Method for finding contours in an image of an object
First Claim
1. A method of identifying contours in an image comprising the steps of:
- extracting features from said image;
forming chains from localized subsets of said features;
computing the values of one or more hash functions for contour segments, wherein each said contour segment comprises at least one of said chains;
identifying, using said values for said contour segments, candidate contour segments for joining;
selecting contour segments for joining from said candidates; and
joining said selected contour segments to form joined contour segments.
2 Assignments
0 Petitions
Accused Products
Abstract
A method is provided for robustly and efficiently identifying contours in an image of an object. Features are extracted from an image, typically using a Sobel edge detector or a set of large oriented filters, and are locally chained to form contour segments of particular character, such as line segments. A spatial hash table is used to access the contour segments according to the location of their endpoints. Candidate sets for joining to grow contours are reduced by considering only segments having ends in adjacent spatial hash bins. Contour constraints are applied to appropriately choose among the candidates to form contours. The method facilitates robust and efficient scratch detection of fiber-optic end faces, or of any surface that may have smoothly curved scratches.
75 Citations
30 Claims
-
1. A method of identifying contours in an image comprising the steps of:
-
extracting features from said image;
forming chains from localized subsets of said features;
computing the values of one or more hash functions for contour segments, wherein each said contour segment comprises at least one of said chains;
identifying, using said values for said contour segments, candidate contour segments for joining;
selecting contour segments for joining from said candidates; and
joining said selected contour segments to form joined contour segments. - View Dependent Claims (2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29)
-
-
30. A method of identifying scratch defects in an image of a fiber end comprising the steps of:
-
extracting features from said image using oriented filters;
forming chains from localized subsets of said features using 3 by 3 connectivity rules;
constructing a spatial hash table for contour segments using contour segment endpoints, wherein each said contour segment comprises at least one of said chains;
identifying, using said spatial hash table, candidate contour segments for joining;
selecting contour segments for joining from said candidates; and
joining said selected contour segments to form identified scratch defects.
-
Specification