Full text
| |
| Author(s): |
Alberto Arruda de Oliveira
Total Authors: 1
|
| Document type: | Master's Dissertation |
| Press: | Campinas, SP. |
| Institution: | Universidade Estadual de Campinas (UNICAMP). Instituto de Computação |
| Defense date: | 2015-04-24 |
| Examining board members: |
Anderson de Rezende Rocha;
Nina Sumiko Tomita Hirata;
Hélio Pedrini
|
| Advisor: | Anderson de Rezende Rocha |
| Abstract | |
Due to the large amount of images shared on the web, tracking the spread and evolution of their content became an increasingly important problem. Recently, several works have been concerned with modeling the parenthood relationships between near duplicates in a set of images. Two images share a parenthood relationship if one is obtained by applying transformations to the other. However, this is not the only form of parenting that can exist among images. An image might be a composition created through the combination of the semantic information existent in two or more source images, establishing a relationship between the sources and the composite. We name as Multiple Parenting Phylogeny the problem of identifying such relationships in a set containing near-duplicate subsets of source and composition images. To tackle this problem, we present in this dissertation a three-step solution: (1) separation of near-duplicate groups; (2) classification of the relations between the groups; and (3) identification of the images used to create the original composition. Moreover, next we extend upon this framework by introducing key improvements, such as better identification of when two images share content, and improved ways to compare this content. We evaluate the proposed approaches in a scenario containing three subsets of near duplicates: one belonging to the composition image, and two to the source images used to create it. Three datasets were developed for this work, two with automatically created compositions and one of realistic, professionally-made compositions. The original and extended frameworks are compared by means of quantitative metrics established for evaluating the accuracy in reconstructing phylogenies and finding multiple parenting relationships. Finally, we discuss some particularities of the framework, such as the importance of an effective reconstruction of phylogenies and how more complex compositions can negatively affect the precision of the method (AU) | |
| FAPESP's process: | 13/21251-1 - Multiple parenting relations in image phylogeny. |
| Grantee: | Alberto Arruda de Oliveira |
| Support Opportunities: | Scholarships in Brazil - Master |