Banded graph cut segmentation algorithms with laplacian pyramids

1. A method for segmenting an object of interest from background in a digital image, comprising:

receiving a high-resolution original digital image, of an object of interest disposed in a background in said image;

receiving a set of partial labels with some pixels being labeled “object”, other pixels being labeled “background”, and still other pixels being unlabeled;

reducing in resolution the partially labeled high-resolution image, both the image intensities and the partial labeling, to a coarse version of both;

segmenting said reduced resolution image to convert the partial labeling on the reduced resolution image into a full labeling of the reduced resolution image;

generating from the reduced resolution image, a new, high-resolution image with all pixels therein being labeled as either object or background; and

segmenting said new, high-resolution image to produce a fully labeled, high-resolution image.

2. The method of claim 1 , wherein generating a new, high-resolution image from the reduced resolution image comprises:

switching the labeling of any pixels near a point where the label is switched from object to background to an become unlabeled pixel; and

comparing pixels in the new fine-level image to the corresponding pixels in the original image and anywhere there is a significant discrepancy in intensity between the two corresponding pixels, the labeling associated with the pixels is changed to “unlabeled”.

3. The method of claim 1 , wherein said segmentations are performed using a graph cuts algorithm.

4. A method for segmenting an object of interest from a background in a digital image, comprising:

receiving an original, high-resolution image of an object of interest disposed in background having seeds on the object being labeled as object pixels, seeds on the background being labeled as background pixels, and with uncertain pixels remaining unlabeled;

reducing the resolution of the labeled image;

segmenting the reduced resolution, labeled image;

producing a band around a fine boundary of the unlabeled pixels;

labeling all pixels interior to the band as foreground pixels and pixels exterior of the band as background pixels;

subtracting the labeled image from the original image;

removing the label of any pixel having a magnitude in the subtracted image that exceeds a predetermined threshold;

segmenting original image again to label all unlabeled pixels; and

repeating the above until a predetermined degree of resolution is achieved in an image segmenting the object from the background.

5. The method of claim 4 , wherein said segmentations are performed using a graph cuts algorithm.

6. A computer implemented method for segmenting an object of interest from a background, the method performed by the computer comprising the steps of:

receiving a master, high-resolution digital image of an object disposed within a background;

applying a first band graph cut process to the master image generating a second image with the object being segmented from the background to a first approximation;

comparing the second image with the master image to produce a comparison image with pixels identified by the comparison to be background images being removed from the comparison image.

7. A computer implemented method for segmenting an object of interest from a background, the method performed by the computer comprising the steps of:

generating an original high-resolution image with seeds being placed on the image to indicate regions on an object of interest and background regions;

reducing the resolution of the original high-resolution image;

segmenting the reduced resolution image with pixels being labeled to identify a boundary between the object and the background, all pixels interior to a band around the object being labeled as foreground pixels and all pixels exterior to the band being labeled as background pixels and all pixels unknown to be either inside the band or exterior to the band being designated as unlabeled pixels;

projecting the reduced resolution segmented image with the labeled and unlabeled pixels into a corresponding high-resolution image; and

comparing the reduced resolution image and original high-resolution image.

8. The method recited in claim 7 wherein the comparing comprises:

subtracting the reduced resolution image from the original high-resolution image; and

removing labeled pixels from a resulting, subtracted, image that are outside the bands.