noise. These objects are termed clutter and represent buildings, trees, airplanes, birds, insects, … One type of techniques mitigates clutter and noise using polarimetric measurements in the spectral domain. This methodology, spectral polarimetric processing, can be reinforced by digital image processing techniques. Image segmentation is the image processing technique chosen due to its similar objectives with spectral polarimetric processing. There are two major challenges for this project. The first one is to search and design effective methods with the help of image segmentation techniques to remove clutter and noise while retaining precipitation. Most of the image segmentation studies were done to tackle images from other areas which have different features from compared to Range-Doppler spectrograms. The methods from image segmentation should be thoroughly tested before being applied. Another problem to be resolved is to assess the techniques applied without any ground truth data. And those methods used in previous studies may not be suitable for testing large data sets. A framework is proposed to understand and quantify the features of polarimetric spectrograms which can be useful for tackling the two challenges. Then, a new filtering technique, criteria for good results and evaluation methods are proposed based on the framework. Three different filtering techniques are tested by the evaluation methods in different scenarios. Finally, some points for future research are also discussed.