Divna Djordjevic

An Object- and User-Driven System for Semantic-Based Image Annotation and Retrieval

In this paper, a system for object-based semi-automatic indexing and retrieval of natural images is introduced. Three important concepts underpin the proposed system: a new strategy to fuse different low-level content descriptions; a learning technique involving user relevance feedback; and a novel object based model to link semantic terms and visual objects. To achieve high accuracy in the retrieval and subsequent annotation processes several low-level image primitives are combined in a suitable multifeatures space. This space is modelled in a structured way exploiting both low-level features and spatial contextual relations of image blocks. Support vector machines are used to learn from gathered information through relevance feedback. An adaptive convolution kernel is defined to handle the proposed structured multifeature space. The positive definite property of the introduced kernel is proven, as essential condition for uniqueness and optimality of the convex optimization in support vector machines. The proposed system has been thoroughly evaluated and selected results are reported in this paper

Video summarisation for surveillance and news domain

Video summarization approaches have various fields of application, specifically related to organizing, browsing and accessing large video databases. In this paper we propose and evaluate two novel approaches for video summarization, one based on spectral methods and the other on ant-tree clustering. The overall summary creation process is broke down in two steps: detection of similar scenes and extraction of the most representative ones. While clustering approaches are used for scene segmentation, the post-processing logic merges video scenes into a subset of user relevant scenes. In the case of the spectral approach, representative scenes are extracted following the logic that important parts of the video are related with high motion activity of segments within scenes. In the alternative approach we estimate a subset of relevant video scene using ant-tree optimization approaches and in a supervised scenario certain scenes of no interest to the user are recognized and excluded from the summary. An experimental evaluation validating the feasibility and the robustness of these approaches is presented.

Relevance feedback for image retrieval in structured multi-feature spaces

An approach for content-based image retrieval with relevance feedback based on a structured multi-feature space is proposed. It uses a novel kernel for merging multiple feature subspaces into a complementary space. The kernel exploits nature of the data by assigning appropriate weights for each feature set. The weights are dynamically adapted to user preferences in a relevance feedback scenario.

Relevance Feedback in Content-based Image Retrieval Systems, an Overview and Analysis

In this paper an overview of relevant developments in visual relevance feedback based image retrieval is presented. Important problems of content-based image retrieval are analyzed and relevant findings from the evaluation of our framework are reported