This paper proposes an adaptive separation method based on dual tree complex wavelet transform (DT-CWT) according to the polished spherical precision optics surface characteristics. DTげCWT exhibited shift invariance superiority, directional selectivity and limited redundancy, and we intensified its distinguishing ability by introducing the adaptive impact factor during refactoring. Finally, we separated various surface components of the measured polishing optical surfaces effectively and characterized their parameters. Experiments showed that the DT-CWT adaptive separation method was more effective in separation and more convenient for subsequent identification and assessment work, compared with the DT-CWT direct separation method.