Search

Automated image annotation (AIA) is an important issue in computer vision and pattern recognition, and plays an extremely important role in retrieving large-scale images. In many image annotation approaches, different regions of the image are processed equally, which is inconsistent with the mechanism by which humans understand images. In order to improve the annotation performance of existing AIA approaches, a hybrid AIA approach based on visual attention mechanism (VAM) and the conditional random field (CRF) is proposed. First, since people pay more attention to the salient region of an image during the image recognition process, VAM is implemented for acquiring the salient and non salient regions of the image. Second, support vector machine (SVM) is used to annotate the salient region, and k nearest neighbor (kNN) voting algorithm is used to annotate the non salient regions. Finally, due to the existence of a certain relationship between any two annotation words (also called labels), CRF is calculated to obtain the final label set of each given image. The experimental results confirm that the proposed hybrid AIA approach has ideal annotation performance.

How to retrieve the desired images quickly and accurately from the large scale image database has become a hot topic in the field of multimedia research. Many content-based image retrieval (CBIR) technologies already exist, but they are not always satisfactory. In many applications, the CBIR model based on machine learning relies heavily on the distance metric between samples. Although the traditional distance metric methods are simple and convenient, it is not always appropriate for CBIR tasks. In this paper, a novel distance metric learning (DML) method based on cost sensitive learning (CSL) is studied, and then it is used in a large margin distribution learning machine (LDM) to replace the traditional kernel functions. The improved LDM also takes into account CSL, and which is called CS-DLDM. Finally, CS-DLDM model is applied to CBIR tasks for implementation classification. We compare the proposed CS-DLDM model with other classifiers based on CSL. The experimental results show that the proposed CS-DLDM model not only has satisfactory classification performance but also the lowest misclassification cost, can effectively avoid the class imbalance of sample.

A number of different gene selection approaches based on gene expression profiles (GEP) have been developed for tumour classification. A gene selection approach selects the most informative genes from the whole gene space, which is an important process for tumour classification using GEP. This study presents an improved swarm intelligent optimisation algorithm to select genes for maintaining the diversity of the population. The most essential characteristic of the proposed approach is that it can automatically determine the number of the selected genes. On the basis of the gene selection, the authors construct a variety of the tumour classifiers, including the ensemble classifiers. Four gene datasets are used to evaluate the performance of the proposed approach. The experimental results confirm that the proposed classifiers for tumour classification are indeed effective.

The proposed approach does not require any assumption for software failure data.Implementation of the proposed approach is very easy.The proposed SRGMTEF has a reasonable predictive capability. Software reliability growth model (SRGM) with testing-effort function (TEF) is very helpful for software developers and has been widely accepted and applied. However, each SRGM with TEF (SRGMTEF) contains some undetermined parameters. Optimization of these parameters is a necessary task. Generally, these parameters are estimated by the Least Square Estimation (LSE) or the Maximum Likelihood Estimation (MLE). We found that the MLE can be used only when the software failure data to satisfy some assumptions such as to satisfy a certain distribution. However, the software failure data may not satisfy such a distribution. In this paper, we investigate the improvement and application of a swarm intelligent optimization algorithm, namely quantum particle swarm optimization (QPSO) algorithm, to optimize these parameters of SRGMTEF. The performance of the proposed SRGMTEF model with optimized parameters is also compared with other existing models. The experiment results show that the proposed parameter optimization approach using QPSO is very effective and flexible, and the better software reliability growth performance can be obtained based on SRGMTEF on the different software failure datasets.

Automatic image annotation (AIA) is a task of assigning one or more semantic concepts to a given image and a promising way to achieve more effective image retrieval and analysis. It is a typical classification problem. Due to the semantic gap between low-level visual features and high-level image semantic, the performances of many existing image annotation algorithms are not satisfactory. This paper presents a novel AIA scheme based on improved quantum particle swarm optimization (IQPSO) algorithm for visual features selection (VFS) and an ensemble stratagem based on boosting technique to improve performance of image annotation approach. To maintain the population diversity, the measure method of population diversity and improvement operation are proposed. To achieve better performance of AIA scheme, the measure of population diversity is as a control condition of VFS process. The classification result of an ensemble classifier is as the final annotation result rather than individual classifier. The experimental results confirm that the proposed AIA scheme is very effectiveness. When using proposed AIA scheme over three image datasets respectively, the annotation results are satisfactory. Display Omitted Measure of population diversity is used as a control condition of feature selection.An improvement operation of QPSO is used to avoid premature convergence.Boosting technique is used for creating an ensemble classifier.

Multi-label image annotation (MIA) has been widely studied during recent years and many MIA schemes have been proposed. However, the most existing schemes are not satisfactory. In this paper, an improved multiple kernel learning (IMKL) method of support vector machine (SVM) is proposed to improve the classification accuracy of SVM, then a novel MIA scheme based on IMKL is presented, which uses the discriminant loss to control the number of top semantic labels, and the feature selection approach is also used for improving the performance of MIA. The experiment results show that proposed MIA scheme achieves higher the performance than the existing other MIA schemes, its performance is satisfactory for large image dataset.

A topic discovery approach of the image has been proposed. First, the training images are segmented into some blocks. After clustering blocks, we obtained class set generated by cluster centers, and extracted all nouns from text annotation of each training image to obtain a keyword set. Secondly, the un-label testing image is also segmented into some blocks as same as training images, we calculated the correlation between the block and keyword, and the keyword set for each block may be obtained. Finally, the number of the same keyword appearing in the each block is calculated, we let the keywords with maximum to appear times be as the topics of the image. The experimental results confirm that proposed approach for the image is effectiveness and has good performance.

Narration of Chinese ancient fiction images has been concerned by many researchers. In the today of the digital technology rapid development, it will affect research of the image narration for Chinese ancient fiction. Based on the existing digital technologies, in this paper, an image narration framework in digital environment for Chinese ancient fiction is proposed. In the proposed framework, we analyze the possibility of using variety digital techniques for achieving the narration of Chinese ancient fiction images, whose implementation can provide support for the digital narration of Chinese ancient fiction images.

Increasing spread of computer viruses has caused huge economic losses and psychological influence on people's lives, so building an effective computer virus propagation model to predict the propagation of viruses is necessary. Classical computer virus propagation models can predict the propagation tendency of computer viruses, but they cannot accurately quantify the loss of the user because of these models only were considered from the view of the node connection rate and did not consider the various factors of the participators oneself. In this paper, under the conditions of bounded rationality, the evolutionary game theory is used for building computer virus propagation model. By referencing “replication dynamic” thought, the long-term evolutionary trend of the normal competitive game between normal users and potential attacks using the evolution strategy can be described. According to the Liapunov theorem, the equilibrium and stability of the proposed model are also discussed. Finally, by simulation analysis, we determine the impact of some important parameters of the proposed model for virus propagation. Simulation results show that the proposed model can provide theoretical support for predicting computer virus propagation. Copyright © 2012 John Wiley & Sons, Ltd.

An intelligence robust digital image watermarking algorithm using artificial neural network (ANN) is proposed. In new algorithm, for embedding watermark, the original image first is divided into some N 1 × N 2 small blocks, different embedding strengths are determined by RBFNN classifier according to different textural features of every block after DCT. The experimental results show that the proposed algorithm are robust against common image processing attacks, such as JPEG compression, Gaussian noise, cropping, mean filtering, wiener filtering, and histogram equalization etc. The proposed algorithm achieves a good compromise between the robustness and invisibility, too.

We present an novel watermark scheme of adaptively selecting embedding positions of watermark. In this scheme, we have exploited the capabilities of Fuzzy Neural Network (FNN) approaches for nonlinearly classification the embedding positions. Fuzzy set theory is adopted in a new neural computing algorithm. It has been proved that the novel scheme gets rid of the slow training speed and network parameter sensibility. It is well suited to select embedding positions tasks. Experimental results show that the proposed scheme has good imperceptibility and high robustness to common image processing operators.

Effective prediction of the fault-proneness plays a very important role in the analysis of software quality and balance of software cost, and it also is an important problem of software engineering. Importance of software quality is increasing leading to development of new sophisticated techniques, which can be used in constructing models for predicting quality attributes. In this paper, we use fuzzy c-means clustering (FCM) and radial basis function neural network (RBFNN) to construct prediction model of the fault-proneness, RBFNN is used as a classificatory, and FCM is as a cluster. Object-oriented software metrics are as input variables of fault prediction model. Experiments results confirm that designed model is very effective for predicting a class's fault-proneness, it has a high accuracy, and its implementation requires neither extra cost nor expert's knowledge. It also is automated. Therefore, proposed model was very useful in predicting software quality and classing the fault-proneness.

A robust digital watermark scheme using wavelet-based is presented for copyright protecting. For security, the watermarking is perturbed firstly. A mark sequence may be gotten. For embed watermark imperceptibility, wavelet decomposition for original image is realized. After the original image decomposed into single resolution level, we connect three detail wavelet coefficient matrixes into a big matrix. Mark sequence is embedded into this big matrix according to modular arithmetic. Wavelet decompositions were used for tested image when the watermark extracted, and then an inverse process is requested. The experimental results demonstrate that the proposed technique is robustness to common image processing operators. The experimental result also shows that the repeating embedding watermarking can improve the ability of resisting cropping.

Focusing on the disadvantageous factors of dynamic environment and node's limited performances of Ad Hoc networks (AHN), a new quality of service (QoS) route discovery method, namely QGAANT, based on quantum genetic algorithm (QGA) and ant colony algorithm (ACA) was presented. Using the route switching policy based on probability, we can reduce network overhead caused by flooding. In this paper, QGA was used to adjust the search direction so as to avoid the problem of stagnancy routes. The simulations were carried out to compare the QGAANT, the algorithm based on ant colony only and the traditional on-demand routing algorithm each other. The simulations results show that QGAANT is more adaptable AHN.

Usually, a mobile ad hoc network is confronted with four different damages: random failure, mobile attack, energy limit and selective attack. These attacks will not only destroy network topology, but also will make the transferring capacity and bandwidth of network to be influenced. An assessment of mobile ad hoc network invulnerability is proposed based on the characteristic of mobile ad hoc network. A definition of invulnerability rate is proposed. The invulnerability rate of network is decided by three factors, i.e., connectivity, transferring capacity, and bandwidth of network. At last, we give two experiments to evaluate the performance of the proposed invulnerability assessment method. The results show that mobile ad hoc network has the very strong invulnerability.

To discover important but previously unknown information in video sequences, a discovery method of unknown visual information based on target tracking is proposed in this paper. It provides a worthy approach for support decision. The proposed discovery method analyzes the video using target tracking. This method, adopting detection and correspondence of feature points, target identification and tracking, and target occlusion techniques etc, analyzes video sequences. In the video, using above techniques, we not only study the spatio-temporal relationships of moving target to obtain important information accurately, but also still study the target model problem. This discovery method can be applied in actual domain.

A wavelet packets-based robust blind watermark scheme using chaotic encryption is proposed in this paper. Two chaotic sequences are modulated according characters of chaotic system, using for sub image extraction and watermark encryption. The binary watermark information is changed into one dimension sequence by module arithmetic. An encrypted watermark sequence can get after do XOR arithmetic with the sequence and a binary chaotic sequence after modulating. For embedding watermark, partition the original image into 8 times 8 blocks, extract sub image about one-fourth of total original image blocks by another chaotic sequence then wavelet packet decomposition is applied for it. Embed watermark by using relationships of wavelet packet coefficients. It realizes a blind extraction because extraction watermark doesnpsilat need original image. The experiment results show that proposed scheme is very robust against for various attacks such as JPEG compression, Gaussian noise, cropping and scaling, etc.

In this paper, we propose an efficient approach for detecting the impulse noise from corrupted images. This method is based on the principle that the feature of the digital image is usually local correlation and the feature of the impulse noise is usually located near one of the two ends of the image's maximum and minimum gray values. After the noisy pixel has been detected by the proposed detector, a modified version of the mean filter is proposed to remove the detected impulse noise. Experimental results show that the implementation of the proposed method is simple, and it has better performance than comparison filters with regard to effective noise suppression and preservation of detail, especially when the noise ratio is very high.