Method and apparatus for segmenting images using constant-time deformable contours
First Claim
Patent Images
1. A method of segmenting a set of images comprising the steps of:
- a) providing a set of search images, each search image containing a view of a target object;
b) providing a priori domain knowledge corresponding to the set of search images;
c) selecting a training image from the set of search images;
d) obtaining at least one training contour from the selected training image, the training contour representing an outline of the target object in the selected training image;
e) utilizing the selected training contour to extract training information corresponding to the target object in the selected training image;
f) associating the training contour with a selected search image of the set of images, the training contour defining a search contour;
g) deforming the search contour relative to the selected search image and maintaining predetermined information corresponding to the deformed search contour;
h) utilizing the training information and the a priori domain knowledge to identify regions of the deformed search contour that correspond within predetermined limits to an outline of the target object in the selected search image; and
i) repeating steps (c) through (h) for each image in the set of search images so as to segment all of the images in the set of search images.
0 Assignments
0 Petitions
Accused Products
Abstract
An image segmentation system uses deformable contours to automatically identify the border of a target object in each image of a set of images. The majority of contour deformations, or moves, are performed in O(1) time. Each contour uses both global and local geometric information and pixel intensity information extracted from a training contour, as well as a priori domain knowledge, to find the outline of the target object.
-
Citations
24 Claims
-
1. A method of segmenting a set of images comprising the steps of:
-
a) providing a set of search images, each search image containing a view of a target object; b) providing a priori domain knowledge corresponding to the set of search images; c) selecting a training image from the set of search images; d) obtaining at least one training contour from the selected training image, the training contour representing an outline of the target object in the selected training image; e) utilizing the selected training contour to extract training information corresponding to the target object in the selected training image; f) associating the training contour with a selected search image of the set of images, the training contour defining a search contour; g) deforming the search contour relative to the selected search image and maintaining predetermined information corresponding to the deformed search contour; h) utilizing the training information and the a priori domain knowledge to identify regions of the deformed search contour that correspond within predetermined limits to an outline of the target object in the selected search image; and i) repeating steps (c) through (h) for each image in the set of search images so as to segment all of the images in the set of search images. - 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. A method of segmenting a set of images comprising the steps of
a) providing a set of search images, each search image containing a view of a target object; -
b) providing a priori domain knowledge corresponding to the set of search images; c) selecting a training image from the set of search images; d) obtaining a training contour from the selected training image, the training contour representing an outline of the target object in the selected training image; e) utilizing the selected training contour to extract training information corresponding to the target object in the selected training image; f) associating the training contour with a selected search image of the set of images, the training contour defining a search contour; g) deforming the search contour relative to the selected search image and maintaining predetermined information corresponding to the deformed search contour; h) utilizing the training information and the a priori domain knowledge to identify regions of the deformed search contour that correspond within predetermined limits to an outline of the target object in the selected search image; i) utilizing selected regions of the search contour and the a priori domain knowledge to fashion a final segmentation contour such that the final segmentation contour forms an outline of the target object in the selected search image; and j) repeating steps (c) through (i) for each image in the set of search images so as to segment all of the images in the set of search images.
-
Specification