Yonggang Huang

Medical Image Retrieval with Query-Dependent Feature Fusion Based on One-Class SVM

Due to the huge growth of the World Wide Web, medical images are now available in large numbers in online repositories, and there exists the need to retrieval the images through automatically extracting visual information of the medical images, which is commonly known as content-based image retrieval (CBIR). Since each feature extracted from images just characterizes certain aspect of image content, multiple features are necessarily employed to improve the retrieval performance. Meanwhile, experiments demonstrate that a special feature is not equally important for different image queries. Most of existed feature fusion methods for image retrieval only utilize query independent feature fusion or rely on explicit user weighting. In this paper, we present a novel query dependent feature fusion method for medical image retrieval based on one class support vector machine. Having considered that a special feature is not equally important for different image queries, the proposed query dependent feature fusion method can learn different feature fusion models for different image queries only based on multiply image samples provided by the user, and the learned feature fusion models can re¿ect the different importances of a special feature for different image queries. The experimental results on the IRMA medical image collection demonstrate that the proposed method can improve the retrieval performance effectively and can outperform existed feature fusion methods for image retrieval.

Collaborative Visualization of Large Scale Datasets Using Web Services

Visualization and collaboration of large scale data sets on Internet is still one of biggest challenges in scientific visualization. A distributed, real-time, collaborative system for large scale data like seismic model can be a valuable tool to support scientific research. In this paper, we present a new approach for Web-based synchronized collaborative visualization of large scale data using Web services and rich Web clients which supports collaborative visualization in Web browsers. We use WS-resources in WSRF (Web services resource framework) to maintain states collaborative server. On client side, we design an Ajax-based (asynchronous JavaScript and XML) application using standard Web technologies. A collaborative demonstration of 3D seismic model which is 2GB size is presented and experimental results are showed finally.

Camera Model Identification With Unknown Models

Feature based camera model identification plays an important role for forensics investigations on images. The conventional feature based identification schemes suffer from the problem of unknown models, that is, some images are captured by the camera models previously unknown to the identification system. To address this problem, we propose a new scheme: Source Camera Identification with Unknown models (SCIU). It has the capability of identifying images of the unknown models as well as distinguishing images of the known models. The new SCIU scheme consists of three stages: 1) unknown detection; 2) unknown expansion; and 3) (K+1 )-class classification. Unknown detection applies a k -nearest neighbours method to recognize a few sample images of unknown models from the unlabeled images. Unknown expansion further extends the set of unknown sample images using a self-training strategy. Then, we address a specific (K+1)-class classification, in which the sample images of unknown (1-class) and known models (K-class) are combined to train a classifier. In addition, we develop a parameter optimization method for unknown detection, and investigate the stopping criterion for unknown expansion. The experiments carried out on the Dresden image collection confirm the effectiveness of the proposed SCIU scheme. When unknown models present, the identification accuracy of SCIU is significantly better than the four state-of-art methods: 1) multi-class Support Vector Machine (SVM); 2) binary SVM; 3) combined classification framework; and 4) decision boundary carving.

A noisy-smoothing relevance feedback method for content-based medical image retrieval

In this paper, we address a new problem of noisy images which present in the procedure of relevance feedback for medical image retrieval. We concentrate on the noisy images, caused by the users mislabeling some irrelevant images as relevant ones, and a noisy-smoothing relevance feedback (NS-RF) method is proposed. In NS-RF, a two-step strategy is proposed to handle the noisy images. In step 1, a noisy elimination algorithm is adopted to identify and eliminate the noisy images. In step 2, to further alleviate the influence of noisy images, a fuzzy membership function is employed to estimate the relevance probabilities of retained relevant images. After noisy handling, the fuzzy support vector machine, which can take into account different relevant images with different relevance probabilities, is adopted to re-rank the images. The experimental results on the IRMA medical image collection demonstrate that the proposed method can deal with the noisy images effectively.

A hierarchical organization approach of multi-dimensional remote sensing data for lightweight Web Map Services

With the rapid development of the World Wide Web, remote sensing (RS) data have become available to a wider range of public/professional users than ever before. Web Map Services (WMSs) provide a simple Web interface for requesting RS data from distributed geospatial databases. RS data providers typically expect to provide lightweight WMSs. They have a low construction cost, and can be easily managed and deployed on standard hardware/software platforms. However, existing systems for WMSs are often heavyweight and inherently hard to manage, due to their improper usage of databases or data storage. That is, they are not suitable for public data services on the Web. In addition, RS data are moving toward the multi-dimensional paradigm, which is characterized by multi-sensor, multi-spectral, multi-temporal and high resolution. Therefore, an efficient organization and storage approach of multi-dimensional RS data is needed for lightweight WMSs, and the efficient WMSs must support multi-dimensional Web browsing. In this paper, we propose a Global Remote Sensing Data Hierarchical Model (GRHM) based on the image pyramid and tiling techniques. GRHM is a logical model that is independent upon physical storage. To support lightweight WMSs, we propose a physical storage structure, and deploy multi-dimensional RS data on Web servers. To further improve the performance of WMSs, a data declustering method based on Hilbert space-filling curve is adopted for the distributed storage. We also provide an Open Geospatial Consortium (OGC) WMS and a Web map system in Web browsers. Experiments conducted on real RS datasets show promising performance of the proposed lightweight WMSs.

ACTGIS: A Web-based collaborative tiled Geospatial image map system

In the past few years, the Web has become a de facto deployment environment for new application systems. Web-based Geospatial image map systems make satellite remote sensing data available to a wider range of public users than ever before. The storage and organization of global massive multi-dimensional remote sensing data have a big impact on the performance of Web mapping systems. To implement a performance-optimized Web mapping system, we propose a global grid division pyramid hierarchical data model for massive remote sensing data management, and adopt Hilbert curve allocation method for data placements. Collaborative web map browsing is also very important for many applications such as crisis management, military activities and government decision-making. However, realizing collaboration functionality on the basis of the stateless HTTP protocol, optimized for client requests, is not trivial. The limitations of the Webs request/response architecture prevent server from pushing real-time dynamic Web data. To address this problem, we propose a web-based interactive collaborative browsing framework based on server push technique (Comet). Moreover, we develop a prototype system called ACTGIS. Finally, we evaluate the prototype system using real remote sensing datasets, which demonstrate the good performance data access in our system.

SOCOM: A Service-Oriented Collaboration Middleware for Multi-User Interaction with Web Services based Scientific Resources

Scientific collaboration has become more and more important in current scientific research. A collaboration environment which simplifies integration and collaboration of heterogeneous scientific resources, for instance scientific computing software and scientific apparatuses, can accelerate progress of scientific research remarkably. We present a novel middleware to facilitate development of distributed and collaborative application for scientific research. A service-oriented collaboration bus is proposed to enable scientific resources to be plugged in and users to interact with these resources collaboratively and transparently. Collaboration bus engines in different organizations can interoperate with each other to support distributed collaboration. We also show our prototype implementation of this middleware and a demonstration of Collaborative Visualization System for Seismic Data in earth science.

Medical image retrieval based on unclean image bags

Traditional content-based image retrieval (CBIR) scheme with assumption of independent individual images in large-scale collections suffers from poor retrieval performance. In medical applications, images usually exist in the form of image bags and each bag includes multiple relevant images of the same perceptual meaning. In this paper, based on these natural image bags, we explore a new scheme to improve the performance of medical image retrieval. It is feasible and efficient to search the bag-based medical image collection by providing a query bag. However, there is a critical problem of noisy images which may present in image bags and severely affect the retrieval performance. A new three-stage solution is proposed to perform the retrieval and handle the noisy images. In stage 1, in order to alleviate the influence of noisy images, we associate each image in the image bags with a relevance degree. In stage 2, a novel similarity aggregation method is proposed to incorporate image relevance and feature importance into the similarity computation process. In stage 3, we obtain the final image relevance in an adaptive way which can consider both image bag similarity and individual image similarity. The experiments demonstrate that the proposed approach can improve the image retrieval performance significantly.

Noisy Smoothing Image Source Identification

Feature based image source identification plays an important role in the toolbox for forensics investigations on images. Conventional feature based identification schemes suffer from the problem of noise, that is, the training dataset contains noisy samples. To address this problem, we propose a new Noisy Smoothing Image Source Identification (NS-ISI) method. NS-ISI address the noise problem in two steps. In step 1, we employ a classifier ensemble approach for noise level evaluation for each training sample. The noise level indicates the probability of being noisy. In step 2, a noise sensitive sampling method is employed to sample training samples from original training set according to the noise level, producing a new training dataset. The experiments carried out on the Dresden image collection confirms the effectiveness of the proposed NS-ISI. When the noisy samples present, the identification accuracy of NS-ISI is significantly better than traditional methods.

Web-Based Multi-Dimensional Medical Image Collaborative Annotation System

Medical image annotation is playing an increasingly important role in clinical diagnosis and medical research. Existing medical image annotation is faced with many demands and challenges. (1) The emergence and sharp increasing speed of multi-dimensional medical images. (2) Image annotation includes not only text annotation, but also graphical annotation, clinical diagnostic information and image content features information. (3) Uneven distribution of medical resources, which makes difficult to aggregate group intelligence from a much larger scale of distributed experts. Most of the present study is texted based within hospitals on single images annotation. It is difficult to organize and manage unstructured medical image annotation and collaborative sharing information. This paper dedicated to the research on collaborative web-based multi-dimensional medical image annotation and retrieval in order to address these problems, overcome the shortcoming of traditional thin client and facilitate medical experts in different locations to exchange views and comments,. It proposed (1) a system architecture that provides authoring, storing, querying, and exchanging of annotations, and supports web-based collaboration. (2) 2D multi-frame and 3D medical image collaborative annotation data model. (3) Collaborative annotation mechanisms.