The Signal-to-Noise Ratio (SNR) and contrast are low in infrared image, which is serious in low target radiation, long detecting distance and bad weather. For enhancing the SNR of infrared image, an adaptive Wiener filtering noise reduction algorithm is proposed. In this algorithm, the technical bottleneck that power spectrums of undegraded image and noise need be known in traditional Wiener filtering is broken. A neighborhood estimation criterion of noise variance of smoothness region is proposed, and the noise adaptive reduction of infrared image is achieved. Experimental results show the proposed algorithm can adjust filtering parameters adaptively, remove the noise, and in the mean time effectively preserve the information of edge and detail, which can be applied to infrared real-time imaging system.