The P-impedance is one of the most important elastic parameters of rocks, and it is commonly used for reservoir characterization. Conventional P-impedance inversion merges a low-frequency log-based model with a high-frequency seismic-derived model. We have proposed a method to estimate the P-impedance by employing dipole-based matching pursuit (DMP) decomposition. The matching pursuit decomposes the seismic traces into a superposition of scaled wavelets, and the associated scalar information represents the reflectivity series, which can be integrated for P-impedance estimation. Unfortunately, DMP analysis is usually performed trace by trace, resulting in a poor lateral continuity. Applying conventional lateral smoothing through mean or median filtering improves the lateral continuity but typically decreases the vertical resolution. We have evaluated an adaptive smoothing strategy that required the filtering to follow bed boundaries in an automated manner, sharpening the boundaries while maintaining the high quality of inversion. We have determined the effectiveness of our algorithm by first applying it to a synthetic wedge model and then to a real seismic data set.