Previous collaborative studies have shown that fringe patterns by optical interference way are used as a direct diagnostic tool for qualitative analysis of defects in Culture Heritage. Nevertheless, the complexity of the fringe patterns prevents the accurate quantification of defects. In this study, a phase separation method based on the weighted least-squares algorithm is proposed to conveniently isolate the local phase distribution from the whole, complex fringe pattern. This method is examined to accurately quantify the deformation and strain distribution of visible and invisible defects. In experimental work, the Non-destructive Acoustic excitation and Digital Holography (NDA-DH) detection is employed for the superficial and subsurface defects in the Ruting murals of the Forbidden City, Beijing. Under the impact of fundamental human voice, the frequency range of the acoustic excitation was considered from 100–800 Hz. All results show that the suggested phase separation method allows direct observation of defects as well as quantified surface deformation providing a precious documentation to the conservators.