Context Search and Recommendation for Large-Scale Community-Sharing Photos

計畫名稱:Context Search and Recommendation for Large-Scale Community-Sharing Photos

所屬單位:資訊系

研究團隊:多媒體實驗室

計畫主持人:徐宏民

研究人員:廖瑋星

資源需求:gcc, g++, matlab, perl

使用期間:2009/03~

研究主題:
Context Search and Recommendation for Large-Scale Community-Sharing Photos

研究內容概述:
Content-based image retrieval (CBIR) is one of the essential techniques for managing exponentially growing photos and the enabling technology for many applications such as annotation by search, computational photography, photo question and answering, etc. Though through decodes of research, the current solutions are limited due to the semantic gap. In this work, we argue to improve CBIR by exploiting the auxiliary knowledge (i.e., tags, photos, meta-data, etc.) from the booming media-sharing services (e.g., Flickr) and search engines (e.g., Google) and propose a semantic expansion framework for boosting CBIR. The framework discovers the semantic similarity between images by constructing a combinational image graph model and utilizing the noisy textual (i.e., tags) and visual information (i.e., visual words). We also consider the efficiency issues as deploying in the large scale media-sharing sites. Meanwhile, the framework is generic and can be extended for keyword-based image retrieval and also image annotation. Experimenting over large-scale photo benchmarks, the proposed semantic expansion framework outperforms traditional CBIR systems 2–3 times, on the average.

詳細計畫內容 回到上一頁