Method that supports the analysis of digital images in a computer cluster environment
AssigneeFaculdades Católicas, Associação sem fins lucrativos, Mantenedora da Pontificia Universidade Católica
InventorsDário Augusto Borges Oliveira, Gilson Alexandre Ostwald Pedro da Costa
Life of this patent
Estimating the term…
Abstract
Methods that support the analysis of digital images through the distributed and integrated processing of raster and vector digital data in a computer cluster environment, the set of methods including a particular strategy for distributing the processing of spatial context-aware operations over distributed datasets, as well as specific methods for the structuring of operations aimed at calculating spectral and topological properties of image objects, and for the resolution of spatial conflicts among objects.
Claims8
A method for the distributed execution of an operation that computes spectral properties of image objects, which utilizes, as a distributed processing strategy, a spatial-aware with replication strategy having a partial results computation, wherein the computation of the partial results of the spatial-aware with replication strategy comprises accessing an image tile that overlaps a geo-cell, for each geo-cell, computing a value of a spectral property for each image object that overlaps the image tile considering only the pixels of the image tile that intersects the image object, and storing the computed value of the spectral property for the image object.
A method for the distributed execution of an operation that performs spatial conflict resolution among image objects, which utilizes, as a distributed processing strategy, a spatial-aware with replication strategy having a partial results reduction step, wherein after the partial results reduction step of the spatial-aware with replication strategy, the image objects are sorted in ascending order, by the value of some stored property of the image objects, and inserted into a list.
A method for the distributed execution of an operation that compute topological properties of image objects, which utilizes, as a distributed processing strategy, a spatial-aware with replication strategy having a partial results computation, wherein the computation of the partial results of the spatial-aware with replication strategy comprises identifying all pairs of neighboring image objects, and creating a list with all geo-cells that overlap each of the pairs of neighboring image objects.
Dependencies are inferred from the claim text.