Abstract:Due to the low-level image features it utilizes, the semantic gap problem is hard to bridge and performance of CBIR systems is still far away from users' expectation. Image annotation, region-based image retrieval and relevance feedback are three main approaches for narrowing the "semantic gap". In this paper, recent development in these fields are reviewed and some future directions are proposed in the end.
“…The detected regions are being represented by its local feature and the weights of its importance. Finally and based on the regions representation, the similarity between the query regions and the other images are calculated to determine the relevant and irrelevant targets [2]. Though, RBIR still lacks the quality for many reasons such as inaccurate region segmentation, high dimensionality of the extracted local features and determining which similarity measure to apply.…”
Section: Region Content Based Image Retrievalmentioning
confidence: 99%
“…CBIR uses the visual contents of an image such as color, shape, texture, and spatial layout to represent and index the image [2]. In typical CBIR system, the visual contents of the images in the database are extracted and described by multi-dimensional feature vectors [3].…”
Section: Content Based Image Retrieval (Cbir)mentioning
confidence: 99%
“…Image query regarding the CBIR applications mainly organized into three paradigms [4]. First paradigm is the Query By Example (QBE) in which the query image is an example shown either from the database itself or some external location [2,4]. Secondly, there is the image query …”
Section: Content Based Image Retrieval (Cbir)mentioning
confidence: 99%
“…IR approaches varied between the Annotation Based Image Retrieval (ABIR) or the meta-data approach which is a traditional method for image retrieval that makes use of the meta data of image such as the textual descriptions, captioning, or the keywords to search and retrieve images [1]. ABIR can be either classification based methods or probabilistic modeling based method [2]. However there is also the Content Based Image Retrieval approach in which database images are indexed on the basis of low level features such as color, texture and shape which can be automatically derived from the visual contents of the images Although many researchers have investigated this topic from many perspectives to develop its performance, Content Based Image Retrieval (CBIR) systems still suffers from a lot of shortcomings.…”
Digital images databases open the way for content-based searching. Content Based Image Retrieval occupies a well ranked position among the research areas as it provides the practical solution for narrowing the semantic gap between the image retrieval process and the human perception. The main objective of this paper is to propose a framework for region content based image retrieval based on a distributed clustered image dataset. The proposed framework introduces a new perspective to measure the similarity between the image query and the clustered dataset images. Moreover, a development by adopting three relevance feedback techniques is used to refine the results of the retrieval system which are the well known Query Point Movement and Query Expansion, besides to the proposed third technique which is Query Modified ReWeighting technique.
“…The detected regions are being represented by its local feature and the weights of its importance. Finally and based on the regions representation, the similarity between the query regions and the other images are calculated to determine the relevant and irrelevant targets [2]. Though, RBIR still lacks the quality for many reasons such as inaccurate region segmentation, high dimensionality of the extracted local features and determining which similarity measure to apply.…”
Section: Region Content Based Image Retrievalmentioning
confidence: 99%
“…CBIR uses the visual contents of an image such as color, shape, texture, and spatial layout to represent and index the image [2]. In typical CBIR system, the visual contents of the images in the database are extracted and described by multi-dimensional feature vectors [3].…”
Section: Content Based Image Retrieval (Cbir)mentioning
confidence: 99%
“…Image query regarding the CBIR applications mainly organized into three paradigms [4]. First paradigm is the Query By Example (QBE) in which the query image is an example shown either from the database itself or some external location [2,4]. Secondly, there is the image query …”
Section: Content Based Image Retrieval (Cbir)mentioning
confidence: 99%
“…IR approaches varied between the Annotation Based Image Retrieval (ABIR) or the meta-data approach which is a traditional method for image retrieval that makes use of the meta data of image such as the textual descriptions, captioning, or the keywords to search and retrieve images [1]. ABIR can be either classification based methods or probabilistic modeling based method [2]. However there is also the Content Based Image Retrieval approach in which database images are indexed on the basis of low level features such as color, texture and shape which can be automatically derived from the visual contents of the images Although many researchers have investigated this topic from many perspectives to develop its performance, Content Based Image Retrieval (CBIR) systems still suffers from a lot of shortcomings.…”
Digital images databases open the way for content-based searching. Content Based Image Retrieval occupies a well ranked position among the research areas as it provides the practical solution for narrowing the semantic gap between the image retrieval process and the human perception. The main objective of this paper is to propose a framework for region content based image retrieval based on a distributed clustered image dataset. The proposed framework introduces a new perspective to measure the similarity between the image query and the clustered dataset images. Moreover, a development by adopting three relevance feedback techniques is used to refine the results of the retrieval system which are the well known Query Point Movement and Query Expansion, besides to the proposed third technique which is Query Modified ReWeighting technique.
“…Therefore, the automatic annotation of medical imaging has become an urgent demand. Traditional medical image classification is mostly based on the basic characteristics of the image, such as color features, texture features, shape features [1,2], which failed to solve the "Semantic gap" problem [3][4][5]. Because the basic characteristics of the image cannot reflect the underlying information in the images, for example the image may imply information of the specific organizational structure which cannot be obtained from the basic characteristics of the image and be a kind of potential information which only can be obtained by the doctors' experience.…”
Abstract. BACKGROUND:With the rapid development of modern medical imaging technology, medical image classification has become more important for medical diagnosis and treatment. OBJECTIVE: To solve the existence of polysemous words and synonyms problem, this study combines the word bag model with PLSA (Probabilistic Latent Semantic Analysis) and proposes the PLSA-BOW (Probabilistic Latent Semantic AnalysisBag of Words) model. METHODS: In this paper we introduce the bag of words model in text field to image field, and build the model of visual bag of words model. RESULTS: The method enables the word bag model-based classification method to be further improved in accuracy.
CONCLUSIONS:The experimental results show that the PLSA-BOW model for medical image classification can lead to a more accurate classification.
PurposeThe purpose of this study aims to identify the impact of verbal-visual cognitive styles on the level of satisfaction and behavior in the textual and content search of Google Images.Design/methodology/approach“Riding” cognitive style test and satisfaction questionnaire were used as data collection tools. Also, to collect data related to the image search behavior, the subjects’ transaction files were recorded using Camtasia software and then the files observed and reviewed. The research sample was 90 postgraduate students of Shiraz University.FindingsThe results showed that cognitive styles in interaction with the text-based and content-based search system of “Google Images” affected user’s satisfaction. Text-based image retrieval, in which vocabulary-based information needs were expressed, was more compatible with the verbal cognitive style and resulted in greater satisfaction. In contrast, in content-based image retrieval, where it was possible to express information needs in the form of images, users were more satisfied with the visual cognitive style. Verbal users performed more positively in text-based search and visual users in content-based search.Originality/valueConsidering the research gap, which has identified the performance of visual text-based and content-based systems in terms of satisfaction and cognitive style search behavior, the present study could be considered a small effort to promote science.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.