Abstract
This paper proposes a hybrid approach of ontology and image clustering to automatically generate hierarchic image database. In the field of computer vision, ”generic object recognition” is one of the most important topics. Generic object recognition needs three types of research: feature extraction, pattern recognition, and database preparation; this paper targets at database preparation. The proposed approach considers both object semantic and visual features in images. In the proposed approach, the semantic is covered by ontology framework, and the visual similarity is covered by image clustering based on Gaussian Mixture Model. The image database generated by the proposed approach covered over 4,800 concepts (where 152 concepts have more than 100 images) and its structure was hierarchic. Through the subjective evaluation experiment, whether images in the database were correctly mapped or not was examined. The results of the experiment showed over 84% precision in average. It was suggested that the generated image database was sufficiently practicable as learning database for generic object recognition.
Highlights
In the field of Computer Vision, “generic object recognition” is one of the most important and difficult topics
In case of an image that is searched with a query “animal” has tags “animal,” “dog,” and “bazooka,” both “animal” and “dog” are remained as appropriate tags; “animal” equals to the search class, Hybrid Approach of Ontology and Image Clustering and “dog” has a IS-A relation with the search class “animal.” images are mapped on hierarchy according to the remained tags while the remained tags are used as labels for images
This paper proposed an approach to automatically generate hierarchic image database using an ontology and Gaussian Mixture Model (GMM)-based image clustering in a hybrid
Summary
In the field of Computer Vision, “generic object recognition” is one of the most important and difficult topics. It is hard work to prepare an appropriate image database in general because of the following three reasons: various concepts should be covered, huge number of images should be prepared for each concept, and appropriate labels should be given to the images. To tackle these problems, Web image mining has recently been reported 4,5. Web Image mining realizes to obtain a huge image dataset from social Web service such as Flickr∗6 These studies about Web image mining directly use the tags given to images on the Web service as the labels. More appropriate images for the tag are detected; noisy images are estimated and removed from the database
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have