Open Access. Powered by Scholars. Published by Universities.®
Social and Behavioral Sciences Commons™
Open Access. Powered by Scholars. Published by Universities.®
- Publication Year
Articles 1 - 30 of 33
Full-Text Articles in Social and Behavioral Sciences
Utilizing Qr Codes To Verify The Visual Fidelity Of Image Datasets For Machine Learning, Yang-Wai Chow, Willy Susilo, Jianfang Wang, Richard Buckland, Joon Sang Baek, Jongkil Kim, Nan Li
Utilizing Qr Codes To Verify The Visual Fidelity Of Image Datasets For Machine Learning, Yang-Wai Chow, Willy Susilo, Jianfang Wang, Richard Buckland, Joon Sang Baek, Jongkil Kim, Nan Li
Faculty of Engineering and Information Sciences - Papers: Part B
Machine learning is becoming increasingly popular in modern technology and has been adopted in various application areas. However, researchers have demonstrated that machine learning models are vulnerable to adversarial examples in their inputs, which has given rise to a field of research known as adversarial machine learning. Potential adversarial attacks include methods of poisoning datasets by perturbing input samples to mislead machine learning models into producing undesirable results. While such perturbations are often subtle and imperceptible from the perspective of a human, they can greatly affect the performance of machine learning models. This paper presents two methods of verifying the …
Automatic Ventricular Nuclear Magnetic Resonance Image Processing With Deep Learning, Binbin Yong, Chen Wang, Jun Shen, Fucun Li, Hang Yin
Automatic Ventricular Nuclear Magnetic Resonance Image Processing With Deep Learning, Binbin Yong, Chen Wang, Jun Shen, Fucun Li, Hang Yin
Faculty of Engineering and Information Sciences - Papers: Part A
Cardiovascular diseases (CVD) seriously threaten the health of human beings, and they have caused widespread concern in recent years. At present, the diagnosis of CVD is mainly conducted by computed tomography (CT), echocardiography and nuclear magnetic resonance (NMR) technologies. NMR imaging technology is widely used in medical applications owing to its characteristics of high resolution and very low radiation. However, manual NMR image segmentation is time-consuming and error-prone, which has led to the research on automatic NMR image segmentation technologies. Researchers tend to explore the ventricular NRM image segmentation to improve the accuracy of CVD diagnosis. In this study, based …
An Assessment Of Image Distortion And Ct Number Accuracy Within A Wide-Bore Ct Extended Field Of View, Bradley Beeksma, D Truant, Lois C. Holloway, Sankar Arumugam
An Assessment Of Image Distortion And Ct Number Accuracy Within A Wide-Bore Ct Extended Field Of View, Bradley Beeksma, D Truant, Lois C. Holloway, Sankar Arumugam
Faculty of Engineering and Information Sciences - Papers: Part A
Although wide bore computed tomography (CT) scanners provide increased space for patients, the scan field of view (sFOV) remains considerably smaller than the bore size. Consequently, patient anatomy which spans beyond the sFOV is truncated and the information is lost. As a solution, some manufacturers provide the capacity to reconstruct CT images from a partial dataset at an extended field of view (eFOV). To assess spatial distortion within this eFOV three phantoms were considered a 30 x 30 x 20 cm3 slab of solid water, the Gammex electron density CT phantom and a female anthropomorphic phantom. For each phantom, …
Ion Radiography As A Tool For Patient Set-Up & Image Guided Particle Therapy: A Monte Carlo Study, Nicolas Depauw, Marta F. Dias, Anatoly B. Rosenfeld, Joao C. Seco
Ion Radiography As A Tool For Patient Set-Up & Image Guided Particle Therapy: A Monte Carlo Study, Nicolas Depauw, Marta F. Dias, Anatoly B. Rosenfeld, Joao C. Seco
Faculty of Engineering and Information Sciences - Papers: Part A
This study investigate the use of ion radiography as a tool for patient set-up and tumor tracking capabilities for image guided particle therapy (IGPT) using Monte Carlo simulations. One pediatric, two lung and one liver cancer patients were considered in this study. For each patient, 230 and 330 MeV proton, and 500 MeV/nucleon carbon ion pencil beams were simulated through their computed tomography (CT) data set using GEANT4.9.0. Energy, position and direction cosines of each particle were recorded in front and behind the patient. Ion radiographs were subsequently reconstructed using a dedicated in-house software. The image quality was assessed by …
Quality Of Experience-Based Image Feature Selection For Mobile Augmented Reality Applications, Yi Cao, Christian H. Ritz, Raad Raad
Quality Of Experience-Based Image Feature Selection For Mobile Augmented Reality Applications, Yi Cao, Christian H. Ritz, Raad Raad
Faculty of Engineering and Information Sciences - Papers: Part A
Mobile augmented reality applications rely on automatically recognising a visual scene through matching of derived image features. To ensure the Quality of Experience (QoE) perceived by users, such applications should achieve high matching accuracy meanwhile minimizing the waiting time to meet real-time requirement. An efficient solution is to develop an effective feature selection method to select the most robust features against distortions caused by camera capture to achieve high matching accuracy whilst transmission and matching process of the features are significant reduced. Feature selection is also beneficial to reducing the computational complexities of the matching system so that waiting time …
Multi-Stage Compressed Sensing And Wall Clutter Mitigation For Through-The-Wall Radar Image Formation, Fok Hing Chi Tivive, Abdesselam Bouzerdoum, Van Ha Tang
Multi-Stage Compressed Sensing And Wall Clutter Mitigation For Through-The-Wall Radar Image Formation, Fok Hing Chi Tivive, Abdesselam Bouzerdoum, Van Ha Tang
Faculty of Engineering and Information Sciences - Papers: Part A
In this paper, a multi-stage through - The-wall radar imaging technique combining wall clutter mitigation and scene reconstruction is proposed. In the first stage, compressed sensing is applied to compressive measurements to recover the radar signals in the wavelet domain. Then, a subspace projection method is employed to remove the wavelet coefficients associated with the exterior wall reflections. In the second stage, the remaining wavelet coefficients are further compressed using principal component analysis. A compact linear measurement model is then formulated which relates the compressed wavelet coefficients to the image of the scene. Finally, the image reconstruction problem is solved …
Characterisation Of Coke Packed Beds After Liquid Slag Flow At 1 500°C By Image Analysis, Hazem Labib George, Raymond Longbottom, Sheng Chew, David Pinson, Brian Monaghan
Characterisation Of Coke Packed Beds After Liquid Slag Flow At 1 500°C By Image Analysis, Hazem Labib George, Raymond Longbottom, Sheng Chew, David Pinson, Brian Monaghan
Faculty of Engineering and Information Sciences - Papers: Part A
The flow of slag in the lower zone of the ironmaking blast furnace was experimentally simulated by adding slag to a high temperature laboratory scale coke packed bed. The flow of slag through packed beds of coke with packing densities varying from 50% to 65% was examined at 1 500°C. Since the liquid flow through a packed bed depends on packing properties such as particle size, particle shape, pore size and pore neck size, it was necessary to characterise these properties of the beds. In this work, image analysis of successive sections of the tested beds was utilised to characterise …
Smoke Detection In Video: An Image Separation Approach, Hongda Tian, Wanqing Li, Lei Wang, Philip Ogunbona
Smoke Detection In Video: An Image Separation Approach, Hongda Tian, Wanqing Li, Lei Wang, Philip Ogunbona
Faculty of Engineering and Information Sciences - Papers: Part A
Existing video-based smoke detection methods often rely on the visual features extracted directly from the original frames. In the case of light smoke, the background is still visible and it deteriorates the quality of the features. This paper presents an approach to separating the smoke component from the background such that visual features can be extracted from the smoke component for reliable smoke detection. Specifically, an image is assumed to be a linear blending of a smoke component and a background image. Given a video frame and its background, the estimation of the blending parameter and the actual smoke component …
Hep-2 Cell Image Classification With Multiple Linear Descriptors, Lingqiao Liu, Lei Wang
Hep-2 Cell Image Classification With Multiple Linear Descriptors, Lingqiao Liu, Lei Wang
Faculty of Engineering and Information Sciences - Papers: Part A
The automatic classification of the HEp-2 cell stain patterns from indirect immunofluorescence images has attracted much attention recently. As an image classification problem, it can be well solved by the state-of-the-art bag-of-features (BoF) model as long as a suitable local descriptor is known. Unfortunately, for this special task, we have very limited knowledge of such a descriptor. In this paper, we explore the possibility of automatically learning the descriptor from the image data itself. Specifically, we assume that a local patch can be well described by a set of linear projections performed on its pixel values. Based on this assumption, …
Image Guidance During Breast Radiotherapy: A Phantom Dosimetry And Radiation-Induced Second Cancer Risk Study, A Quinn, L Holloway, Peter Metcalfe
Image Guidance During Breast Radiotherapy: A Phantom Dosimetry And Radiation-Induced Second Cancer Risk Study, A Quinn, L Holloway, Peter Metcalfe
Faculty of Engineering and Information Sciences - Papers: Part A
Imaging procedures utilised for patient position verification during breast radiotherapy can add a considerable dose to organs surrounding the target volume on top of therapeutic scatter dose. This study investigated the dose from a breast kilovoltage cone-beam CT (kV-CBCT), a breast megavoltage fan-beam CT (MV-FBCT), and a TomoDirectTM breast treatment. Thermoluminescent dosimeters placed within a female anthropomorphic phantom were utilised to measure the dose to various organs and tissues. The contralateral breast, lungs and heart received 0.40 cGy, 0.45 cGy and 0.40 cGy from the kV-CBCT and 1.74 cGy, 1.39 cGy and 1.73 cGy from the MV-FBCT. In comparison to …
Optical Image Encryption Based On Chaotic Baker Map And Double Random Phase Encoding, Ahmed M. Elshamy, Ahmed N. Z Rashed, Abd El-Naser A. Mohamed, Osama S. Faragalla, Yi Mu, Saleh A. Alshebeili, F E Abd El-Samie
Optical Image Encryption Based On Chaotic Baker Map And Double Random Phase Encoding, Ahmed M. Elshamy, Ahmed N. Z Rashed, Abd El-Naser A. Mohamed, Osama S. Faragalla, Yi Mu, Saleh A. Alshebeili, F E Abd El-Samie
Faculty of Engineering and Information Sciences - Papers: Part A
This paper presents a new technique for optical image encryption based on chaotic Baker map and Double Random Phase Encoding (DRPE). This technique is implemented in two layers to enhance the security level of the classical DRPE. The first layer is a pre-processing layer, which is performed with the chaotic Baker map on the original image. In the second layer, the classical DRPE is utilized. Matlab simulation experiments show that the proposed technique enhances the security level of the DRPE, and at the same time has a better immunity to noise.
An Image-Based Approach For Classification Of Human Micro-Doppler Radar Signatures, Fok Hing Chi Tivive, Son Lam Phung, Abdesselam Bouzerdoum
An Image-Based Approach For Classification Of Human Micro-Doppler Radar Signatures, Fok Hing Chi Tivive, Son Lam Phung, Abdesselam Bouzerdoum
Faculty of Engineering and Information Sciences - Papers: Part A
With the advances in radar technology, there is an increasing interest in automatic radar-based human gait identification. This is because radar signals can penetrate through most dielectric materials. In this paper, an image-based approach is proposed for classifying human micro-Doppler radar signatures. The time-varying radar signal is first converted into a time-frequency representation, which is then cast as a two-dimensional image. A descriptor is developed to extract micro-Doppler features from local time-frequency patches centered along the torso Doppler frequency. Experimental results based on real data collected from a 24-GHz Doppler radar showed that the proposed approach achieves promising classification performance.
Accuracy Of Deformable Image Registration For Contour Propagation In Adaptive Lung Radiotherapy, Nicholas Hardcastle, Wouter Van Elmpt, Dirk De Ruysscher, Karl Bzdusek, Wolfgang A. Tome
Accuracy Of Deformable Image Registration For Contour Propagation In Adaptive Lung Radiotherapy, Nicholas Hardcastle, Wouter Van Elmpt, Dirk De Ruysscher, Karl Bzdusek, Wolfgang A. Tome
Faculty of Engineering and Information Sciences - Papers: Part A
Background: Deformable image registration (DIR) is an attractive method for automatic propagation of regions of interest (ROIs) in adaptive lung radiotherapy. This study investigates DIR for automatic contour propagation in adaptive Non Small Cell Lung Carcinoma patients.
Scalable Fragile Watermarking For Image Authentication, Angela Piper, Reihaneh Safavi-Naini
Scalable Fragile Watermarking For Image Authentication, Angela Piper, Reihaneh Safavi-Naini
Faculty of Engineering and Information Sciences - Papers: Part A
Semi-fragile watermarks are used to detect unauthorised changes to an image, whereas tolerating allowed changes such as compression. Most semi-fragile algorithms that tolerate compression assume that because compression only removes the less visually significant data from an image, tampering with any data that would normally be removed by compression cannot affect a meaningful change to the image. Scalable compression allows a single compressed image to produce a variety of reduced resolution or reduced quality images, termed subimages, to suit the different display or bandwidth requirements of each user. However, highly scaled subimages remove a substantial fraction of the data in …
The Joint Effect Of Image Blur And Illumination Distortions For Mobile Visual Search Of Print Media, Yi Cao, Christian Ritz, Raad Raad
The Joint Effect Of Image Blur And Illumination Distortions For Mobile Visual Search Of Print Media, Yi Cao, Christian Ritz, Raad Raad
Faculty of Engineering and Information Sciences - Papers: Part A
Mobile Visual Search (MVS) is an emerging area of research given the explosion of smart and powerful mobile devices. Typically, the performance of MVS applications is influenced by the implemented image processing algorithms as well as various distortions that occur when capturing images by the mobile camera. This paper examines the joint effect of two common distortions, namely the illumination changes and image blurring on image matching accuracy for print media when using four state-of-the art local feature algorithms. Results obtained for a database of real camera images captured by two different camera models show that the illumination changes have …
Two-Stage Through-The-Wall Radar Image Formation Using Compressive Sensing, Van Ha Tang, Abdesselam Bouzerdoum, Son Lam Phung
Two-Stage Through-The-Wall Radar Image Formation Using Compressive Sensing, Van Ha Tang, Abdesselam Bouzerdoum, Son Lam Phung
Faculty of Engineering and Information Sciences - Papers: Part A
We introduce a robust image-formation approach for through-the-wall radar imaging (TWRI). The proposed approach consists of two stages involving compressive sensing (CS) followed by delay-and-sum (DS) beamforming. In the first stage, CS is used to reconstruct a complete set of measurements from a small subset collected with a reduced number of transceivers and frequencies. DS beamforming is then applied to form the image using the reconstructed measurements. To promote sparsity of the CS solution, an overcomplete Gabor dictionary is employed to sparsely represent the imaged scene. The new approach requires far fewer measurement samples than the conventional DS beamforming and …
On The Dosimetric Effect And Reduction Of Inverse Consistency And Transitivity Errors In Deformable Image Registration For Dose Accumulation, Edward T. Bender, Nicholas Hardcastle, Wolfgang A. Tome
On The Dosimetric Effect And Reduction Of Inverse Consistency And Transitivity Errors In Deformable Image Registration For Dose Accumulation, Edward T. Bender, Nicholas Hardcastle, Wolfgang A. Tome
Faculty of Engineering and Information Sciences - Papers: Part A
Purpose: Deformable image registration (DIR) is necessary for accurate dose accumulation between multiple radiotherapy image sets. DIR algorithms can suffer from inverse and transitivity inconsistencies. When using deformation vector fields (DVFs) that exhibit inverse-inconsistency and are nontransitive, dose accumulation on a given image set via different image pathways will lead to different accumulated doses. The purpose of this study was to investigate the dosimetric effect of and propose a postprocessing solution to reduce inverse consistency and transitivity errors. Methods: Four MVCT images and four phases of a lung 4DCT, each with an associated calculated dose, were selected for analysis. DVFs …
Key-Based Scrambling For Secure Image Communication, Prashan Premaratne, Malin Premaratne
Key-Based Scrambling For Secure Image Communication, Prashan Premaratne, Malin Premaratne
Faculty of Engineering and Information Sciences - Papers: Part A
Secure image communication is becoming increasingly important due to theft and manipulation of its content. Law enforcement agents may find it increasingly difficult to stay afloat above the ill intentions of hackers. We have been able to develop an image scrambling algorithm that is very simple to implement but almost impossible to breach with a probability less than 5x10− 300. This is possible due to the fact that a user may purchase or acquire rights for an intended image by specifying a 'key' that can form a sequence of numbers 10 to 100 in length. The content provider uses this …
New Structural Similarity Measure For Image Comparison, Prashan Premaratne, Malin Premaratne
New Structural Similarity Measure For Image Comparison, Prashan Premaratne, Malin Premaratne
Faculty of Engineering and Information Sciences - Papers: Part A
Subjective quality measures based on Human Visual System for images do not agree well with well-known metrics such as Mean Squared Error and Peak Signal to Noise Ratio. Recently, Structural Similarity Measure (SSIM) has received acclaim due to its ability to produce results on a par with Human Visual System. However, experimental results indicate that noise and blur seriously degrade the performance of the SSIM metric. Furthermore, despite SSIM's popularity, it does not provide adequate insight into how it handles 'structural similarity' of images. We propose a structural similarity measure based on approximation level of a given Discrete Wavelet Decomposition …
A Reduced Reference Image Quality Metric Based On Feature Fusion And Neural Networks, Aladine Chetouani, Azeddine Beghdadi, Mohamed Deriche, Abdesselam Bouzerdoum
A Reduced Reference Image Quality Metric Based On Feature Fusion And Neural Networks, Aladine Chetouani, Azeddine Beghdadi, Mohamed Deriche, Abdesselam Bouzerdoum
Faculty of Engineering and Information Sciences - Papers: Part A
A Global Reduced Reference Image Quality Metric (IQM) based on feature fusion using neural networks is proposed. The main idea is the introduction of a Reduced Reference degradation-dependent IQM (RRIQM/D) across a set of common distortions. The first stage consists of extracting a set of features from the wavelet-based edge map. Such features are then used to identify the type of degradation using Linear Discriminant Analysis (LDA). The second stage consists of fusing the extracted features into a single measure using Artificial Neural Networks (ANN). The result is a degradation- dependent IQM measure called the RRIQM/D. The performance of the …
Development Of Fully 3d Image Reconstruction Techniques For Pinhole Spect Imaging, Xuezhu Zhang, Yujin Qi
Development Of Fully 3d Image Reconstruction Techniques For Pinhole Spect Imaging, Xuezhu Zhang, Yujin Qi
Faculty of Engineering and Information Sciences - Papers: Part A
The purpose of this study was to develop fully 3D image reconstruction techniques for pinhole SPECT imaging with our Micro-SPECT system. Our studies involve in the derivation of projection operators, analysis of the sampling characteristics of pinhole SPECT imaging in Radon space, development of effective geometric calibration method for system misalignment, and 3D image reconstruction development and implementation with quantitative degrading compensation for pinhole SPECT with both circular and helical scan. The performances of pinhole SPECT imaging were evaluated using computer simulations and experiments with the Ultra-Micro Hot-Spot phantom, Ultra-Micro Defrise phantom and small-animal imaging. The results from the computer …
Learning-Based Prostate Localization For Image Guided Radiation Therapy, Luping Zhou, Shu Liao, Wei Li, Dinggang Shen
Learning-Based Prostate Localization For Image Guided Radiation Therapy, Luping Zhou, Shu Liao, Wei Li, Dinggang Shen
Faculty of Engineering and Information Sciences - Papers: Part A
Accurate prostate localization is the key to the success of radiotherapy. It remains a difficult problem for CT images due to the low image contrast, the prostate motion, and the uncertain presence of rectum gas. In this paper, a learning based framework is proposed to improve the accuracy of prostate detection in CT. It adaptively determines distinctive feature types at distinctive image regions, thus filtering out features that are salient in image appearance, but irrelevant to prostate localization. Furthermore, an image similarity function is learned to make the image appearance distance consistent with the underlying prostate alignment. The efficacy of …
Discriminative Maximum Margin Image Object Categorization With Exact Inference, Qinfeng Shi, Luping Zhou, Li Cheng, Dale Schuurmans
Discriminative Maximum Margin Image Object Categorization With Exact Inference, Qinfeng Shi, Luping Zhou, Li Cheng, Dale Schuurmans
Faculty of Engineering and Information Sciences - Papers: Part A
Categorizing multiple objects in images is essentially a structured prediction problem: the label of an object is in general dependent on the labels of other objects in the image. We explicitly model object dependencies in a sparse graphical topology induced by the adjacency of objects in the image, which benefits inference, and then use maximum margin principle to learn the model discriminatively. Moreover, we propose a novel exact inference method, which is used in training to find the most violated constraint required by cutting plane method. A slightly modified inference method is used in testing when the target labels are …
Pavement Scene Interpretation And Obstacle Detection By Large Margin Image Labeling, Ke Jia, Nianjun Liu, Lei Wang, Li Cheng
Pavement Scene Interpretation And Obstacle Detection By Large Margin Image Labeling, Ke Jia, Nianjun Liu, Lei Wang, Li Cheng
Faculty of Engineering and Information Sciences - Papers: Part A
This paper presents a novel discriminative approach for pave-ment scene understanding and obstacle detection in real-world images. It overcomes the heavy constraints in previous systems such as a simple background, a specic obstacle, etc. The approach we exploited extends the bundle method to incorporate pairwise correlations among neighboring pixels, and adopts graph-cuts as the inference engine to attain the approximation efficiently. A set of robust features on both local and multi-scale level is also introduced that captures the general statistical properties of pavements and obstacles. The proposed approach is validated on real-world image database, and outperforms the current state-of-the-art visioned-based …
Unified Geometric Calibration And Image Registration For Detached Small Animal Spect/Ct, Xuezhu Zhang, Fangfu Chen, Yongping Li, Qin Wei, Yujin Qi
Unified Geometric Calibration And Image Registration For Detached Small Animal Spect/Ct, Xuezhu Zhang, Fangfu Chen, Yongping Li, Qin Wei, Yujin Qi
Faculty of Engineering and Information Sciences - Papers: Part A
The aim of this study is to develop a unified optimization method to estimate the complete parameters about geometric calibration of system misalignment for both cone-beam CT and pinhole SPECT, and parameters about coordinate transformation of image registration for these detached systems. The uniform projection equations are derived for both cone-beam and pinhole imaging geometry. Complete geometric parameters are estimated by point object phantom with a priori relative position information. The cost function is structured as the least-squares about residual error. The implementation of nonlinear estimation utilizes the Powell method so as to constrain the optimization problem of this study …
Implementing Analytical Geometric And Penetration Response Correction For Keel-Edge Pinhole Spect Image Reconstruction, Xuezhu Zhang, Qiusheng Dai, Yujin Qi
Implementing Analytical Geometric And Penetration Response Correction For Keel-Edge Pinhole Spect Image Reconstruction, Xuezhu Zhang, Qiusheng Dai, Yujin Qi
Faculty of Engineering and Information Sciences - Papers: Part A
The collimator response compensation is very important in high-resolution pinhole SPECT imaging for resolution recovery and quantitative imaging. In this study the pinhole collimator response of the keel-edge aperture was investigated in terms of the geometric response function and penetration response function (GPRF). An approximate numerical method was proposed to implement the geometric and penetration response correction (GPRC) for keel-edge pinhole SPECT image reconstruction. A lookup table for the GPRC was calculated and then was utilized in the 3D pinhole iterative OSEM reconstruction procedure. The performance of the image reconstruction with the GPRC was evaluated using both the phantom and …
Feature Subset Selection For Multi-Class Svm Based Image Classification, Lei Wang
Feature Subset Selection For Multi-Class Svm Based Image Classification, Lei Wang
Faculty of Engineering and Information Sciences - Papers: Part A
Multi-class image classification can benefit much from feature subset selection. This paper extends an error bound of binary SVMs to a feature subset selection criterion for the multi-class SVMs. By minimizing this criterion, the scale factors assigned to each feature in a kernel function are optimized to identify the important features. This minimization problem can be efficiently solved by gradient-based search techniques, even if hundreds of features are involved. Also, considering that image classification is often a small sample problem, the regularization issue is investigated for this criterion, showing its robustness in this situation. Experimental study on multiple benchmark image …
Image Retrieval With Svm Active Learning Embedding Euclidean Search, Lei Wang, Kap Luk Chan, Yap Peng Tan
Image Retrieval With Svm Active Learning Embedding Euclidean Search, Lei Wang, Kap Luk Chan, Yap Peng Tan
Faculty of Engineering and Information Sciences - Papers: Part A
Image retrieval with relevance feedback suffers from the small sample problem. Recently, SVM active learning has been proposed to tackle this problem, showing promising results. However, a small but sufficient number of initially labelled samples are still required to ensure the subsequent active learning efficient and good retrieval performance. In the existing method, the user is asked to label more images before active learning starts. In this paper, a method of embedding Euclidean search into SVM active learning is proposed. With the help of Euclidean search, not only the adverse effect on retrieval performance due to lack of initially labelled …
Bootstrapping Svm Active Learning By Incorporating Unlabelled Images For Image Retrieval, Lei Wang, Kap Luk Chan, Zhihua Zhang
Bootstrapping Svm Active Learning By Incorporating Unlabelled Images For Image Retrieval, Lei Wang, Kap Luk Chan, Zhihua Zhang
Faculty of Engineering and Information Sciences - Papers: Part A
The performance of image retrieval with SVM active learning is known to be poor when started with few labelled images only. In this paper, the problem is solved by incorporating the unlabelled images into the bootstrapping of the learning process. In this work, the initial SVM classifier is trained with the few labelled images and the unlabelled images randomly selected from the image database. Both theoretical analysis and experimental results show that by incorporating unlabelled images in the bootstrapping, the efficiency of SVM active learning can be improved, and thus improves the overall retrieval performance.
An Image Database Semantically Structured Based On Automatic Image Annotation For Content-Based Image Retrieval, Xuejian Xiong, Kap Luk Chan, Lei Wang
An Image Database Semantically Structured Based On Automatic Image Annotation For Content-Based Image Retrieval, Xuejian Xiong, Kap Luk Chan, Lei Wang
Faculty of Engineering and Information Sciences - Papers: Part A
In this paper, we presented a semantically structured image database for content-based image retrieval. A class descriptor is proposed to represent each class using a multiprototype model, which can be obtained by using a learning scheme, such as the Unsupervised Optimal Fuzzy Clustering algorithm, on a group of sample images manually selected from the class. Based on the proposed Image-Class Matching Distance, a similarity measure at the semantic level between an image and classes, images can be annotated by tokens of classes. Hence, composite features of images, including low-level descriptors, class descriptors, and image annotation, are stored into a structured …