| 主权项 |
1. A method of segmenting an image using a computer, the method comprising:
using a set of user input foreground markers identifying a foreground of an image to be segmented, defining Nf foreground regions and Nb background regions, where Nf is an integer greater than zero and Nb is an integer at least zero; for each respective border region that defines a border of the image and is not one of the Nf foreground regions, performing a multi-level assessment of the respective border region to determine whether the respective border region is a possible foreground region, where each level of the multi-level assessment compares color similarity, with respect to a threshold that is specific for that level, between the respective border region and any of the Nf foreground regions or of a region that satisfied the threshold of a previous level of the multi-level assessment; from the multi-level assessment and the foreground markers, classifying each region of the image to be segmented as one of a strong foreground region, a strong background region or an ambiguous region; classifying each of the respective ambiguous regions as a weak foreground region or a weak background region; and merging the strong foreground regions with the weak foreground regions and output therefrom, to a tangible memory or to a graphical display, a foreground image of the image to be segmented, separate and distinct from any region classified as one of the strong background regions or weak background regions. |
