Application of polynomial interpolation to salt and pepper noise removal

In this paper, we apply the one-dimensional polynomial interpolation (1-D PI) to salt and pepper noise removal. The difference between the proposed approach and median based schemes is that the noise pixel is replaced by an interpolated value instead of median in the window. With 3 x 3 window, two non-noise neighbor pixels are used to estimate the noise pixel by first-order PI. Then the interpolated value takes the place of noise pixel. The proposed approach is justified by two examples. The results indicate that the proposed approach has better visual quality in the examples when compared with conventional median filter and adaptive median filter.