Outliers and noise will cause difficulties during processing and using flight data. This paper proposes an outlier detection method based on histogram analysis of wavelet transform residuals, which can locate outliers in time-domain precisely, and can recognize little outliers in succession. Then according to the characteristics of flight data noise and its denoising demand, edge detection is introduced and a two-step denoising method including dyadic wavelet coefficients product and wavelet shrinkage is put forward, which can keep the characteristic of extremum points very well. Finally the experiment shows that the method presented in this paper is effective on flight data cleaning, with which outliers can be recognized exactly and denoising effect is good. The method can also be used for reference in processing other similar data.