A novel approach to enable semantic and visual image summarization for exploratory image search

Abstract

In this paper, we have developed a novel scheme to incorporate topic network and representativeness-based sampling for achieving semantic and visual summarization and visualization of large-scale collections of Flickr images. First, topic network is automatically generated for summarizing and visualizing large-scale collections of Flickr images at a semantic level, so that users can select more suitable keywords for more precise query formulation. Second, the diverse visual similarities between the semantically-similar images are characterized more precisely by using a mixture-of-kernels and a representativeness-based image sampling algorithm is developed to achieve similarity-based summarization and visualization of large amounts of images under the same topic, so that users can find some particular images of interest more effectively. Our experiments on large-scale image collections with diverse semantics have provided very positive results.

Visualizing Web Images Using Fisher Discriminant Locality Preserving Canonical Correlation Analysis

“…It has been reported in [27] and [28] that a scheme that utilizes several kinds of features provides better performance for multimedia analysis than does a scheme that utilizes only one feature. However, since the above dimensionality reduction methods can utilize only one kind of feature, it is difficult for these methods to use several kinds of modalities effectively when they are applied to multimedia data.…”

“…Although these methods integrate different kinds of features for realizing dimensionality reduction, they cannot consider the unique characteristics of each modality. In [27], it has been reported that considering the characteristics of target modalities is necessary for successful multimodal data analysis.…”

COMPENDIUM: a text summarisation tool for generating summaries of multiple purposes, domains, and genres

“…As it can be seen in the figure, although it has traditionally focused on text (Yu et al 2007), the input to the summarisation process can also be multimedia information, such as images (Fan et al 2008); video (Dumont and Mérialdo 2009), or audio (Liu and Liu 2009), as well as on-line information or hypertexts (Steinberger, Jezek and Sloup 2008;Tigelaar, Op Den Akker and Hiemstra 2010). Furthermore, we can talk about summarising only one document (single-document summarisation) (Svore, Vanderwende and Burges 2007) or multiple ones (multidocument summarisation) (Haghighi and Vanderwende 2009).…”

An Interactive Approach for Filtering Out Junk Images From Keyword-Based Google Search Results

“…Image seekers often express a desire for a user interface that can organize the search results into meaningful groups, in order to help them make sense of the search results, and to help them decide what to do next. Some pioneer works have been done on supporting similarity-based image visualization [19][20][26][27][28][29][30], but most existing techniques for image projection and visualization may perform well when the images belong to one single cluster, and fail to project the images nicely when they are spread among multiple clusters with diverse visual properties.…”

“…To capture users' query intentions precisely and generate better hypotheses for junk image filtering, three key issues should be addressed jointly: (a) incremental kernel learning should be supported for reducing the computational cost [31][32][33], so that users can interactively change the underlying hypotheses for filtering out the junk images in real time or nearly in real time; (b) the convergence of the underlying techniques for incremental kernel learning should be guaranteed; (c) an interactive interface should be developed to enable similaritybased visualization of large amounts of returned images [19][20][26][27][28][29][30], generate more understandable assessment of the hypotheses for junk image filtering (i.e., make the margin between the relevant images and the junk images to be more visible and more assessable), and allow users to express their query intentions more precisely and make better hypotheses for junk image filtering.…”