Accuracy enhancement of 3D profilometric human face reconstruction using undecimated wavelet analysis

摘要

Three-dimensional profilometric object reconstruction is a challenging topic; among the various methods available, we implement the line projection technique, which has superiorities over other methods. In order to increase the accuracy of measurement, a wavelet transform analysis is used in two stages of denoising and phase extraction. Because of the denoising capability and multiresolution characteristics of wavelet transforms, we employ an undecimated wavelet transform for noise reduction and a continuous wavelet transform in the phase extraction stage. The aim is to add a preprocessing stage of denoising based on the undecimated wavelet transform to enhance the accuracy of measurement in noisy patterns. The experimental results on the human face as a complex object demonstrate that the combination of undecimated and continuous wavelet transforms could increase measurement accuracy in noise-contaminated patterns.