A selective and adaptive image filtering approach based on impulse noise detection

The paper proposes a novel approach to remove impulse noises selectively and adaptively in a grayscale image based on impulse noise detection results. It first detects impulse noises in an image one by one in a serial mode. The threshold, which is used to estimate noise pixels, can be modified adaptively according to the mean brightness value of different regions. The results collected in the previous detection processes can be used in the subsequent detection processes. Modified median filtering is adopted to remove the noise, whose window lengths and selected samples are decided on the basis of the noise detection results. This method can effectively remove the impulse noises while successfully preserving the image edge details. The experimental results show that this approach is valid and worthy of application to image processing.