A novel de-noise method based on the grey relational analysis

The images always are corrupted by noises, which have a bad influence on the subsequently processing. There are various types of filters developed to reduce noises. But unfortunately, there are no any kinds of filters, which are perfect for de-noise in any cases. We try our best to resolve this problem, and design a novel filter based on the grey relational analysis (GRA). The flag matrix, which marks the type of every pixel, can be obtained by using GRA between the two sequences in four operators. The noise pixel is removed by the mean of the 3 by 3 window, and non-noise pixel is kept its intensity value. We put an emphasis on removing the Gaussian noise, and compare the three more popular filters with our new filter. We also study the effect on removing the salt and pepper noise. The results show that the approach cannot only remove the Gaussian noise in infrared image or visible light image, but also the non-Gaussian noise, such as salt and pepper noise. In all cases, our new filter gives improved results when compared to some more popular de-noising methods.