High-fidelity filter based on medians for chemograms

A filtering method using the median in a moving data window and an algorithm avoiding level distortion on the signal peak top and valley bottom are proposed. The distinct feature of a median filter is its opposite effect on peaks with different widths. Peaks with small widths are removed thoroughly, whereas peaks with large widths are kept intact; thus the median filter can eliminate high-frequency noises effectively and keep the low-frequency signals nearly undistorted. Compared with other moving data window filtering methods (in particular, the widely used ones based on weighted sum), the median filter can remove spike noises more effectively because the median is not a compromise between noise and signal, and it is not affected by abnormal data or tendentious noises. The median filter can be connected in series, and its only parameter is easily set for a definite purpose. Keeping high fidelity to the width, height, and position of signal peaks, this method is especially feasible for use as a high-frequency noise eliminator in chemograms sampling data processing.