Method, apparatus and computer program product for image segmentation

摘要

In an example embodiment, a method, apparatus and computer program product are provided. The method includes facilitating receipt of a plurality of superpixels of an input image and an object selection input for selecting part of a foreground object in the input image. The method includes determining first set of superpixels as first set of foreground seeds and second set of superpixels as first set of background seeds. The method includes filtering the first set of background seeds to generate second set of background seeds based on geodesic distances. The method includes adding superpixels of first set of unclassified superpixels to the second set of background seeds to generate third set of background seeds, and adding superpixels of second set of unclassified superpixels to the third set of background seeds to generate fourth set of background seeds based on applying a classifier on the second set of unclassified superpixels.

主权项

1. A method comprising:
facilitating receipt of a plurality of superpixels of an input image; facilitating receipt of an object selection input for selecting at least a part of a foreground object in the input image; determining a first set of superpixels of the plurality of superpixels as a first set of foreground seeds based on the object selection input; determining a second set of superpixels of the plurality of superpixels as a first set of background seeds, the second set of superpixels comprised within a pre-determined region of the input image; filtering the first set of background seeds to generate a second set of background seeds based on geodesic distances between the first set of foreground seeds and the first set of background seeds, a geodesic distance of the geodesic distances between a foreground seed and a background seed being based on a spatial distance and an appearance metric between the foreground seed and the background seed; adding one or more superpixels of a first set of unclassified superpixels to the second set of background seeds to generate a third set of background seeds, the one or more superpixels of the first set of unclassified superpixels selected based on geodesic distances of superpixels of the first set of unclassified superpixels from the first set of foreground seeds and from the second set of background seeds, wherein the first set of unclassified superpixels comprises superpixels of the plurality of superpixels other than the first set of foreground seeds and the second set of background seeds; and adding one or more superpixels of a second set of unclassified superpixels to the third set of background seeds to generate a fourth set of background seeds based on applying a classifier on the second set of unclassified superpixels, wherein the second set of unclassified superpixels comprises superpixels of the plurality of superpixels other than the first set of foreground seeds and the third set of background seeds.