Towards Semantic Clustering : Grouping Image Visual Features Through Exploratory Factor Analysis

Narayanan, A/L N. Kulathu Ramaiyer and Lim, Phei Chin and Dayang Nurfatimah, Binti Awang Iskandar and Chiew, Kang Leng (2012) Towards Semantic Clustering : Grouping Image Visual Features Through Exploratory Factor Analysis. In: Proceedings of the IIEEJ Image Electronics and Visual Computing Workshop 2012, 21-24 November 2012, Kuching, Malaysia.

[img] PDF
Towards Semantic Clustering - Copy.pdf

Download (143kB)


Current image clustering schemes tend to cluster images based on similarity of low-level image visual features. Our previous work has demonstrated the need for organizing groups of low-level image visual features into composite feature sets that can then be mapped to semantically relevant abstractions. Symbolic terms such as wing ratio and tailed-wings and many more have been obtained from mapping clusters from a single-feature clustering and visual knowledge acquisition. Current focus is the explorations on the extraction and transformation of groupings of low-level image visual features into factor space before mapped to these meaningful terms. Preliminary results from exploratory factor analyses with different settings suggested the solution of forming four groups of features. The selected visual feature groupings have also been shown to correspond to the user-relevant symbolic terms. We hope to highlight these mapped relationships at the conference.

Item Type: Proceeding (Paper)
Uncontrolled Keywords: image clustering, image visual, visual knowledge, mapping clusters, mapping clusters, unimas, university, universiti, Borneo, Malaysia, Sarawak, Kuching, Samarahan, ipta, education, research, Universiti Malaysia Sarawak.
Subjects: Q Science > QA Mathematics > QA75 Electronic computers. Computer science
Divisions: Academic Faculties, Institutes and Centres > Faculty of Computer Science and Information Technology
Depositing User: Gani
Date Deposited: 01 Jul 2020 02:29
Last Modified: 01 Jul 2020 02:29

Actions (For repository members only: login required)

View Item View Item