Impulse noise removal by the difference-type noise detector and the cost function-type filter

All kinds of median filters have been often adopted in impulse noise reduction. However, many details (e.g., thin lines and step edges) are easily lost, especially at high noise situation. Although many new techniques were added to these filters, the results of image filtering are still not satisfactory. Thus, in this paper, a cost function-based filter is proposed. The corrupted pixels are processed by differentiating the cost function. During image processing, values of the uncorrupted pixels are required to be transformed into constant terms of the cost function, thus, a difference-type noise detector is presented in this paper. During simulations, the proposed filter is compared with several filters in the numerical value and the vision, respectively. Experimental results show that the proposed filter can very effectively remove impulse noise and preserve more details of original images.