Supporting Keyword Search for Image Retrieval with Integration of Probabilistic Annotation

Sustainability

Tie Hua Zhou¹  Ling Wang²  Keun Ho Ryu¹ 

[1] Database/Bioinformatics Laboratory, School of Electrical & Computer Engineering, Chungbuk National University, Cheongju 362-763, Korea; E-Mail:;Department of Information Engineering, Northeast Dianli University, Jilin 132000, China

关键词: multi-label image;    image annotation;    annotation integration;    semantic matching;    keyword search;    image retrieval;   

DOI  :  10.3390/su7056303

来源: mdpi

【 摘 要 】

The ever-increasing quantities of digital photo resources are annotated with enriching vocabularies to form semantic annotations. Photo-sharing social networks have boosted the need for efficient and intuitive querying to respond to user requirements in large-scale image collections. In order to help users formulate efficient and effective image retrieval, we present a novel integration of a probabilistic model based on keyword query architecture that models the probability distribution of image annotations: allowing users to obtain satisfactory results from image retrieval via the integration of multiple annotations. We focus on the annotation integration step in order to specify the meaning of each image annotation, thus leading to the most representative annotations of the intent of a keyword search. For this demonstration, we show how a probabilistic model has been integrated to semantic annotations to allow users to intuitively define explicit and precise keyword queries in order to retrieve satisfactory image results distributed in heterogeneous large data sources. Our experiments on SBU (collected by Stony Brook University) database show that (i) our integrated annotation contains higher quality representatives and semantic matches; and (ii) the results indicating annotation integration can indeed improve image search result quality.

【 授权许可】

CC BY   
© 2015 by the authors; licensee MDPI, Basel, Switzerland.