A fusion algorithm of infrared polarization and intensity based on discrete wavelet transform (DWT) is put forward. DWT is adopted in the algorithm to decompose a source image into multi-scale images. Then the low frequency sub-band coefficient is fused by the local energy. And the high frequency sub-band coefficient is fused by eliminating high frequency noise and incorporating with local mean square error rule. The reconstructed fusion images are got finally through inverse transformation. Experimental results show that local standard deviation, roughness, contrast and local entropy are increased by 1.1%, 3.0%, 15.2%, 2.6% respectively comparing with the fusion rules of weighted average and the largest absolute value. So the fusion method has a certain advantage and reality.