DocumentCode
3457485
Title
Frequency tuned salient edge detection
Author
Saber, Yusuf ; Kyan, M.
Author_Institution
Dept. of Electr. & Comput. Eng., Ryerson Univ., Toronto, ON, Canada
fYear
2011
fDate
8-11 May 2011
Abstract
This paper presents a novel method for salient edge detection based on the frequency tuning principle. Limiting the operation to a single wavelet sub-band, the frequency components of the wavelet are analyzed. The average value in each frequency component (horizontal, vertical, and diagonal) is computed and subtracted from each pixel in its respective component. This creates a variance which allows the salient edges to pop out more significantly. Reconstructing the wavelet using only the frequency components yields a strong response to salient edges. Comparisons are given to classic salient edge detection methods.
Keywords
edge detection; wavelet transforms; frequency component; frequency tuning principle; salient edge detection; single wavelet subband; wavelet reconstruction; Conferences; Equations; Feature extraction; Frequency domain analysis; Image edge detection; Image reconstruction; Strontium; Saliency; edge detection; spectral residual; visual attention; wavelets;
fLanguage
English
Publisher
ieee
Conference_Titel
Electrical and Computer Engineering (CCECE), 2011 24th Canadian Conference on
Conference_Location
Niagara Falls, ON
ISSN
0840-7789
Print_ISBN
978-1-4244-9788-1
Electronic_ISBN
0840-7789
Type
conf
DOI
10.1109/CCECE.2011.6030664
Filename
6030664
Link To Document