Title :
Rate distortion optimal contour compression using cubic B-splines
Author :
Zaletelj, J. ; Pecci, R. ; Spaan, F. ; Hanjalic, A. ; Lagendijk, R.L.
Author_Institution :
Fac. of Electr. Eng., Univ. of Ljubljana, Ljubljana, Slovenia
Abstract :
Object contours resulting from segmenting images or video frames can be efficiently encoded using B-spline functions. An unsolved problem is how to divide a given contour into segments such that the resulting compression is optimal in rate-distortion sense. In this paper we describe two techniques for finding a close-to-optimal knot assignment. The first technique prunes an accurate B-splines approximation until the desired rate is achieved. The second technique analyzes the curvature function of the original contour to obtain a suboptimal knot assignment. The resulting algorithms are compared experimentally.
Keywords :
data compression; function approximation; image coding; image segmentation; rate distortion theory; splines (mathematics); close-to-optimal knot assignment; cubic B-spline approximation; curvature function analysis; image segmentation; object contour; rate distortion optimal contour compression; suboptimal knot assignment; video frame segmentation; Approximation methods; Bit rate; Image coding; Merging; Rate-distortion; Shape; Splines (mathematics);
Conference_Titel :
Signal Processing Conference (EUSIPCO 1998), 9th European
Conference_Location :
Rhodes
Print_ISBN :
978-960-7620-06-4