Authors: Žunić, Joviša 
Acketa, Dragan
Title: Least squares fitting of digital polynomial segments
Journal: Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume: 1176
First page: 17
Last page: 23
Conference: 6th International Workshop on Discrete Geometry for Computer Imagery, DGCI 1996; Lyon; France; 13 November 1996 through 15 November 1996
Issue Date: 1-Jan-1996
ISBN: 978-3-540-62005-1
ISSN: 0302-9743
DOI: 10.1007/3-540-62005-2_2
It is proved that digital polynomial segments and their least squares polynomial fits are in one-to-one correspondence. This enables an efficient representation of digital polynomial segments by n+3 parameters, under the condition that an upper bound, say n, for the degrees of the digitized polynomials is assumed. One of such representations is (x 1, m, an, an−1,…, a 0), where x 1 and m are the x-coordinate of the left endpoint and the number of digital points, respectively, while a n, a n−1,..., a 0 are the coefficients of the least squares polynomial fit Y=a nXn+an− 1Xn−1+ ...+a0, for a given digital polynomial segment.
Keywords: Coding | Computer vision | Digital polynomial segment | Image processing | Least squares fitting
Publisher: Springer Link

Show full item record


checked on May 28, 2024

Page view(s)

checked on May 9, 2024

Google ScholarTM




Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.