Data fusion method for 3-D object reconstruction from range images

A method of data fusion from a set of range images for 3-D object surface reconstruction is presented. The two major steps (multiview registration and data integration) of data fusion are carefully discussed. Firstly, the range images taken from multiple views are accurately registered through a set of translation and rotation matrices whose coefficients are carefully calculated through the developed methodology. Then, three criteria for overlapping-data elimination are provided as the foundation of data integration. Compared with the most other methods, which mesh all multiple views or compute an implicit surface function for the object before integrating the data, our integration method manipulates surface data directly, thus providing a straightforward way for overlap removal. A surface-based smoothing filter and a resampling operation are also developed for data quality improvement and data size reduction. The approach is applied to various range data sets of objects with different geometric shapes. The experimental results demonstrate the efficiency and applicability of the proposed method. (c) 2005 Society of Photo-Optical Instrumentation Engineers.