Improvisation of Retrieval Effectiveness by CR-Reranking Method

Web-scale image search engines (For e.g. Google Image Search, Bing Image Search, Pinterest) mostly rely on surrounding text features. It is difficult for them to predict user's intention only with the query they are giving and this leads to ambiguous and noisy search results from the search engines which are far from satisfactory. In this paper, we propose a novel Internet image search approach. Which requires the user only to click on one query image with the minimum effort and images from the database retrieved by text-based search are re-ranked based on both visual and textual content. Our goal is to capture the user's search intention from this one-click query image in four steps as follows. (1) The query image which the users search for is categorized into one of the predefined adaptive weight categories, which reflect user's search intention at some level. (2) Based on the visual content of the searching image selected by the user and through image clustering mechanism, the query keywords are expanded to capture user intention by one click. (3) Expanded keywords are used to enlarge the image pool to contain more relevant images in which the user search for. (4) Expanded keywords from the above stage are also used to expand the query image to numerous positive visual examples from which new query specific visual and textual similarity metrics are learned to further improvise the results of content-based image re-ranking for improvised results. All these steps are automatic without extra effort from the user.

2014

In this paper, we present an approach for topic-level search and hyperlinking of video snippets, which relies on contentbased recommendation and multimodal re-ranking techniques. We identify topic-level segments using transcripts or subtitles and enrich them with other metadata. Segments are indexed in a word vector space. Given a text query or an anchor, the most similar segments are retrieved using cosine similarity scores, which are then combined with visual similarity scores, computed as the distance from the anchor's visual concept vector. This approach has performed well on the MediaEval 2013 Search and Hyperlinking task, evaluated over 1260 hours of BBC TV broadcast, in terms of overall mean average precision. Experiments showed that topic-segments based on transcripts from automatic speech recognition level systems (ASR) led to better performance than the ones based on subtitles for both search and hyperlinking. Moreover, by analyzing the effect of multimodal re-ranking on hyperlinking performance, we emphasize the merits of rich visual information available in the anchors for the hyperlinking task, and the merits of ASR for large-scale search and hyperlinking.

Proceeding of the ACM International Conference on Image and Video Retrieval - CIVR '09, 2009

In this paper, we describe one of the largest multi-site interactive video retrieval experiments conducted in a laboratory setting. Interactive video retrieval performance is difficult to cross-compare as variables exist across users, interfaces and the underlying retrieval engine. Conducted within the framework of TRECVID 2008, we completed a multi-site, multi-interface experiment. Three institutes participated involving 36 users, 12 each from Dublin City University (DCU, Ireland), University of Glasgow (GU, Scotland) and Centrum Wiskunde & Informatica (CWI, the Netherlands). Three user interfaces were developed which all used the same search service. Using a latin squares arrangement, each user completed 12 topics, leading to 6 TRECVID runs per site, 18 in total. This allowed us to isolate the factors of users and interfaces from retrieval performance. In this paper we present an analysis of both the quantitative and qualitative data generated from this experiment, demonstrating that for interactive video retrieval with "novice" users, performance can vary by up to 300% for the same system using different sets of users, whilst differences in performance of interface variants was in comparison not statistically different. Our results have implications for the manner in which interactive video retrieval experiments using non-expert users are evaluated. The primary focus of this paper is in highlighting that non-expert users generate very large performance fluctuations, which may either mask or create system variability. The discussion of why this happened is not covered by this paper.

Procedia - Social and Behavioral Sciences, 2013

The goal of our new e-science platform is to support collaborative research communities by providing a simple solution to jointly develop semantic-and media search algorithms on common and challenging datasets processed by novel feature extractors. Querying of nearest neighbor (NN) elements on large data collections is an important task for several information or content retrieval tasks. In the paper a flexible framework for research purposes is introduced for testing features, metrics, distances and indexing structures. The core part of the content based retrieval system is the LHI-tree, a disk-based index scheme for fast retrieval of multimodal features. Additionally, we compare LHI-tree to FLANN, an effective implementation of ANN search and show that LHI-tree gives similar list of retrieved images.

Image re-ranking, as an effective way to improve the results of web-based image search, has been adopted by current commercial search engines such as Bing and Google. Given a query keyword, a pool of images are first retrieved based on textual information. By asking the user to select a query image from the pool, the remaining images are re-ranked based on their visual similarities with the query image. A major challenge is that the similarities of visual features do not well correlate with images' semantic meanings which interpret users' search intention. Recently people proposed to match images in a semantic space which used attributes or reference classes closely related to the semantic meanings of images as basis. However, learning a universal visual semantic space to characterize highly diverse images from the web is difficult and inefficient. In this paper, we propose a novel image re-ranking framework, which automatically offline learns different semantic spaces for different query keywords. The visual features of images are projected into their related semantic spaces to get semantic signatures. At the online stage, images are re-ranked by comparing their semantic signatures obtained from the semantic space specified by the query keyword. The proposed query-specific semantic signatures significantly improve both the accuracy and efficiency of image re-ranking. The original visual features of thousands of dimensions can be projected to the semantic signatures as short as 25 dimensions. Experimental results show that 25-40 percent relative improvement has been achieved on re-ranking precisions compared with the state-of-the-art methods.

Web image search, with one click method is adopted by several commercial searches. Image re- raking approach is taken in account for current web based image search .image re-ranking starts with the intake of query keyword and search engine do first retrieve images according to the keyword given. The user now selects an image from the group retrieved and the images will be re-ranked according to users one click. The major challenge is the image with similar visual features cannot be correlated by which the user clicks. Learning of a universal visual semantic Space which depicts the characteristics of images which are highly different to each other is also a difficult task. Thus the proposed Approach to overcome the problem is an image re-ranking framework. It has the property of Automatic offline learning. It learns different visual semantic spaces for different query keyword. Thus a semantic signature will be produced by projecting visual features of images to their visual semantic space. At its online stage the produced semantic signature is compared with the images so that images get re-ranked as the semantic signature is obtained from keywords visual semantic space. As a result the proposed image re-ranking framework thus increases accuracy and efficiency of image re-ranking

International journal of engineering research and technology, 2018

Image re-ranking is a successful method to get better results of text-based image search. User click data has been used in image re-ranking, clicks have been exposed to more exactly express the relevance of retrieved images to search query. A significant problem for click-based method is the need of click data, since only a short number of web images have been clickedon by users. To solve this problem with predicting image clicks, we propose a query specific method for image click prediction, and apply the obtained click information to the image reranking. A query keywords given, from a pool of images is first retrieved based on textual information. The user have to pick a query picture from the pool, the remaining images are re-ranked based on their optical similarity with the query picture. A most important challenge is that the similarities of visual features do not well associate with images semantic meanings which interpret users search purpose. The proposed method is based on ...

IOSR Journal of Computer Engineering, 2016

New approach is presented for reranking of the images. While searching images on the web to increase the accuracy of the image search result Image search reranking is used. Image search reranking is an effective approach to refine the initial text-based image search result. In the reranking process,image visual contents are used to refine the initial text based result. The image re-ranking restricts users' attempt to just one-click feedback is a useful way to improve search results and its interaction is simple enough. One click event of the user is considered as user intention. We propose a feature extraction algorithm named Hyperspherebased Relevance Preserving Projection (HRPP) and a ranking function called Hypersphere-based Rank (H-Rank).A main challenge in the research of image Reranking is that the similarities of the visual features do not perfectly associate with actual semantic meaning of images which intends users search goals. H-Rank is an effective ranking algorithm to sort the images by their distances to the hypersphere center. Moreover, to capture the user's intent with minimum human interaction, a reversed KNN algorithm is proposed as One-Click-based HRPP (OC-HRPP). Finally, OC-HRPP algorithm and H-Rank algorithm form a new ISR method.

2016

Image search reranking (ISR) has been implemented in order to get a refined image search as per the user needs. ISR fails to capture user’s intension and needs for the search of a particular query. In order to improve this reranking with user interactions or active ranking is highly demanded to improve the image search performance. Use of visua l information can be used to solve text-based image retrieval. Among the papers referred for literature survey the techniques that can be implemented are hypersphere-based relevance preserving projection (HRPP), reverse k nearest neighbor (KNN) and ranking function hypersphere-based rank (H-rank). HRPP is used to sort the images into various categories and sort. Reverse KNN is used in order to review OnClick Hypersphere based Relevance Preserving Projection (OC-HRPP) to know the exact search of the user. H-rank technique uses Euclidean distance formula and gives us all user relevant refined image result.

2016

Image re-ranking, as an effective way to improve the results of web based image search. This have been adopted by current commercial search engines. The query-specific semantic signature is also striking in this application, where it is crucial to reduce the semantic gap when computing the similarities of images. Due to the ambiguity of query keywords, there may be multiple semantic categories under one keyword query. These approaches cannot absolutely capture users’ search intention without query images selected by users The visual features of images are projected into their related visual semantic spaces to get their semantic signatures.