DocumentCode
2683160
Title
Edge detection on object color
Author
Tsang, Peter W M ; Tsang, W.H.
Author_Institution
Dept. of Electron. Eng., City Univ. of Hong Kong, Hong Kong
Volume
3
fYear
1996
fDate
16-19 Sep 1996
Firstpage
1049
Abstract
In this paper a novel scheme for color edge detection is presented to minimize the false edge detection caused by specular reflection. The method is based on the dichromatic reflection model proposed by Shafer (1985) to describe the object reflection properties in the HSV color space, and is different from existing color edge detection techniques in three major aspects. First, our analysis accounts for multiple light sources and is less restrictive than techniques which are based on the assumption of single light source. Second, we have incorporated the V-component on top of the hue H to include the shading information which is absent in the latter. This effectively added in a new dimension in separating the regions which are similar in hue, but different in shade. Third, we had discovered that a single color edge operator is inadequate for detecting discontinuities in all conditions with equal precision. The V-component, for example, is more relevant than H in describing the edge contour when S is low, and vice versa. In between these two extremes, both components are found to have equal significance
Keywords
edge detection; feature extraction; image colour analysis; image segmentation; light reflection; HSV color space; V-component; color images; dichromatic reflection model; discontinuities detection; edge contour; edge detection; gradient based edge operator; hue; multiple light sources; object reflection properties; regions separation; shading information; specular reflection; Color; Colored noise; Computer vision; Data mining; Dielectric materials; Image edge detection; Image segmentation; Light sources; Optical reflection; Solid modeling;
fLanguage
English
Publisher
ieee
Conference_Titel
Image Processing, 1996. Proceedings., International Conference on
Conference_Location
Lausanne
Print_ISBN
0-7803-3259-8
Type
conf
DOI
10.1109/ICIP.1996.561021
Filename
561021
Link To Document