Methods and systems of performing video object segmentation
US-2017236290-A1 · Aug 17, 2017 · US
US10078886B2 · US · B2
| Field | Value |
|---|---|
| Publication number | US-10078886-B2 |
| Application number | US-201715455849-A |
| Country | US |
| Kind code | B2 |
| Filing date | Mar 10, 2017 |
| Priority date | Mar 22, 2016 |
| Publication date | Sep 18, 2018 |
| Grant date | Sep 18, 2018 |
A practical reading order for non-experts. Skip the full description unless you need deep technical detail.
What the patent document calls the invention.
A short plain-language summary of the technical disclosure.
Who owns or filed the patent and who is credited as inventor.
Filing, priority, publication, and grant dates set the timeline.
The legal scope of protection — read this for what is actually claimed.
Technology tags used to group this patent with similar filings.
Prior art links and similar publications in this corpus.
Official abstract text for this publication.
A method for adjusting an image using message passing comprises associating each pixel of an image with a node of a graph and one or more cliques of nodes, determining for a node of the graph a respective set of possible pixel labels for which a unary potential is known, computing for that node a unary potential of a possible pixel label for which the unary potential is unknown, adjusting a clique potential associated with each clique to which that node belongs based on the unary potentials, and adjusting, based on the adjusted clique potential associated with each clique to which that node belongs, at least one of the messages between that node and the other nodes of each clique. Once a convergence criterion is met, an adjusted image is produced having pixel labels determined from the adjusted messages.
Opening claim text (preview).
The invention claimed is: 1. A computer implemented method for producing an adjusted image using message passing, the method comprising performing the following steps: i) receiving an input image that comprises a plurality of pixels, each pixel having an input value and each input value corresponding to one of a number of possible pixel labels; ii) associating each pixel with a node of a graph and one or more cliques of nodes, wherein each possible pixel label is associated with a unary potential, and wherein, for each permutation of possible pixel labels of the pixels associated with the nodes of each clique, a clique potential is associated; iii) for each clique, initialising at least one message between each pair of nodes of that clique; iv) for each node: a) determining, for that node, a respective set of possible pixel labels for which the unary potential is known; b) computing a unary potential of a selected possible pixel label for which the unary potential is unknown and updating the respective set of possible pixel labels to include the selected possible pixel label; c) for each clique to which that node belongs, based on the unary potentials associated with the updated respective set of possible pixel labels, adjusting the clique potential associated with that clique; and d) for each clique to which that node belongs, adjusting, based on the adjusted clique potential associated with that clique, at least one of the messages between that node and the other nodes of that clique; v) repeating step iv) until a convergence criterion is met; and vi) producing an adjusted image having pixel labels determined from the adjusted messages. 2. A method according to claim 1 , wherein step iv) further comprises, after step a) and before step c), updating the respective set of possible pixel labels by removing one of the possible pixel labels therefrom. 3. A method according to claim 2 , wherein for each clique to which that node belongs there is an associated parameter related to messages passed between that node and other nodes of that clique, the parameter dependent on the possible pixel labels for that node; and wherein the one of the possible pixel labels that is removed from the respective set of possible pixel labels is the possible pixel label of the respective set of possible pixel labels for which the sum of the associated parameters is greatest. 4. A method according to claim 1 , wherein, at step c), adjusting the clique potential associated with that clique comprises adjusting the clique potential based on a weighted sum of the unary potentials associated with the respective set of possible pixel labels. 5. A method according to claim 4 , wherein, at step c), adjusting the clique potential associated with that clique further comprises adding a constant value to the weighted sum for each possible pixel label which is not included in the respective set of possible pixel labels. 6. A method according to claim 1 , wherein, at step b), the selected possible pixel label is selected from a candidate set of possible pixel labels, wherein the candidate set of possible pixel labels is based on an estimated set of possible pixel labels of an adjacent node for which the unary potential is known. 7. A method according to claim 1 , wherein repeating step iv) until a convergence criterion is met comprises repeating step iv) a predetermined number of times. 8. A method according to claim 1 , wherein repeating step iv) until a convergence criterion is met comprises repeating step iv) until repetition of step iv) does not change the messages. 9. A method according to claim 1 , wherein at step b), computing a unary potential of a selected possible pixel label comprises estimating a unary potential of the selected possible pixel label based on a unary potential associated with a possible pixel label of an adjacent node in the graph. 10. A method according to claim 1 , wherein, at step d), adjusting at least one of the messages comprises adjusting the at least one of the messages according to one or more predetermined message weightings. 11. A method according to claim 1 , wherein the unary potentials are configured so as to cause the method to perform one or more of the following operations upon the input image: segmentation; distortion correction; optical flow; dense stereo correspondence; image denoising, deblurring and zooming. 12. A method according to claim 1 , wherein the clique potentials are initially configured so as to cause the method to perform one or more of the following operations upon the input image: segmentation; distortion correction; optical flow; dense stereo correspondence; image denoising, deblurring and zooming. 13. An apparatus or system comprising: input/output means arranged to receive an image; a processor; and a memory storing instructions for causing a processor to perform the following steps: i) receiving an input image that comprises a plurality of pixels, each pixel having an input value and each input value corresponding to one of a number of possible pixel labels; ii) associating each pixel with a node of a graph and one or more cliques of nodes, wherein each possible pixel label is associated with a unary potential, and wherein, for each permutation of possible pixel labels of the pixels associated with the nodes of each clique, a clique potential is associated; iii) for each clique, initialising at least one message between each pair of nodes of that clique; iv) for each node: a) determining, for that node, a respective set of possible pixel labels for which the unary potential is known; b) computing a unary potential of a selected possible pixel label for which the unary potential is unknown and updating the respective set of possible pixel labels to include the selected possible pixel label; c) for each clique to which that node belongs, based on the unary potentials associated with the updated respective set of possible pixel labels, adjusting the clique potential associated with that clique; and d) for each clique to which that node belongs, adjusting, based on the adjusted clique potential associated with that clique, at least one of the messages between that node and the other nodes of that clique; v) repeating step iv) until a convergence criterion is met; and vi) producing an adjusted image having pixel labels determined from the adjusted messages. 14. A non-transitory computer-readable medium comprising machine-readable instructions arranged, upon execution by one or more processors, to cause the one or more processors to perform the following steps: i) receiving an input image that comprises a plurality of pixels, each pixel having an input value and each input value corresponding to one of a number of possible pixel labels; ii) associating each pixel with a node of a graph and one or more cliques of nodes, wherein each possible pixel label is associated with a unary potential, and wherein, for each permutation of possible pixel labels of the pixels associated with the nodes of each clique, a clique potential is associated; iii) for each clique, initialising at least one message between each pair of nodes of that clique; iv) for each node: a) determining, for that node, a respective set of possible pixel labels for which the unary potential is known; b) computing a unary potential of a selected possible pixel label for which the unary potential is unknown and updating the respective set of possible pixel labels to include the selected possible pixel label; c) for each clique to which that node belongs, based on the unary potentials associated with
Related publications grouped by family.
Answers are generated from the same data shown on this page.