Robust Detection Technique for Removing Random-Valued Impulse Noise

This paper presents a new detection technique to detect the random valued-impulse noise. In this technique, the noisy pixels are detected progressively through several phases. In the first phase, the pixel that has a strange value amongst its neighboring pixels in the local window is detected first as a noisy pixel. Then in the next phase, the pixel that has less than three similar pixels amongst its neighboring is detected as noisy pixel, and so on. Extensive simulations demonstrate that this method comparable to the best method at low noise rates, but it outperforms other state of the art methods at high noise rates. This method suppresses impulse noise efficiently, while at the same time preserving the image details.