Image segmentation device and image segmentation method

1. An image segmentation method comprising:

displaying, through a display component, an original designation region relative to an image, wherein the image comprises a plurality of pixels;

after the original designation region is displayed, sensing a stroke on the image, wherein the stroke passes through a plurality of pixels of the image;

after the stroke on the image is sensed, determining a regional area according to stroke information corresponding to the stroke on the image, wherein the regional area at least partially overlaps with the original designation region:

generating a plurality of seeds corresponding to the stroke, the regional area, and the original designation region, wherein the seeds comprise foreground seeds and background seeds; if the stroke is a background stroke, the pixels passed by the background stroke are the background seeds; if the stroke is a foreground stroke, the pixels passed by the foreground stroke are the foreground seeds; if the original designation region is used to select a foreground, the pixels around the regional area and within the original designation region are the foreground seeds, and the pixels around the regional area and outside the original designation region are the background seeds; if the original designation region is used to select a background, the pixels around the regional area and within the original designation region are the background seeds, and the pixels around the regional area and outside the original designation region are the foreground seeds; and the pixels in the regional area not passed by the stroke are not classified as the foreground seeds or the background seeds;

performing a segmentation within the regional area according to the seeds to update the original designation region;

wherein under a first condition, the segmentation within the regional area is performed to increase an area of the original designation region within the regional area, and wherein under a second condition, the segmentation within the regional area is performed to decrease the area of the original designation region within the regional area; and

wherein the regional area is determined according to the stroke by extending the regional area from the pixels passed by the stroke for w pixels, wherein w is an integer and is directly proportional to a length of the stroke.

2. The image segmentation method as claimed in claim 1 ,

wherein the segmentation is a graph cut algorithm and is performed according to the seeds.

3. The image segmentation method as claimed in claim 1 , wherein all of pixels passed by the stroke are either within the updated designation region or outside the updated designation region according to whether the stroke is the background stroke or the foreground stroke.

4. An image segmentation device comprising:

a display component configured for displaying an original designation region relative to an image, wherein the image comprises a plurality of pixels;

a user input interface configured for sensing a stroke on the image after the original designation region is displayed, wherein the stroke passes through a plurality of pixels of the image; and

a processing component configured for:

determining a regional area according to stroke information corresponding to the stroke on the image, wherein the regional area at least partially overlaps with the original designation region;

generating a plurality of seeds corresponding to the stroke, the regional area, and the original designation region, wherein the seeds comprise foreground seeds and background seeds; if the stroke is a background stroke, the pixels passed by the background stroke are the background seeds: if the stroke is a foreground stroke, the pixels passed by the foreground stroke are the foreground seeds; if the original designation region is used to select a foreground, the pixels around the regional area and within the original designation region are the foreground seeds, and the pixels around the regional area and outside the original designation region are the background seeds; if the original designation region is used to select a background, the pixels around the regional area and within the original designation region are the background seeds, and the pixels around the regional area and outside the original designation region are the foreground seeds: and the pixels in the regional area not passed by the stroke are not classified as the foreground seeds or the background seeds;

performing a segmentation within the regional area according to the seeds to update the original designation region,

wherein under a first condition, the segmentation within the regional area is performed to increase an area of the original designation region within the regional area,

wherein under a second condition, the segmentation within the regional area is performed to decrease the area of the original designation region within the regional area,

wherein the regional area is determined according to the stroke by extending the regional area from the pixels passed by the stroke for w pixels, wherein w is an integer and is directly proportional to a length of the stroke.

5. The image segmentation device as claimed in claim 4 ,

wherein the segmentation is a graph cut algorithm and is performed according to the seeds.

6. The image segmentation device as claimed in claim 4 , wherein all of pixels passed by the stroke are either within the updated designation region or outside the updated designation region according to whether the stroke is the background stroke or the foreground stroke.

7. An image segmentation method comprising: displaying, through a display component, an original designation region relative to an image, wherein the image comprises a plurality of pixels; sensing a stroke on the image; determining a regional area according to stroke information corresponding to the stroke on the image, wherein the original designation region at least partially overlaps with the regional area; and generating a plurality of seeds corresponding to the stroke, the regional area, and the original designation region, wherein the seeds comprise foreground seeds and background seeds; if the stroke is a background stroke, the pixels passed by the background stroke are the background seeds; if the stroke is a foreground stroke, the pixels passed by the foreground stroke are the foreground seeds; if the original designation region is used to select a foreground, the pixels around the regional area and within the original designation region are the foreground seeds, and the pixels around the regional area and outside the original designation region are the background seeds; if the original designation region is used to select a background, the pixels around the regional area and within the original designation region are the background seeds, and the pixels around the regional area and outside the original designation region are the foreground seeds, and the pixels in the regional area not passed by the stroke are not classified as the foreground seeds or the background seeds; performing a segmentation within the regional area according to the seeds to update a contour of the original designation region within the regional area to generate an updated designation region, wherein a portion of pixels in the regional area are located inside the updated designation region, while another portion of pixels in the regional area are located outside the updated designation region.