US 8,180,154 B2Granted

Method and apparatus for region-based segmentation image processing using region mapping

AssigneeTechFinity, Inc.

InventorsJarrell D. Collier, Michael P. Davenport, H. K. John Armenian

Life of this patent

Estimating the term…

Loading patent…

Abstract

Object detection in an image using an image processing approach in which the image is updated based on an energy function. In one aspect, the exemplary image update process attempts to homogenize each region by associating each pixel with a particular region such that the total level of energy for the image based on the pixel located in that region is minimized with respect to the energy function. For example, the method of object detection in an image having a plurality of pixels by image segmentation includes dividing the image into a plurality of regions; assigning each pixel to one of the plurality of regions based on a characteristic; performing an energy reduction on the image based on a region reassignment of at least one pixel; and producing an output image based on the energy reduction.

Claims25

Dependent claims

A method of object detection in an image comprising a plurality of pixels by image segmentation comprising:

dividing the image into a plurality of regions;
assigning each pixel to one of the plurality of regions based on a characteristic;
performing an energy reduction on the image based on a region reassignment of at least one pixel and a comparison based on regions to which pixels surrounding the at least one pixel are assigned, wherein reducing the energy of the image comprises:
a) selecting a candidate pixel;
b) determining a change in energy of the image comprising a difference in energy of the image based on a movement of the candidate pixel from a current location in a current region to a new location in a new region, wherein determining the change in energy comprises:
i) determining a plurality of energy levels for the image, wherein each energy level is determined as if the candidate pixel has already been assigned to a different region, and wherein the energy level comprises a sum of a plurality of region energies, with each region energy comprising a weighted sum of the intensity of the region and a length of at least a portion of a perimeter of the region;
and
ii) identifying the new region as being one with a lowest energy level in the plurality of energy levels;
and
c) reassigning the candidate pixel to the new region if there is a decrease in energy of the image;
and
producing an output image based on the energy reduction.

9 more elements

An apparatus for object detection in an image comprising a plurality of pixels by image segmentation comprising:

means for dividing the image into a plurality of regions;
means for assigning each pixel to one of the plurality of regions based on a characteristic;
means for performing an energy reduction on the image based on a region reassignment of at least one pixel and a comparison based on regions to which pixels surrounding the at least one pixel are assigned, wherein the means for reducing the energy of the image comprises:
a) means for selecting a candidate pixel;
b) means for determining a change in energy of the image comprising a difference in energy of the image based on a movement of the candidate pixel from a current location in a current region to a new location in a new region, wherein the means for determining the change in energy comprises:
i) means for determining a plurality of energy levels for the image, wherein each energy level is determined as if the candidate pixel has already been assigned to a different region, and wherein the energy level comprises a sum of a plurality of region energies, with each region energy comprising a weighted sum of the intensity of the region and a length of at least a portion of a perimeter of the region;
and
ii) means for identifying the new region as being one with a lowest energy level in the plurality of energy levels;
and
c) means for reassigning the candidate pixel to the new region if there is a decrease in energy of the image;
and
means for producing an output image based on the energy reduction.

9 more elements

A computer program product for object detection in an image comprising a plurality of pixels by image segmentation comprising:

a non-transitory computer readable medium comprising codes executable to:
divide the image into a plurality of regions;
assign each pixel to one of the plurality of regions based on a characteristic;
perform an energy reduction on the image based on a region reassignment of at least one pixel and a comparison based on regions to which pixels surrounding the at least one pixel are assigned, wherein the codes executable to reduce the energy of the image comprises codes executable to:
a) select a candidate pixel;
b) determine a change in energy of the image comprising a difference in energy of the image based on a movement of the candidate pixel from a current location in a current region to a new location in a new region, wherein the codes executable to determine the change in energy comprises codes executable to:
i) determine a plurality of energy levels for the image, wherein each energy level is determined as if the candidate pixel has already been assigned to a different region, and wherein the energy level comprises a sum of a plurality of region energies, with each region energy comprising a weighted sum of the intensity of the region and a length of at least a portion of a perimeter of the region;
and
ii) identify the new region as being one with a lowest energy level in the plurality of energy levels;
and
c) reassign the candidate pixel to the new region if there is a decrease in energy of the image;
and
produce an output image based on the energy reduction.

10 more elements

Dependencies are inferred from the claim text.