To achieve better infrared focal plane array (IRFPA) noise suppression with minimum edge blurring, an anisotropic filtering algorithm is proposed based on Tukey model. This algorithm distills edge by differential, and then the image preprocessing is carried out to improve ill-posed problem of Perona-Malik (P-M) equation based on the distribution properties of signal and noise. A new diffusion coefficient was established to satisfy Charbonnier rule. Initialized edge conditions are presented to preserve edge. It has run on the DSP+FPGA platform, and has good effect in the respect of noise removal and edge preservation. The visual effect of the treated image is promoted.