While various approaches are suggested in the literature to describe and generalize relaxation processes concerning to several objectives, the wider problem addressed here is to find the best-suited relaxation process for a given assignment problem, or better still, to construct a task-dependent relaxation process. For this, we develop a general framework for the theoretical foundations of relaxation processes in pattern recognition. The resulting structure enables (1) a description of all known relaxation processes in general terms and (2) the design of task-dependent relaxation processes. We show that the well-known standard relaxation formulas verify our approach. Referring to the common problem of generating a generalized description of a contour we demonstrate the applicability of the suggested generalization in detail. Important characteristics of the constructed task-dependent relaxation process are: (1) the independency of the segmentation from any parameters, (2) the invariance to geometric transformations, (3) the simplicity, and (4) efficiency.

The ability to construct CAD or other object models from edge and range data has a fundamental meaning in building a recognition and positioning system. While the problem of model fitting has been successfully addressed, the problem of efficient high accuracy and stability of the fitting is still an open problem. In the past researchers have used approximate distance functions rather than the real Euclidean distance because of computational efficiency. We now feel that machine speeds are sufficient to ask whether it is worth considering Euclidean fitting again. This paper address the problem of estimation of elliptical cylinder and cone surfaces to 3D data by a constrained Euclidean fitting. We study and compare the performance of various distance functions in terms of correctness, robustness and pose invariance, and present our results improving known fitting methods by closed form expressions of the real Euclidean distance. 1