Mixed gaussian and impulse noise removal based on kernel observation and edge direction

Two stages filtering process to reduce the mixed Gaussian plus impulse noise is proposed. The first stage is reducing impulse noise. Furthermore, the second stage is reducing Gaussian noise. A merger between the decision based method and kernel observation is conducted to induce impulse noise. Meanwhile, five Gaussian masks are used to reduce Gaussian noise. The use of the mask is adjusted to the image texture. In contrast to the previous method, our method can give a much better restoration in some particular cases, not only in the quality of the filtering result, but also in the fast computation time. In this study, we can improve the speed of previous method. Experimental rsults have shown that the proposed method outperforms the alternative filters in suppressing the mixed Gaussian plus impulse noise.