Semantic multimedia modelling & interpretation for search & retrieval
Aslam, Nida (2011) Semantic multimedia modelling & interpretation for search & retrieval. PhD thesis, Middlesex University.
PDF (PhD thesis)
With the axiomatic revolutionary in the multimedia equip devices, culminated in the proverbial proliferation of the image and video data. Owing to this omnipresence and progression, these data become the part of our daily life. This devastating data production rate accompanies with a predicament of surpassing our potentials for acquiring this data. Perhaps one of the utmost prevailing problems of this digital era is an information plethora.
Until now, progressions in image and video retrieval research reached restrained success owed to its interpretation of an image and video in terms of primitive features. Humans generally access multimedia assets in terms of semantic concepts. The retrieval of digital images and videos is impeded by the semantic gap. The semantic gap is the discrepancy between a user’s high-level interpretation of an image and the information that can be extracted from an image’s physical properties. Content- based image and video retrieval systems are explicitly assailable to the semantic gap due to their dependence on low-level visual features for describing image and content. The semantic gap can be narrowed by including high-level features. High-level descriptions of images and videos are more proficient of apprehending the semantic meaning of image and video content.
It is generally understood that the problem of image and video retrieval is still far from being solved. This thesis proposes an approach for intelligent multimedia semantic extraction for search and retrieval. This thesis intends to bridge the gap between the visual features and semantics. This thesis proposes a Semantic query Interpreter for the images and the videos. The proposed Semantic Query Interpreter will select the pertinent terms from the user query and analyse it lexically and semantically. The proposed SQI reduces the semantic as well as the vocabulary gap between the users and the machine. This thesis also explored a novel ranking strategy for image search and retrieval. SemRank is the novel system that will incorporate the Semantic Intensity (SI) in exploring the semantic relevancy between the user query and the available data. The novel Semantic Intensity captures the concept dominancy factor of an image. As we are aware of the fact that the image is the combination of various concepts and among the list of concepts some of them are more dominant then the other. The SemRank will rank the retrieved images on the basis of Semantic Intensity.
The investigations are made on the LabelMe image and LabelMe video dataset. Experiments show that the proposed approach is successful in bridging the semantic gap. The experiments reveal that our proposed system outperforms the traditional image retrieval systems.
|Item Type:||Thesis (PhD)|
|Research Areas:||A. > Institute for Work Based Learning
B. > Theses
|Depositing User:||Aran Lewis|
|Date Deposited:||11 May 2012 14:35|
|Last Modified:||30 Apr 2015 15:50|
Actions (login required)