Multi-Instance Learning for Image Retrieval with Relevance Feedback

2013 ◽  
Vol 427-429 ◽  
pp. 1606-1609 ◽  
Author(s):  
Tao Chen ◽  
Hui Fang Deng
Keyword(s):  
Image Retrieval ◽  
Objective Function ◽  
Relevance Feedback ◽  
Image Data ◽  
Experimental Results ◽  
Data Sets ◽  
Digital Repository ◽  
Novel Method

In this paper, we propose a novel method for image retrieval based on multi-instance learning with relevance feedback. The process of this method mainly includes the following three steps: First, it segments each image into a number of regions, treats images and regions as bags and instances respectively. Second, it constructs an objective function of multi-instance learning with the query images, which is used to rank the images from a large digital repository according to the distance values between the nearest region vector of each image and the maximum of the objective function. Third, based on the users relevance feedback, several rounds may be needed to refine the output images and their ranks. Finally, a satisfying set of images will be returned to users. Experimental results on COREL image data sets have demonstrated the effectiveness of the proposed approach.

scholarly journals Convolutional neural networks for relevance feedback in content based image retrieval

2020 ◽  
Vol 79 (37-38) ◽  
pp. 26995-27021
Author(s):  
Lorenzo Putzu ◽  
Luca Piras ◽  
Giorgio Giacinto
Keyword(s):  
Image Retrieval ◽  
Relevance Feedback ◽  
Image Representation ◽  
Semantic Gap ◽  
Content Based Image Retrieval ◽  
Data Sets ◽  
Great Success ◽  
Query Image ◽  
Pros And Cons ◽  
The One

Abstract Given the great success of Convolutional Neural Network (CNN) for image representation and classification tasks, we argue that Content-Based Image Retrieval (CBIR) systems could also leverage on CNN capabilities, mainly when Relevance Feedback (RF) mechanisms are employed. On the one hand, to improve the performances of CBIRs, that are strictly related to the effectiveness of the descriptors used to represent an image, as they aim at providing the user with images similar to an initial query image. On the other hand, to reduce the semantic gap between the similarity perceived by the user and the similarity computed by the machine, by exploiting an RF mechanism where the user labels the returned images as being relevant or not concerning her interests. Consequently, in this work, we propose a CBIR system based on transfer learning from a CNN trained on a vast image database, thus exploiting the generic image representation that it has already learned. Then, the pre-trained CNN is also fine-tuned exploiting the RF supplied by the user to reduce the semantic gap. In particular, after the user’s feedback, we propose to tune and then re-train the CNN according to the labelled set of relevant and non-relevant images. Then, we suggest different strategies to exploit the updated CNN for returning a novel set of images that are expected to be relevant to the user’s needs. Experimental results on different data sets show the effectiveness of the proposed mechanisms in improving the representation power of the CNN with respect to the user concept of image similarity. Moreover, the pros and cons of the different approaches can be clearly pointed out, thus providing clear guidelines for the implementation in production environments.

Research on Related Technology of Content-Based Image Retrieval

2013 ◽  
Vol 448-453 ◽  
pp. 3616-3620
Author(s):  
Yu Bing Dong ◽  
Ming Jing Li ◽  
Bai Chuan Li
Keyword(s):  
Image Retrieval ◽  
Relevance Feedback ◽  
Image Feature ◽  
Content Based Image Retrieval ◽  
The Novel ◽  
Effective Solution ◽  
Retrieval Accuracy ◽  
Novel Method ◽  
Feedback Method ◽  
High Level

Content-Based Image Retrieval (CBIR) system existed a gap between high-level concepts and low-level features. As an effective solution, the Relevance Feedback (RF) technique has been used on many CBIR systems to improve the retrieval precision. In order to further improve convergence speed and retrieval accuracy, a novel relevance feedback method was proposed. According to feedback from user, image feature was weighted and adjusted in the novel method.

A CONTEXT-BASED APPROACH FOR COLOR IMAGE RETRIEVAL

2002 ◽  
Vol 16 (02) ◽  
pp. 239-255 ◽  
Author(s):  
JAU-LING SHIH ◽  
LING-HWEI CHEN
Keyword(s):  
Image Retrieval ◽  
Similarity Measure ◽  
Relevance Feedback ◽  
Color Image ◽  
Experimental Results ◽  
Retrieval Method ◽  
Feature Vectors ◽  
The Mean ◽  
Feedback Algorithm

In this paper, a color image retrieval method based on the primitives of images will be proposed. First, the context of each pixel in an image will be defined. Then, the contexts in the image are clustered into several classes based on the algorithm of fast noniterative clustering. The mean of the context in the same class is considered as a primitive of the image. The primitives are used as feature vectors. Since the numbers of primitives between images are different, a specially designed similarity measure is then proposed to do color image retrieval. To better adapt to the preferences of users, a relevance feedback algorithm is provided to automatically determine the weight of each primitive according to the user's response. To demonstrate the effectiveness of the proposed system, several test databases from Corel are used to compare the performances of the proposed system with other methods. The experimental results show that the proposed system is superior to others.

scholarly journals Multimodal Markov Random Field for Image Reranking Based on Relevance Feedback

2013 ◽  
Vol 2013 ◽  
pp. 1-16 ◽  
Author(s):  
Ricardo Omar Chávez ◽  
Hugo Jair Escalante ◽  
Manuel Montes-y-Gómez ◽  
Luis Enrique Sucar
Keyword(s):  
Image Retrieval ◽  
Random Field ◽  
Markov Random Field ◽  
Relevance Feedback ◽  
Retrieval System ◽  
Experimental Results ◽  
Feedback Information ◽  
Markov Random ◽  
Initial Retrieval ◽  
Image Reranking

This paper introduces a multimodal approach for reranking of image retrieval results based on relevance feedback. We consider the problem of reordering the ranked list of images returned by an image retrieval system, in such a way that relevant images to a query are moved to the first positions of the list. We propose a Markov random field (MRF) model that aims at classifying the images in the initial retrieval-result list as relevant or irrelevant; the output of the MRF is used to generate a new list of ranked images. The MRF takes into account (1) the rank information provided by the initial retrieval system, (2) similarities among images in the list, and (3) relevance feedback information. Hence, the problem of image reranking is reduced to that of minimizing an energy function that represents a trade-off between image relevance and interimage similarity. The proposed MRF is a multimodal as it can take advantage of both visual and textual information by which images are described with. We report experimental results in the IAPR TC12 collection using visual and textual features to represent images. Experimental results show that our method is able to improve the ranking provided by the base retrieval system. Also, the multimodal MRF outperforms unimodal (i.e., either text-based or image-based) MRFs that we have developed in previous work. Furthermore, the proposed MRF outperforms baseline multimodal methods that combine information from unimodal MRFs.

Applied Technology in Image Retrieval via 8-Direction Gaussian Density in the DCT Domain

2014 ◽  
Vol 977 ◽  
pp. 431-434
Author(s):  
Jian Feng Wang ◽  
Wen Ming Wu ◽  
Xiao Rong Zhao
Keyword(s):  
Image Retrieval ◽  
Processing Speed ◽  
Feature Vector ◽  
Image Data ◽  
Experimental Results ◽  
Dct Domain ◽  
Gaussian Density ◽  
Applied Technology

In this paper, A new algorithm for compressed image retrieval is proposed based on Gaussian Density Feature Vector(GDFV). This algorithm directly extract gaussian density of 8 direction from compressed image data to construct a 2-dimention array (8*4) as an indexing key to retrieve images based on their content features. To test and evaluate the proposed algorithms, we carried out experiments with a database of 1000 images. In comparison with existing representative techniques, the experimental results show the superiority of the proposed method in terms of retrieval precision and processing speed.

scholarly journals A State of Art on Content Based Image Retrieval Systems

2019 ◽  
Vol 8 (2S4) ◽  
pp. 297-301
Keyword(s):  
Image Retrieval ◽  
Relevance Feedback ◽  
Image Data ◽  
Distance Measures ◽  
Content Based Image Retrieval ◽  
Shape Features ◽  
Query Image ◽  
Indexing Methods ◽  
Retrieval Systems ◽  
State Of Art

The digital image data is quick expanding in capacity and heterogeneity. The customary information retrieval approaches are cannot fulfill the client's need, so there isneed to present a proficient framework for Content Based Image Retrieval(CBIR). The CBIR is an appealing wellspring of precise and quick retrieval. CBIR goes for discovering imagedatabases for explicit images that are like a given query image dependent on its features.In this paper the methodology of content based image retrieval are examined, investigated and thought about. Here, the different image substance, for example, colour, texture and shape features are mined by utilizing differentfeature extraction procedures, and furthermore extraordinary distance measures, Relevance Feedback (RF) and indexing methods are used to improve the execution of the CBIR system.The existing exploration strategies are talked about with their benefits and negative marks, so the further research works can be focused more.

scholarly journals Optimal Features Selection for X-Ray Image Retrieval with Regular and Irregular Zoning Methods

2018 ◽  
Vol 7 (4.5) ◽  
pp. 87
Author(s):  
P. Nalini ◽  
Dr B. L. Malleswari
Keyword(s):  
Image Retrieval ◽  
Medical Image ◽  
Image Annotation ◽  
Image Data ◽  
Image Features ◽  
Data Sets ◽  
Features Selection ◽  
X Ray ◽  
Eigen Values ◽  
Textural Attributes

Medical Image Retrieval is mainly meant for enhancing the healthcare system by coordinating physicians and interact with computing machines. This helps the doctors and radiologists in understanding the case and leads to automatic medical image annotation process. The choice of image attributes have crucial role in retrieving similar looking images of various anatomic regions.  In this paper we presented an empirical analysis of an X-Ray image retrieval system with intensity, statistical features, DFT and DWT transformed coefficients and Eigen values using Singular Valued Decomposition techniques as parameters. We computed these features by dividing the images in five different regular and irregular zones. In our previous work we proved that analyzing the image with local attributes result in better retrieval efficiency and hence in this paper we computed the attributes by dividing the image into 64 regular and irregular zones. This experimentation carried out on IRMA 2008 and IRMA 2009 X-Ray image data sets. In this work we come up with some conclusions like wavelet based textural attributes, intensity features and Eigen values extracted from different regular zones worked well in retrieving the images over the features computed over irregular zones. We also determined like the set of image features in which form of zoning for different anatomical regions  result in excellent retrieval of  similar looking X-Ray images.

Image Retrieval using Conceptual Fuzzy Sets

2000 ◽  
Vol 4 (6) ◽  
pp. 450-456
Author(s):  
Tomohiro Takagi ◽  
◽  
Kazushi Kawase ◽  
Kazuhiko Otsuka ◽  
Keyword(s):  
Image Retrieval ◽  
Fuzzy Sets ◽  
Image Data ◽  
Experimental Results ◽  
Abstract Concepts

An algorithm is described that uses fuzzy sets to handle word ambiguity, the main cause of vagueness in the meaning of a word. It is based on conceptual fuzzy sets (CFSs), which represent the meaning of words by linking other related words. A trial application of this algorithm to image retrieval showed that it can retrieve images that conceptually fit the meanings of the entered keyword based on the context understood from the characteristics of images. Experimental results showed that the proposed algorithm works well to represent various meanings of a keyword by linking it to other words and to connect words directly to image data. In addition, image retrieval starting with a sample image also worked well. First, a selected sample image was translated into abstract concepts, and images fitting the concepts were chosen.

A Two-Stage of Relevance Feedback for Content-Based Image Retrieval

2011 ◽  
Vol 467-469 ◽  
pp. 1627-1632
Author(s):  
Xue Feng Wang ◽  
Xing Su Chen
Keyword(s):  
Logistic Regression ◽  
Image Retrieval ◽  
Relevance Feedback ◽  
Retrieval System ◽  
Image Database ◽  
Experimental Results ◽  
Content Based Image Retrieval ◽  
Target Image ◽  
Two Stage ◽  
Predictive Probabilities

In this paper, an effective relevance feedback (RF) approach is proposed in content-based image retrieval (CBIR). In the first stage, according to the user’s marked images, we get theirs predictive probabilities based-on Bayesian methodology which yields the posteriori of the images in the database; second via justify the weight of elements in each feature extracted of images, we refine features by logistic regression with positive features which get from the first stage. Then we rank the images according to the probability of the images in the database. The retrieval system is repeating until the user is satisfied with the feedback results or the target image has been found. Experimental results are shown to evaluate the method on a large image database in terms of precision and recall.

Sign in / Sign up

Export Citation Format

Share Document