The enhanced infrared noisy image usually involves noise which is difficult to be eliminated by traditional methods. A new adaptiveenhancing algorithm for infrared noisy image was proposed to overcome the contradiction of enhancing edge and suppressing noise. Firstly, the decomposition of noisy image was accomplished by using a new Contourlet transform with sharp frequency localization. Then, the coefficients were denoising in different scale and direction according to the shrinkage thresholds, which were estimated with layering adaptivethreshold algorithm. Finally, an image sharpness metric, which was used to calculate semi-soft threshold enhancement function, was applied to the denois high frequency sub-band images. Thus, the local pixels of noisy image were adaptively enhanced. Experimental results show that the proposed algorithm is effective.