Using savitzky Golay to achieve data smoothing filtering, the principle is to take n points around the filtering point, a total of N + 1 points for fitting, and then use the least square method to determine the correlation coefficient, so as to get the final value of the filtering point. It can be extended to 2D applications