As phase shift errors will lead to severe ringing artifacts in the restoration results of optical synthetic aperture images, a method to detect and remove the ringing artifacts was proposed based on visual perception. Firstly, the texture region was obtained by calculating the scales of gradient direction randomness and local variance distribution. Then, the whole edges and main edges were separately detected by the Canny operator with different thresholds and whole edges belong to the neighborhood of all edges were considered as the edge ringing candidate. Finally, the ringing region and flat region were respectively derived by combining the texture masking of human visual system. Based on the image classification, the adaptive bilateral filter was applied to removal of the ringing artifacts. Experimental results show that the Peak Signal to Noise Ratio ( PSNR) of a Lena image after processing is improved by 10.8% while the Structural Similarity(SSIM) of resolution board image is improved by 0.057 as well. It condudes that the proposed method classifies the image on pixel scale, so it eliminates the ringing artifacts dramatically while preserving the edges and textures effectively. It is capable of improving the image restoration quality.