The main instrument of the French-Indian Altika/ SARAL mission scheduled for launch in 2010, is the Kaband Altika altimeter. The high attenuation due to atmospheric water (liquid or vapor) at this frequency band is the major drawback of the use of Ka-band. In part I of this paper, the impact of rain/clouds on Ka-band data and on the accuracy of the estimates of the geophysical parameters have been analyzed and quantified using an analytical model of waveform. Waveform distortion and errors on the geophysical parameters can be significant especially for small dense clouds and rain cells. It is thus necessary to flag the data potentially affected by rain and clouds. The use of a single channel for Altika prevents the use of the classical dual-frequency rain flag used for Topex or Jason altimeters, and requires the definition of a new flag based on the altimeter signal alone. Past studies showed that clouds and rain are characterized by sharp coherent along-track fluctuations of the off-nadir angle estimates. The new flagging algorithm is based on the analysis of the variations of this parameter by Matching Pursuit (MP) algorithm. MP allows the decomposition of a signal into a few salient features or atoms chosen from a dictionary of elementary functions. The dictionary is here defined by the wavelet decomposition of the signal. The method has been tested on an ensemble of Altika passes simulated for cloudy, rainy and cloud/rain free situations. The false alarm rate is almost nil while the detection performances are better than 90% at a range error of 5 cm and significant wave height error of 20 cm. The flag can be easily adapted to other altimeters' data and has been used to flag several Jason-1 passes. The comparison to the operational dual-frequency flag shows that the MP flag performs better in detecting range errors and waveforms distortion, while its performances are inferior in detecting samples attenuated by rain.