Realistic human trunk surfaces from unorganized three-dimensional data

The problems of displaying an irregularly shaped surface created from noisy data manifest themselves in unrealistic and inaccurate surface characteristics. These problems are relevant to medical visualization in which the surface data must appear as close to the actual surface as possible. It is discussed in detail how to transform three dimensional discrete points from the surface of a human trunk into a realistic and accurate surface visualization. Realistic human back surfaces were created that can be used in the assessment of subjects with scoliosis. Methods are presented which minimize the effects of noise and create realistic solid surfaces that can be viewed and manipulated in six degrees of freedom. Automated techniques of boundary detection, triangulation and noise reduction are discussed. This work determines the best techniques for transforming three dimensional back surface data, with the criteria being: realism of presentation and surface accuracy. Specifically, a linear triangulation algorithm is developed which defines the data boundaries. A subsequent non-linear filtering of the triangulated data produces the desired noise reduction effects.