Text-Based Image Retrieval

Introduction

The technological advancements of Internet have increased the population of images, which demands effective retrieval mechanisms. Search engines have become indispensable tools for retrieving relevant images from WWW. In general, there are two flavors of retrieval namely, Text Based Image Retrieval (TBIR) and Content Based Image Retrieval (CBIR). The CBIR systems are suitable only for domain specific applications. It is also difficult to capture the semantics of images using low-level features alone and the semantic gap posed between the query image and database images remains as challenge. The extracted semantic features of the query image may not effectively match with the database images. Additionally, the cost of time associated for extracting image features among large database is also very high. In a typical CBIR system, the query image is matched with database images and the retrieval set is ranked using a suitable similarity measure. The image retrieval can be broadly classified as Text-Based Image Retrieval and Content-Based Image Retrieval. Below, the functionality of each of these methods are discussed in detail with suitable example.