SUBSAMPLING FOR SURFACE FITTING - A WAVELET TRANSFORM APPLICATION

A 3-D range sensor using triangulation collects large files of range d ata for an object surface. The available range data may be reduced to a more manageable set for surface reconstruction by using multiresolut ion analysis and wavelet transformation. This reduction of the data fi le size may be achieved without losing the essential characteristics o f the surface by imposing proper requirements on the scaling function of the wavelet transform. These requirements are determined by conside ring either the type of scanned surface or the reconstruction method u sed here: a polyhedral surface fitting. Wavelet transformations corres ponding to different scaling functions are used to produce reduced dat a sets from the original one; subsampling is achieved using the Battle -Lemarie linear spline and cubic spline scaling functions and an eight -coefficient Daubechies compactly supported scaling function. A short description of the acquired range data, the application of wavelet tra nsformation to the data, and test results on surface reconstruction ar e presented.