The Length of Digital Curves

The paper discusses one of the elementary subjects in image analysis: how to measure the length of a digital curve? A digital curve in the plane is defined to be a cycle given either as an alternating sequence of vertices and edges, or an alternating sequence of edges and squares. The paper reports about two length estimators, one based on the partition of a frontier of a simply-connected isothetic polygon into digital straight segments, and one based on calculating the minimum-length polygon within an open boundary of a simply-connected isothetic polygon. Both techniques are known to be implementations of convergent estimators of the perimeter of bounded, polygonal or smooth convex sets in the euclidean plane. For each technique a linear-time algorithm is specified, and both algorithms are compared with respect to convergence speed and number of generated segments. The experiments show convergent behavior also for perimeters of non-convex bounded subsets of the euclidean plane. 1 The University of Auckland, Computer Science Department, CITR, Tamaki Campus, Glen Innes, Auckland, New Zealand The Length of Digital Curves Reinhard Klette and Ben Yip CITR Tamaki, University of Auckland Tamaki Campus, Building 731, Auckland, New Zealand Abstract. The paper discusses one of the elementary subjects in image analysis: how to measure the length of a digital curve? A digital curve in the plane is de ned to be a cycle given either as an alternating sequence of vertices and edges, or an alternating sequence of edges and squares. The paper reports about two length estimators, one based on the partition of a frontier of a simply-connected isothetic polygon into digital straight segments, and one based on calculating the minimumlength polygon within an open boundary of a simply-connected isothetic polygon. Both techniques are known to be implementations of convergent estimators of the perimeter of bounded, polygonal or smooth convex sets in the euclidean plane. For each technique a linear-time algorithm is speci ed, and both algorithms are compared with respect to convergence speed and number of generated segments. The experiments show convergent behavior also for perimeters of non-convex bounded subsets of the euclidean plane. The paper discusses one of the elementary subjects in image analysis: how to measure the length of a digital curve? A digital curve in the plane is de ned to be a cycle given either as an alternating sequence of vertices and edges, or an alternating sequence of edges and squares. The paper reports about two length estimators, one based on the partition of a frontier of a simply-connected isothetic polygon into digital straight segments, and one based on calculating the minimumlength polygon within an open boundary of a simply-connected isothetic polygon. Both techniques are known to be implementations of convergent estimators of the perimeter of bounded, polygonal or smooth convex sets in the euclidean plane. For each technique a linear-time algorithm is speci ed, and both algorithms are compared with respect to convergence speed and number of generated segments. The experiments show convergent behavior also for perimeters of non-convex bounded subsets of the euclidean plane.

[1]  E. Creutzburg,et al.  Geometric methods for on-line recognition of digital straight-line segments , 1988 .

[2]  Bryan H. Bunch Mathematical Fallacies and Paradoxes , 1982 .

[3]  Krzysztof P. Rybakowski,et al.  An abstract approach to smoothness of invariant manifolds , 1993 .

[4]  Reinhard Klette,et al.  Linear Time Calculation of 2D Shortest Polygonal Jordan Curves , 1998 .

[5]  V. Kovalevsky,et al.  A New Concept for Digital Geometry , 1994 .

[6]  Z. Kulpa On the properties of discrete circles, rings, and disks , 1979 .

[7]  AZRIEL ROSENFELD,et al.  Digital Straight Line Segments , 1974, IEEE Transactions on Computers.

[8]  Klaus Voss,et al.  Digitalisierungseffekte in der Automatischen Bildverarbeitung , 1975, J. Inf. Process. Cybern..

[9]  Marcel Worring,et al.  Digitized Circular Arcs: Characterization and Parameter Estimation , 1995, IEEE Trans. Pattern Anal. Mach. Intell..

[10]  Vladimir Kovalevsky Applications of Digital Straight Segments to Economical Image Encoding , 1997, DGCI.

[11]  John N. Tsitsiklis,et al.  Local Versus Nonlocal Computation of Length of Digitized Curves , 1994, IEEE Trans. Pattern Anal. Mach. Intell..

[12]  Reinhard Klette,et al.  Length estimation of digital curves , 1999, Optics & Photonics.

[13]  Reinhard Klette,et al.  Topology of grid continua , 1998, Optics & Photonics.

[14]  V. A. Kovalevsky,et al.  Finite Topology and Image Analysis , 1992 .

[15]  Fridrich Sloboda,et al.  The minimum perimeter polygon and its application , 1992, Theoretical Foundations of Computer Vision.

[16]  K. Voss Discrete Images, Objects, and Functions in Zn , 1993 .

[17]  Vladimir A. Kovalevsky,et al.  Finite topology as applied to image analysis , 1989, Comput. Vis. Graph. Image Process..

[18]  Roland T. Chin,et al.  On digital approximation of moment invariants , 1986, Computer Vision Graphics and Image Processing.

[19]  Reinhard Klette,et al.  Determination of the Convex Hull of a Finite Set of Planar Points Within Linear Time , 1981, J. Inf. Process. Cybern..

[20]  Gabor T. Herman,et al.  Boundary detection in 3-dimensions with a medical application , 1981, COMG.

[21]  Behrooz Kamgar-Parsi,et al.  Evaluation of quantization error in computer vision , 1988, Proceedings CVPR '88: The Computer Society Conference on Computer Vision and Pattern Recognition.

[22]  V. Kovalevsky,et al.  New definition and fast recognition of digital straight segments and arcs , 1990, [1990] Proceedings. 10th International Conference on Pattern Recognition.

[23]  Azriel Rosenfeld,et al.  Advances in Digital and Computational Geometry , 1999 .