Title :
An effective resolution definition or how to choose an edge detector, its scale parameter and the threshold?
Author :
Demigny, D. ; Karabernou, M.
Author_Institution :
Equipe Traitement des Images et du Signal, ENSEA-UCP, Cergy Pontoise, France
Abstract :
In this paper, we propose a new quantitative method to choose parameters values of edge detectors based on linear first derivative filters. First, the resolution is explicitly defined as the smallest distance d between edges that the user wants to detect. A new criterion Cd of signal to noise ratio improvement is defined as a function of d. Maximizing this criterion for a given filter defines its best scale parameter value. Then, we analyze quantitatively the effect of the third Canny criterion C3 (mean distance between noise maxima), and the effect of Cd on the number of false detections. Finally, we discuss quantitatively the choice of the scale parameter and threshold as a function of the desired resolution, output quality, and of the input signal to noise ratio
Keywords :
edge detection; filtering theory; image resolution; edge detector; effective resolution definition; false detections; linear first derivative filters; mean distance between noise maxima; output quality; quantitative method; scale parameter; signal to noise ratio improvement; third Canny criterion; threshold; Detectors; Energy resolution; Image edge detection; Noise reduction; Nonlinear filters; Performance analysis; Roads; Signal resolution; Signal to noise ratio; Technological innovation;
Conference_Titel :
Image Processing, 1996. Proceedings., International Conference on
Conference_Location :
Lausanne
Print_ISBN :
0-7803-3259-8
DOI :
10.1109/ICIP.1996.559627