Abstarct: This study aims to use new techniques to estimate reservoir parameters such as porosity using seismic attributes within an Iranian oil field. To estimate the porosity of the reservoir quantitatively, a combination of multivariable regression and various artificial neural networks are used to process seismic attributes. In addition to three dimensional (3D) seismic data, a set of resistivity, sonic, density and porosity well logs data are utilized. In order to correct sonic logs velocity data, information of well data (check shot), seismic horizon picking, the wavelet and synthetic seismic trace, correlation of well and seismic data and the results of seismic inversion are used. To achieve the goal, the inversion was firstly carried out on seismic and well logs data and the required seismic attributes were then extracted from the seismic data using Hampson-Russell software. The technique of cross-validation was used to show which attributes are significant. Once the most appropriate seismic attributes have been found, the relationship between them and reservoir porosity in the well locations has then been determined to be used for porosity calculation by sets of appropriate seismic attributes throughout the reservoir volume. The results indicate six seismic attributes including acoustic impedance, filter15/20-25/30, time, amplitude weighted frequency, cosine instantaneous phase, and derivative instantaneous amplitude are the most relevant attributes in porosity estimation. The Hampson-Russell software was then used to estimate the reservoir porosity using multivariable regression analysis, and three different probabilistic, multi-laxyer feed- forward, and radial basis function neural networks (i.e. PNN, MLFN, and RBFN). The results of study show that the probabilistic neural network (PNN) with root mean squares error (RMS) 4.69% and correlation coefficient (R) 68.26%, is the best method for porosity estimation using the above six attributes. At the next stage a few more relevant seismic attribute were extracted from 3D seismic data using Petrel software and transferred to be used within the Hampson-Russell software to enhance the quality of porosity estimation. The obtained results indicate that nine seismic attributes (instantaneous quality, apparent polarity, iso-frequency components, real acoustic impedance log, instantaneous bandwidth, time gain, cosine instantaneous phase, inversion results, and amplitude envelope) are the best for porosity estimation in this phase. The results also show the PNN is again the best approach among three artificial neural networks and regression analysis for estimation as its estimation error (RMS=3.85%) is the least and its correlation coefficient (R=79%) is the most among four estimation approaches during validation step.