Permanent downhole gauges (PDG) for pressure and temperature are very common in oil and gas wells. The data collected by PDG´s are used for monitoring the well condition and reservoir performance and, additionally, may provide new and complementary information for reservoir characterization. PDG record data at high sample rates, usually as high as thousands per hour. The computational analysis of this large volume of data requires a previous processing for the removal of noise and outliers and for the reduction of the filtered data. The Discrete Wavelet Transform (DWT) and the Multiresolution analysis were used successfully for this purpose. In the Multiresolution analysis, the DWT is used to decompose a signal in approximation and detail coefficients through low pass and high pass filters, respectively. The coefficients are modified, according to a thresholding method, to remove the noise and outliers before the reconstruction of the original signal. The final result depends on the kind of wavelet and resolution level used for the decomposition process, the method used to estimate the noise level and the rules used to modify the coefficients. In this work, the performance of different wavelet decomposition, noise threshold estimators and thresholding methods was investigated. Several combinations were tested on synthetic and actual data to establish a procedure that efficiently removes the noise and preserves the sharp features, characteristic of changes in production or injection rates, with minimum user intervention.