Aiming at the existing methods being short of effectiveness and robustness for high density impulse noise removal, and failing to preserve and restore the edges and detail structures of image, a twice iterative trimmed mean filter with wavelet detection is proposed. This method takes full benefit of characteristic of impulse noise and the highly correlation of wavelet denoising image, identifies the noisy pixel with extreme intensity values and wavelet denoising image, and uses the iterative trimmed mean filter with 3×3 and 5×5 neighborhood successfully for noisy pixel restoration. The experimental results show that in comparison to existing methods proposed very recently, the proposed method achieves higher PSNR and EPI values, and better visual perception, showing superior capability in denoising and detail structure preservation.