DocumentCode :
2283767
Title :
Role of prefiltering in unsupervised video segmentation
Author :
Karaca, Hnkan M. ; Anarm, E. ; Morgul, Avni
Author_Institution :
Dept. of Electr. & Electron. Eng., Bogazici Univ., Istanbul, Turkey
Volume :
6
fYear :
2000
fDate :
2000
Firstpage :
1999
Abstract :
In this study, we address the problem of prefiltering before segmentation. Prefiltering is quite important since it simplifies the data, producing flat zones without computing the contour information (Salembier and Pardas, 1994). Morphological filters with reconstruction restore the contour of the components. However, covered and uncovered areas in moving objects may be connected during the reconstruction process. In this case, open-close by partial reconstruction filters remove those artifacts in the background. After that, the contours of the regions can be located by the watershed algorithm. Although the number of resulting regions at the end of watershed with open-close by partial reconstruction is more than local averaging, applying a premerging by comparing the difference of mean gray values to these regions results in less number and greater-sized regions. Therefore, open-close filtering makes merging of more regions possible
Keywords :
filtering theory; image reconstruction; image segmentation; image sequences; contour information; image sequence; local averaging; mean gray values; morphological filters; moving objects; partial reconstruction filters; prefiltering; premerging; unsupervised video segmentation; watershed algorithm; Cryptography; Filtering; Filters; Image processing; Image reconstruction; Image restoration; Image segmentation; Image sequences; Merging; Object segmentation;
fLanguage :
English
Publisher :
ieee
Conference_Titel :
Acoustics, Speech, and Signal Processing, 2000. ICASSP '00. Proceedings. 2000 IEEE International Conference on
Conference_Location :
Istanbul
ISSN :
1520-6149
Print_ISBN :
0-7803-6293-4
Type :
conf
DOI :
10.1109/ICASSP.2000.859224
Filename :
859224
Link To Document :
بازگشت