The partial differential equations (PDE) filtering model is an important filtering method in electronic speckle pattern interferometry (ESPI) technology. The differential operator is often approximated by difference scheme in PDE models. Three kinds of difference scheme such as central difference, nine point difference and higher difference are introduced. The representative orientation second order PDE filtering model is selected and analyzed, approximating differential operator in PDE model with three different difference schemes using simulated fringe image, simulated phase image and experiment fringe image. The result indicates the high density region and sparse density region can be balanced preferably with higher difference for large density change image. Nine point difference and central difference schemes should be deposed with average filtering. The processing rate with central difference scheme is the fastest, the higher difference takes the second place, and the nine point difference is the slowest.