IP Library Granted Patent US 10,102,450
Granted Patent B2
US 10,102,450 · App. 14/784,026 · Granted Oct 16, 2018

Superpixel generation with improved spatial coherency

View Patent ↗
Loading inventors, assignments & file history…
Monitor This Case
Get email alerts when status or documents change.
Order Certified Copies
Most orders are placed with the USPTO same day — all within 24 business hours.
Order via The Patent Place →
Pre-filled with this patent's details
Quick Facts
Patent No.
US 10,102,450
App. No.
14/784,026
Granted
Oct 16, 2018
Kind
B2
Abstract

A method for generating superpixels for an image comprising pixels and an apparatus configured to perform the method. A clustering stage clusters the pixels into initial clusters. A determination stage then selects spatially coherent parts of the clusters and determines unconnected fragments of the clusters. Pixels of these unconnected fragments are iteratively assigned to adjacent clusters by an assignment stage using a contour evolution process.

Claims (26)

1. A method for generating superpixels for an image having pixels, the method comprising:

clustering the pixels into initial clusters;

determining one spatially coherent part of each of the clusters, all other parts of each cluster being considered as unconnected fragments and being set to unassigned and marked as mutable; and

assigning pixels of the unconnected fragments to adjacent clusters using a contour evolution process based on the spatially coherent parts of the clusters as initial adjacent clusters that processes only the pixels that are in a region marked as mutable.

2. The method according to claim 1 , wherein in each iteration of the contour evolution cluster assignments for pixels at a boundary within a region marked as mutable can be changed.

3. The method according to claim 2 , wherein the cluster assignment of a pixel is changed if the pixel has no assignment.

4. The method according to claim 3 , wherein the pixel is assigned to that cluster of one of its adjacent pixels which minimizes an energy term used for the initial clustering.

5. The method according to claim 1 , wherein an assignment of a pixel is changed to the cluster of one of its adjacent pixels if an energy term is smaller for this cluster than for the cluster to which the pixel was previously assigned.

6. An apparatus configured to generate superpixels for an image having pixels, wherein the apparatus comprises:

at least one processor configured to:

cluster the pixels into initial clusters;

determine one spatially coherent part of each of the clusters, all of the other parts of each cluster being considered as unconnected fragments and being set to unassigned and marked as mutable; and

assign pixels of the unconnected fragments to adjacent clusters, using a contour evolution process based on the spatially coherent parts of the clusters as initial adjacent clusters that processes only the pixels that are in a region marked as mutable.

7. The apparatus according to claim 6 , wherein in each iteration of the contour evolution cluster assignments for pixels at a boundary within a region marked as mutable can be changed.

8. The apparatus according to claim 7 , wherein said at least one processor is configured to change the cluster assignment of a pixel if the pixel has no assignment.

9. The apparatus according to claim 8 , wherein said at least one processor is configured to assign the pixel to that cluster of one of its adjacent pixels which minimizes an energy term used for the initial clustering.

10. The apparatus according to claim 6 , wherein said at least one processor is configured to change an assignment of a pixel to the cluster of one of its adjacent pixels if an energy term is smaller for this cluster than for the cluster to which the pixel was previously assigned.

11. A computer readable non-transitory storage medium having stored therein instructions enabling generating superpixels for an image having pixels, which, when executed by a computer, cause the computer to:

cluster the pixels into initial clusters;

determine one spatially coherent part of each of the clusters;

determine unconnected fragments of the clusters, all other parts of each cluster being considered as unconnected fragments and being set to unassigned and marked as mutable; and

assign pixels of the unconnected fragments to adjacent clusters using a contour evolution process based on the spatially coherent parts of the clusters as initial adjacent clusters that processes only the pixels that are in a region marked as mutable.

12. The computer readable non-transitory storage medium according to claim 11 , wherein in each iteration of the contour evolution cluster assignments for pixels at a boundary within a region marked as mutable can be changed.

13. The computer readable non-transitory storage medium according to claim 12 , wherein the instructions cause the computer to change the cluster assignment of a pixel if the pixel has no assignment.

14. The computer readable non-transitory storage medium according to claim 13 , wherein the instructions cause the computer to assign the pixel to that cluster of one of its adjacent pixels which minimizes an energy term used for the initial clustering.

15. The computer readable non-transitory storage medium according to claim 11 , wherein the instructions cause the computer to change an assignment of a pixel to the cluster of one of its adjacent pixels if an energy term is smaller for this cluster than for the cluster to which the pixel was previously assigned.

Assignments (2)
ASSIGNMENT OF ASSIGNOR'S INTEREST Recorded Oct 23, 2018
From: THOMSON LICENSING
To: INTERDIGITAL VC HOLDINGS, INC.
Reel/Frame 047289/0698 →
ASSIGNMENT OF ASSIGNOR'S INTEREST Recorded Apr 18, 2018
From: JACHALSKY, JOERN; RESO, MATTHIAS; ROSENHAHN, BODO
To: THOMSON LICENSING
Reel/Frame 045576/0247 →