Abstract: With regard to the difficult problem of extracting the accurate blend feature parameters from a set of discrete points, a new data processing method based on the theory of statistics and probability is proposed. This method, including four steps：initial regression analysis, hypothesis test, removing of noise points, regression analysis, can accurately extract out the blend feature from the coarse parameters. The practical application proved that the theory of statistics and probability play a very important role in reverse engineering on feature extraction, analysis and representation.

Key words: Blend feature, Reverse engineering, Statistics