Patent · US Active

Image foreground segmentation algorithm based on edge knowledge transformation

US12067731B2 · kind B2 · utility

0Cited by
2References
3Claims
0Family size

Assignee

Inventors

Key dates

Filing dateJan 28, 2022
Grant dateAug 20, 2024
Priority date
Expiry dateJan 16, 2043

Classification

  • Technology area (CPC G)Physics
  • CPC primaryG06T2207/20084
  • WIPO fieldComputer technology
  • WIPO sectorElectrical engineering

Abstract

An image foreground segmentation algorithm based on edge knowledge transformation includes the following steps: 1) construction of an image segmentation framework with edge self-supervised mechanism; 2) construction of an inner edge and outer edge discriminator; 3) generation of pseudo-segmented triplet data; 4) two edge adversary foreground segmentation guided by a very few labeled samples of the target category. According to the image foreground segmentation algorithm based on edge knowledge transformation established in the above steps, under the guidance of a very small number of labeled segmentation samples of the target category, the inner and outer edge discrimination network transforms edge knowledge of a large amount of open-source labeled data of a non-target category into a foreground segmentation network for the target category image by the way of adversary of segmentation results, and realize the segmentation of foreground target category objects of the image.

Source: USPTO / EPO open patent data. Objective bibliographic and citation counts.