Based on the existing standard adaptive median filtering algorithm and gradient inverse weight smoothing algorithm, we propose an image filtering algorithm based on adaptive median and gradient inverse weight, aiming at the difficulty of mixed noise filtering by either single liner filter or non-linear filter. In order to filter the mixed noise, this algorithm adaptively adjust the window size to detect the salt & pepper noise and does median filtering, meanwhile, the gradient inverse weight smoothing is performed for the Gaussian noise by setting threshold. Experimental results show that this algorithm filters well for Gaussian noise, salt & pepper noise and the mixed noise, compared with the standard adaptive median filtering algorithm and the gradient inverse weight smoothing algorithm. And it has obvious filtering advantages for the high intensity image noise.