Rainfall estimation from satellite data have many applications in climatology and meteorology but calculation associated requires a rapid processing to large amounts of data in order to achieve significant result.The neural networks(NN) method is one of the several techniques employed to extract meteorologically-useful information from remotely sensed data.However this method is hardly used independently to yield quasi-real time rainfall estimates since a large amount of satellite data are needed to generate the input/output data for the NN training.In order to overcome this shortage,multiresolution wavelet transform(WT) technique is proposed to decompose the images to obtain the key information for further analysis.As a result,the NN training becomes easier and faster.In the paper a case study to estimate rainfall over the central part of the Sá Paulo state,Brazil using both the NN and WT techniques is given.The analyses were performed using GOES-8 brightness temperature data and meteorological radar data from Bauru,SP.It is concluded that NN can be successfully used to estimate rainfall from remotely sensed imagery.