Shu-Yuan Chen

Retrieval of translated, rotated and scaled color textures

A new method for color texture retrieval using color and edge features is proposed in this study. The proposed method unifies color and edge features rather than simply analyzing only color characteristics. First, the distributions of color and local edge patterns are used to derive a similarity measure for a pair of textures. Then, a retrieval method based on the similarity measure is proposed to retrieve texture images from a database of color textures. Finally, the similarity measure is extended to retrieve texture regions from a database of natural images. Since the proposed feature distributions can resist variations in translation, rotation and scale, our method has the ability to retrieve texture images or regions that change in translation, rotation and/or scale. The effectiveness and practicability of the proposed method have been demonstrated by various experiments.

Pattern classification in DNA microarray data of multiple tumor types

In this paper, we propose a genetic algorithm with silhouette statistics as discriminant function (GASS) for gene selection and pattern recognition. The proposed method evaluates gene expression patterns for discriminating heterogeneous cancers. Distance metrics and classification rules have also been analyzed to design a GASS with high classification accuracy. Moreover, the proposed method is compared to previously published methods. Various experimental results show that our method is effective for classifying the NCI60, the GCM and the SRBCTs datasets. Moreover, GASS outperforms other existing methods in both the leave-one-out cross validations and the independent test for novel data.

Deformed trademark retrieval based on 2D pseudo-hidden Markov model ☆

Abstract A new deformed trademark retrieval method based on two-dimensional pseudo-hidden Markov model (2D PHMM) is proposed in this paper. Most trademark retrieval systems focus on color features, shape silhouettes, or the combination of color and shape. However, these approaches adopted individual silhouettes as shape features, leading to the following two crucial problems. First, most trademarks have various numbers of decomposed components, while the silhouette-based approaches cannot handle the variety correctly. Second, the infringement cases in which trademarks are changed by non-rigid deformation, in particular nonlinear deformation, may escape detection. Thus, our method focuses on the overall appearance of trademarks and incorporates color and shape features into 2D PHMM to tackle the above two problems. The reason to involve 2D PHMM is that it has high tolerance to noise and distortion, moreover, contextual information can be incorporated into it in a natural and elegant way. However, 2D PHMM is computation intensive and sensitive to rotation, scale and translation variations. Thus, it is the main originality of this paper to include the advantages of 2D PHMM but to exclude its disadvantages. As a result, similar trademarks can be retrieved effectively, even those with different numbers of components or non-rigid deformation. Various experiments have been conducted on a trademark database to prove the effectiveness and practicability of the proposed method.

Spot detection for a 2-DE gel image using a slice tree with confidence evaluation

Spot detection is an essential step in 2-DE gel image analysis. The results of protein spot detection may substantially influence subsequent stages of analysis. This study presents a novel method for spot detection with the addition of confidence evaluation for each detected spot. The confidence of a spot provides useful hints for subsequent processing, such as landmark selection, spot quantification and gel image registration. The proposed method takes slices of a gel image in the gray level direction, and builds them into a slice tree, which in turn is adopted to perform spot detection and confidence evaluation. The spot detection software is implemented on Windows using the proposed slice tree. Building a slice tree for a gel image of resolution 1262x720 takes about 1.5 s on an Intel^(C)Pentium^(C)III 1.2 GHz machine with 512 MB of RAM. Spot detection takes about 43 ms after building the slice tree. The detected spots are shown by different colors based on their respective confidence values. Moreover, pointing a mouse over a detected spot shows detailed information about the spot, including the confidence value. Experimental results indicate that confidence values are close to a subjective judgment.

Complementary retrieval for distorted images

In todays computer networks, the amount of digital images increases rapidly and enormously. However, images may be distorted through different types of processing such as histogram equalization, quantization, smoothing, compression, noise corruption, geometric transformation, and changing of illumination. It is imperative to develop an effective method to retrieve the original images from very large image databases because only the original images are stored for economy. In this study, a new image normalization method is first proposed to solve the problem with illumination varying. A complementary retrieval method is then proposed to resist various types of processing. According to the type of distortion, all processing are classified into three distortion categories, low frequency, high frequency and geometric transformation. In addition, different features are resistant to different distortion categories. However, the distortion by which a query image is corrupted is usually unknown. Hence, a complementary analysis is proposed to determine the distortion category for each query image and the feature resistant to the estimated category is used to retrieve the desired original image. As a result, an effective retrieval method is achieved. The feasibility and effectiveness of our method are demonstrated by experimental results.

Automatic Broadcast Soccer Video Analysis, Player Detection, and Tracking Based on Color Histogram

In this chapter, a broadcast soccer video analysis system is proposed for the detection and tracking of the players. Our method consists of two phases. The first one is the scene analysis phase which automatically classifies the video into different scenes based on 2-D Gaussian color model of hue and saturation. An adaptive incremental model update scheme is proposed so that even under shadow condition, good scene analysis result can still be provided. The second one is the player analysis phase. A color histogram-based method is proposed to classify the player with a decision tree, and a linear prediction model based on spatial similarity matrix (SSM) is used for tracking of the players. Experimental results show that the proposed method is simple yet effective.

Classifying subtypes of acute lymphoblastic leukemia using silhouette statistics and genetic algorithms.

Correct classification and prediction of tumor cells is essential for a successful diagnosis and reliable future treatment. In this study, we aimed at using genetic algorithms for feature selection and proposed silhouette statistics as a discriminant function to distinguish between six subtypes of pediatric acute lymphoblastic leukemia by using microarray with thousands of gene expressions. Our methods have shown a better classification accuracy than previously published methods and obtained a set of genes effective to discriminate subtypes of pediatric acute lymphoblastic leukemia. Furthermore, the use of silhouette statistics, offering the advantages of measuring the classification quality by a graphical display and by an average silhouette width, has also demonstrated feasibility and novelty for more difficult multiclass tumor prediction problems.

Prediction of outer membrane proteins by support vector machines using combinations of gapped amino acid pair compositions

Discriminating outer membrane proteins from proteins with other subcellular localizations and with other folding classes are both important to predict farther their functions and structures. In this paper, we propose a method for discriminating outer membrane proteins from other proteins by support vector machines using combinations of gapped amino acid pair compositions. Using 5-fold cross-validation, the method achieves 95% precision and 92% recall on the dataset of proteins with well-annotated subcellular localizations, consisting of 471 outer membrane proteins and 1,120 other proteins. When applied on another dataset of 377 outer membrane proteins and 674 globular proteins belonging to four typical structural classes, the method reaches 96% precision and recall and correctly excludes 98% of the globular proteins. Our method outperforms the OM classifier of PSORTb v.2.0 and a method based on dipeptide composition.

Multiclass microarray data classification using GA/ANN method

This work aims to explore the use of gene expression data in discriminating heterogeneous cancers. We introduce hybrid learning methodology that integrates genetic algorithms (GA) and artificial neural networks (ANN) to find optimal subsets of genes for tissue/cancer classification. This method was tested on two published microarray datasets: (1) NCI60 cancer cell lines and (2) the GCM dataset. Experimental results on classifying both datasets show that our GA/ANN method not only outperformed many reported prediction approaches, but also reduced the number of predictive genes needed in classification analysis.

Genetic algorithms and silhouette measures applied to microarray data classification.

Microarray technology allows large-scale parallel measurements of the expression of many thousands genes and thus aiding in the development of efficient cancer diagnosis and classification platforms. In this paper, we apply the genetic algorithm and the silhouette statistic in conjunction with several distance functions to the problem of multi-class prediction. We examine two widely used sets of gene expression data, measured across sets of tumors, and present the results of classification accuracy on these two datasets by our methods. Our best success rate of tumor classification has better accuracy than many previously reported methods and it provides a useful method towards a complete tool in this domain.