DocumentCode :
3058109
Title :
Breakpoint skeletal representation and compression of document images
Author :
Tam, David ; Barrett, William ; Morse, Bryan ; Mortensen, Eric
Author_Institution :
Dept. of Comput. Sci., Brigham Young Univ., Provo, UT, USA
fYear :
1998
fDate :
30 Mar-1 Apr 1998
Firstpage :
575
Abstract :
Summary form only given. We present a new method for representation and (lossy) compression of bitonal document images. The technique extracts a skeletal medial axis from each object using a true Euclidean distance map of the image and then finds piecewise linear breakpoints in the skeleton to create a breakpoint skeletal representation, b.p.s. The b.p.s. is encoded for each object as a set of triples. The original binary object is reconstructed by first reconstructing the skeleton using linear interpolation between breakpoints and then fractionally dilating each point on the skeleton with the (linearly interpolated) radius, ri. For noninteger r i fractional dilation provides a natural antialiasing in the reconstructed image. Breakpoints can be extracted to preserve fine detail or a more coarse representation by tightening or relaxing the pruning radius respectively. If, in extracting breakpoints, the pruning radius is set to zero, the reconstruction is almost lossless, but the compression is worse
Keywords :
data compression; document image processing; feature extraction; image coding; image reconstruction; image representation; interpolation; antialiasing; binary object; bitonal document images; breakpoint skeletal compression; breakpoint skeletal representation; fractional dilation; linear interpolation; lossless image reconstruction; lossy compression; piecewise linear breakpoints; pruning radius; skeletal medial axis extraction; true Euclidean distance map; 1f noise; Computer science; Euclidean distance; Image coding; Image reconstruction; Image segmentation; Interpolation; Piecewise linear techniques; Sampling methods; Skeleton;
fLanguage :
English
Publisher :
ieee
Conference_Titel :
Data Compression Conference, 1998. DCC '98. Proceedings
Conference_Location :
Snowbird, UT
ISSN :
1068-0314
Print_ISBN :
0-8186-8406-2
Type :
conf
DOI :
10.1109/DCC.1998.672317
Filename :
672317
Link To Document :
بازگشت