ABSTRACT
As already stated in the opening chapter the fundamental problem in data smoothing is the choice of input parameter. Too large a parameter will smooth out many important features (corners, vertices, zero-crossings) and too small a parameter will produce many redundant features. This is the fundamental problem of scale because features appearing on any curve vary enormously in size and extent and there is seldom any basis of choosing a particular parameter for a particular feature size. This problem may be resolved by automatic parameter tuning. The parameter size can be tuned on the basis of the local properties of the curve using suitable criterion function. Teh and Chin [170] resolve this problem using the chord length and perpendicular distance as the criterion function to determine the region of support. In Chapter 9 and 10 we too have developed two such schemes in which automatic parameter tuning on the basis of the local properties of a curve is suggested. Another approach to solution of the problem of scale is scale space analysis.
