Content-based multimedia information retrieval makes available of brand new ideas and means for searching through gazillions type of media all over the world. It is literally, the improvised version of multimedia information retrieval (MIR), for MIR system had been developed to become more user-friendly by using content-based methods.
Content-based methods are needed whenever text annotations are non-existent or incomplete. Plus, content-based methods also improve the information retrieval accuracy even when text annotations are present that is by providing additional insights into the media collections.
The typical content-based information retrieval (CBIR) system e.g., an image retrieval system has three major aspects that are (1) feature extraction, (2) high dimensional indexity and (3) system design. Among these three aspects, feature extraction is the basis of CBIR.
How do content-based methods actually improve things ?
The earliest years of multimedia information retrieval had worked using the mechanism of computer vision algorithms of which the mechanism stressed on feature-based similarity search over images, videos and audio. However, this feature-based similarity search concept was inconvenient for the users, for this concept worked in such a way that only scientists could fathom. Therefore, the new approaches were conducted to overcome this problem and the approaches are; (1) relevance feedback and, (2) hidden annotation-- (*relevance feedback is one of the interactive tools in content-based image retrieval). These two approaches have been working with an objective to have the system of retrieval understood the semantics of a query and not simply the low-level underlying computational features. This is because the low-level underlying computational features have a flaw of which the two semantically similar objects may lie far from each other in the feature space, while two completely different objects may stay close to each other. Therefore, this process of improvising is called as "Bridging the semantic gap". This literally means having low-level underlying computational features translated to high-level concepts or terms which would be intuitive to the user. In this brand new improvised system, it meets the two basic necessities for a multimedia information retrieval functions that are (1) searching for a particular media item and, (2) browsing and summarizing a media collection.
With content-based information retrieval, the process of searching pictures online becomes easier as guidance and navigational tools are given for users to get the information they want. These are some of the examples of content-based information retrieval:
1)
2)
Sources :
Paper project of Cha Zhang, Student Member, IEEE, and Tsuhan Chen, Member, IEEE; An Active Learning Framework for Content-Based Information Retrieval.
No comments:
Post a Comment