Nonlinear filters for multimedia applications

A new type of nonlinear image processing filters for multimedia applications is proposed. The filters remove small-amplitude fluctuations while keeping edges sharp. The filter algorithm is simply realized by combination of a usual linear filter and a comparator, which is suitable for hardware implementation. Here, two types of multimedia applications are considered, one is to post-process block-based encoded images which contain blocking artifacts, and the other to remove undesirable skin components in facial images, which is useful for facial beautification. Computer simulations verify the high performance of the proposed methods.