Piecewise Feature Extraction and Artificial Neural Networks: An Approach towards Curve Reconstruction

Background/Objectives: Reconstruction of shapes from unorganized data points is a problem with lot of practical significance in which a piecewise linear approximation to the shape is computed from the sample of the unknown shape. Methods/Statistical Analysis: An approach based on reconstruction of curves by using the feature points as control points is presented. At the transmission end the curve is represented using the feature points and at the receiver side the reconstruction of curve is established by optimizing the parameters of a radial basis function (RBF) neural network. Findings: The method reduces the complexity in terms of time and space. In other words it reduces the informational complexity of the RBF network for the problem of curve reconstruction. It also allows for noise in the data by using the inherent capabilities of a RBF neural network. Applications/Improvements: Scenes generated by modeling and animation using multimedia techniques contain curves and surfaces and thus the presented approach is useful in efficient transmission of images and video sequences.