Kien-Ping Chung

A parallel architecture for feature extraction in content-based image retrieval system

Although it is possible to retrieve images from database using a unique identification defined by a human operator as an index to images, it is more convenient and natural to search images based on their contents. The principle of content-based image retrieval (CBIR) system is to retrieve images based on the content of the images. One of the important components in CBIR system is to extract the visual features of the images for performing more abstract analysis. However, some of these features are computationally expensive. To solve this issue, a flexible parallel architecture has been proposed to improve the extraction time for the system. This architecture provides the software system with the flexibility of adding and removing any visual features from the system. Thus, a system becomes more intelligent and so it is able to adapt changes caused by the replacement of more appropriate visual features for representing the images.

A hierarchical nonparametric discriminant analysis approach for a content-based image retrieval system

This paper proposes a hierarchical nonparametric discriminant analysis (HNDA) content-based image retrieval (CBIR) system for e-business applications. It has the potential to become an important and integral component for future e-business applications. Developments in CBIR have drawn interest from many researchers and practitioners in recent years. The challenge is how to retrieve the most appropriate or relevant images at the fastest speed. To increase the retrieval speed, most of the systems pre-process the stored images and extract out the essential features. Such scheme only works well for the server type database system. Such approach is not feasible for systems that analyze images in real-time. In this paper, a hierarchical multi-layer statistical discriminant framework is proposed. The system is able to select the most appropriate features by analyzing the newly received images, and then apply a relevance feedback (RF) approach to improve the retrieval accuracy. As the number of features being analyzed is less, an improvement in performance is achieved

The Application of User Log for Online Business Environment using Content-Based Image Retrieval System

Over the past few years, inter-query learning has gained much attention in the research and development of content-based image retrieval (CBIR) systems. This is largely due to the capability of inter-query approach to enable learning from the retrieval patterns of previous query sessions. However, much of the research works in this field have been focusing on analyzing image retrieval patterns stored in the database. This is not suitable for a dynamic environment such as the World Wide Web (WWW) where images are constantly added or removed. A better alternative is to use an images visual features to capture the knowledge gained from the previous query sessions. Based on the previous work (Chung et al., 2006), the aim of this paper is to propose a framework of inter-query learning for the WWW-CBIR systems. Such framework can be extremely useful for those online companies whose core business involves providing multimedia content-based services and products to their customers

Multiple Layar Kernel-Based Approach in Relevance Feedback Content-Based Image Retrieval System

Relevance feedback has drawn intense interest from many researchers in the field of content-based image retrieval (CBIR). In recent years, kernel-based approach has been a popular choice for the implementation of the relevance feedback based CBIR system. This is largely due to its ability to classify patterns with limited sample data. Since most of the kernel approaches reported have been treating the input as a long flat vector, such arrangement may increase the chances of “polluting” the feature element that uniquely identifies the selected image group. This paper proposes a two layer kernel configuration with an objective to improve the retrieval accuracy. While the performance of the two configurations is similar in certain conditions, the proposed configuration has shown to superior when dominant feature element exists that is capable to uniquely identify the selected image group.

A hierarchical discriminant analysis framework for content-based image retrieval system for industrial applications

Content-based image retrieval (CBIR) systems have drawn interest from many researchers in recent years. One of the potential applications of CBIR is in industrial areas where the most relevant drawings or images can be retrieved speedily without the need to memorize any file name or specific key-words. To increase the retrieval speed, most of the systems pre-process the stored images by extracting a set of predefined features. Such scheme only works well for the server type database systems where the images have been stored previously. It is not feasible for systems that analyze images in real-time where the images are stored or added on an ongoing basis. For instance, personal image search engine for the World-Wide-Web is such an example. In this paper, the authors propose a multi-layer statistical discriminant framework which is able to select the most appropriate features to analyze newly received images thereby improving the retrieval accuracy and efficiency.

A Nonparametric Discriminant Approach in Resolving Complex Multi-class Query for Content-based Image Retrieval

Content-based image retrieval (CBIR) systems have drawn intense interest from many researchers in recent years. Over this period, certain degree of success has been achieved in domain-oriented systems for applications such as facial recognition and medical diagnosis. However, the machine learning techniques used in these systems mostly assume that all the targeted images belong to a single group. Thus, most of the research efforts so far have been trying to search for one or a combination of global image features that can be used to differentiate the targeted images from the rest. This is not the case for a generic image database. Quite often, images that are similar semantically may be completely different with the visual context. In this paper, the authors propose a local grouping strategy together with a multiple Gaussian distributions distance ranking approach in an attempt to address the retrieval and ranking of images that belong to multiple disjoint groups.

Application of Hough transform for the identification of secondary flow patterns

This paper reports the application of Hough transform image processing technique to identify the existence of secondary flow patterns in a curved channel. When a fluid is forced through a passage with curvature, secondary flow appearing in spiral motion will superimpose on the main flow. Such phenomena are commonly found in many heat exchange equipment. While secondary flow promotes mixing of the fluids, thereby improving the heat exchange efficiency, limited attempts were made to improve the heat exchange process through the control of the secondary flow. In addition to the highly complex nature of the secondary heat transfer process, it is difficult to quantify or measure the extent of the secondary flow. Hence, much research and investigations have concentrated on the study of the exact mechanism and characteristics of the secondary flow. A novel approach using an image processing technique based on the Hough transform is used to locate and extent of the secondary vortices. The Hough transform, is used to capture the secondary flow system pattern and then it is used to reconstruct the boundary of the secondary flow patterns.

A Feature Vector Approach for Inter-Query Learning for Content-Based Image Retrieval

Use of relevance feedback (RF) in the feature vector model has been one of the most widely used approaches to fine tuning queries for content-based image retrieval (CBIR). We propose a framework that extends RF to capturing the inter-query relationship between current and previous queries. Using the feature vector model, this avoids the need to memorize actual retrieval relationships between actual image indexes and the previous queries. This approach is suited to image database applications in which images are frequently added and removed. In the previous work, we developed a feature vector framework for inter-query learning using statistical discriminant analysis. One weakness of the previous framework is that the criteria for exploring and merging with an existing visual group are based on two constant thresholds, which are selected through trial and error. Another weakness is that it is not suited to mutually interrelated data clusters. Instead of using constant values, we have further extended the framework using positive feedback sample size as a factor for determining thresholds. Experiments demonstrated that our proposed framework outperforms the previous framework.

BER bounds of a trellis coded GMSK signal

In this paper, the error bounds for trellis coded modulation, formed by combining rate-1/2 convolution coding and GMSK (B/sub 0/T=0.3), are presented for an additive white Gaussian noise (AWGN) channel as well as a channel suffering from amplitude fading. In the derivations, a Viterbi decoder is used to select the most likely transmitted sequence. The bounds obtained for the AWGN channel are tight to within 2.5 dB for the code. Computer simulations have also been carried out to verify the analytical bounds over AWGN, and frequency-flat fading channels. Simulated results over AWGN fall within 1 dB of the theoretical lower bound.

Identification of secondary flow pattern in a heated curved rectangular channel using image processing technique

This paper introduces a new application of pattern recognition techniques to the problem of automatic identification of secondary flow patterns in heated curved rectangular channels. The proposed algorithm is based on two observations. Firstly, in the study of secondary flow patterns, it is found that a smaller pair of secondary vortices is generated when the flowrate of a fluid in a curved passage is beyond certain threshold value. Secondly, humans can easily identify the centre of the vortex by visual inspection. This paper proposes an algorithm that can automatically realise the existence and location of such vortices. Such information can be used to provide a feedback signal for the control of the flow within the channel.