| Full text | |
| Author(s): |
Barrera, Junior
;
Hashimoto, Ronaldo F.
;
Hirata, Nina S. T.
;
Hirata, R., Jr.
;
Reis, Marcelo S.
Total Authors: 5
|
| Document type: | Journal article |
| Source: | SAO PAULO JOURNAL OF MATHEMATICAL SCIENCES; v. 16, n. 1, p. 42-pg., 2022-04-21. |
| Abstract | |
Morphological image operators are a class of non-linear image mappings studied in Mathematical Morphology. Many significant theoretical results regarding the characterization of families of image operators, their properties, and representations are derived from lattice theory, the underlying foundation of Mathematical Morphology. A fundamental representation result is a pair of canonical decompositions for any translation-invariant operator as a union of sup-generating or an intersection of inf-generating operators, which in turn can be written in terms of two basic operators, erosions and dilations. Thus, in practice, a toolbox with functional operators can be built by composing erosions and dilations, and then operators of the toolbox can be further combined to solve image processing problems. However, designing image operators by hand may become a daunting task for complex image processing tasks, and this motivated the development of machine learning based approaches. This paper reviews the main contributions around this theme made by the authors and their collaborators over the years. The review covers the relevant theoretical elements, particularly the canonical decomposition theorem, a formulation of the learning problem, some methods to solve it, and algorithms for finding computationally efficient representations. More recent contributions included in this review are related to families of operators (hypothesis spaces) organized as lattice structures where a suitable subfamily of operators is searched through the minimization of U-curve cost functions. A brief account of the connections between morphological image operator learning and deep learning is also included. (AU) | |
| FAPESP's process: | 15/22308-2 - Intermediate representations in Computational Science for knowledge discovery |
| Grantee: | Roberto Marcondes Cesar Junior |
| Support Opportunities: | Research Projects - Thematic Grants |
| FAPESP's process: | 13/07467-1 - CeTICS - Center of Toxins, Immune-Response and Cell Signaling |
| Grantee: | Hugo Aguirre Armelin |
| Support Opportunities: | Research Grants - Research, Innovation and Dissemination Centers - RIDC |
| FAPESP's process: | 19/21619-5 - Finding the Goldilocks zones of cell signaling pathways in cancer therapy |
| Grantee: | Marcelo da Silva Reis |
| Support Opportunities: | Regular Research Grants |
| FAPESP's process: | 17/25835-9 - Understanding images and deep learning models |
| Grantee: | Nina Sumiko Tomita Hirata |
| Support Opportunities: | Research Grants - Research Partnership for Technological Innovation - PITE |