Search

Paper checks may have complex background features including fine lines and patterns, which make forgery more difficult. Also, halftoning techniques are used to produce continuous tones and to prevent copies with void pantograph features. When these kinds of checks are scanned, Moire patterns may occur. These patterns make it difficult for customers to examine the scanned check images on ATM (Automated Teller Machine) displays. They also can decrease the classification accuracy of check recognition systems. In this paper, we propose an algorithm to enhance the perceptual quality of scanned check images by reducing the Moire patterns. The proposed algorithm consists of foreground extraction, Moire detection and Moire removal. Subjective image quality assessment was performed to evaluate the degree of improvement. Experimental results show that the proposed algorithm improves perceptual quality while maintaining check recognition accuracy.

International audience; This paper presents a novel method for image manipulation iden- tification of natural images in JPEG format. The image forgery detection technique is based on a signal-dependent noise model that is relevant to de- scribe a natural image acquired by a digital camera. This parametric model is characterized by two fingerprints which are used to falsification identification. The problem is cast in the framework of the hypothesis testing theory. For practical use, the Generalized Likelihood Ratio Tests (GLRT) are presented and their performance is theoretically established. There are different types of image forgery which have been considered in this paper for example re- sampling, Gaussian filtering and median filtering. Experiments with real and simulated images highlight the relevance of the proposed approach.

This paper considers the area of digital forensics (DF). One of the problem in DF is the issue of identification of digital cameras based on images. This aspect has been attractive in recent years due to popularity of social media platforms like Facebook, Twitter etc., where lots of photographs are shared. Although many algorithms and methods for digital camera identification have been proposed, there is lack of research about their robustness. Therefore, in this paper the robustness of digital camera identification with the use of convolutional neural network is discussed. It is assumed that images may be of poor quality, for example, degraded by Poisson noise, Gaussian blur, random noise or removing pixels’ least significant bit. Experimental evaluation conducted on two large image datasets (including Dresden Image Database) confirms usefulness of proposed method, where noised images are recognized with almost the same high accuracy as normal images.

Extreme learning machine (ELM) is a training algorithm for single hidden layer feedforward neural network (SLFN), which converges much faster than traditional methods and yields promising performance. In this paper, we hope to present a comprehensive review on ELM. Firstly, we will focus on the theoretical analysis including universal approximation theory and generalization. Then, the various improvements are listed, which help ELM works better in terms of stability, efficiency, and accuracy. Because of its outstanding performance, ELM has been successfully applied in many real-time learning tasks for classification, clustering, and regression. Besides, we report the applications of ELM in medical imaging: MRI, CT, and mammogram. The controversies of ELM were also discussed in this paper. We aim to report these advances and find some future perspectives.

In this paper, we propose a new method for improving human activity recognition (HAR) datasets in order to increase their classification accuracy when trained with a certain classifier like a Neural Network. In this paper a novel training/testing process for building/testing a classification model for human activity recognition (HAR) is proposed. Traditionally, HAR is done by a classifier that learns what activities a person is doing by training with skeletal data obtained from a motion sensor such as Microsoft Kinect or accelerometer sensors. These skeletal data are the spatial coordinates (x, y, z) of different parts of the human body. In addition to the spatial features that describe current positions in the skeletal data, new features called Shadow Features are used to improve the supervised learning efficiency and accuracy of Neural Network classifiers. Shadow Features are inferred from the dynamics of body movements, thereby modelling the underlying momentum of the performed activities. They provide extra dimensions of information for characterizing activities in the classification process and thus significantly improving the accuracy. These Shadow Features are generated based on the existing features obtained from sensor datasets. In this paper we show that the accuracy of a neural network classifier can be significantly improved by the addition of Shadow Features and we also show that the generation of Shadow Features can be achieved with little time cost, on the fly, with the NVIDIA GPU technology and the CUDA programming model, this way we can improve the Neural Network accuracy at almost no time cost. GPUs are particularly suitable for generating Shadow Features, since they possess multiple cores which can be taken advantage of, in order to generate Shadow Features for multiple data columns in parallel, therefore reducing a lot of processing time, especially when dealing with huge datasets.

Gaming video streaming services are growing rapidly due to new services such as passive video streaming of gaming content, e.g. Twitch.tv, as well as cloud gaming, e.g. Nvidia GeForce NOW and Google Stadia. In contrast to traditional video content, gaming content has special characteristics such as extremely high and special motion patterns, synthetic content and repetitive content, which poses new opportunities for the design of machine learning-based models to outperform the state-of-the-art video and image quality approaches for this special computer generated content. In this paper, we train a Convolutional Neural Network (CNN) based on an objective quality model, VMAF, as ground truth and fine-tuned it based on subjective image quality ratings. In addition, we propose a new temporal pooling method to predict gaming video quality based on frame-level predictions. Finally, the paper also describes how an appropriate CNN architecture can be chosen and how well the model performs on different contents. Our result shows that among four popular network architectures that we investigated, DenseNet performs best for image quality assessment based on the training dataset. By training the last 57 convolutional layers of DenseNet based on VMAF values, we obtained a high performance model to predict VMAF of distorted frames of video games with a Spearman’s Rank correlation (SRCC) of 0.945 and Root Mean Score Error (RMSE) of 7.07 on the image level, while achieving a higher performance on the video level leading to a SRCC of 0.967 and RMSE of 5.47 for the KUGVD dataset. Furthermore, we fine-tuned the model based on subjective quality ratings of images from gaming content which resulted in a SRCC of 0.93 and RMSE of 0.46 using one-hold-out cross validation. Finally, on the video level, using the proposed pooling method, the model achieves a very good performance indicated by a SRCC of 0.968 and RMSE of 0.30 for the used gaming video dataset.

In this paper visualization techniques for modern closed circuit television (CCTV) smart city services are discussed with application to prevention of threats. Unconventional approaches to the intelligent visual data processing are proposed in order to support video surveillance operators, thus to make their work less exhaustive and more effective. Although registration of a huge amount of video data requires development of intelligent and automatic signal processing information extraction techniques, improvement of visualization methods for operators is also a very important task, because of the crucial role the human factor plays and should always play in the decision making, e.g. in the operator reactions to various crisis situations, which can never be fully eliminated by artificial intelligence. Four software based mechanisms connected with a standard or with a slightly extended hardware are proposed as options for the CCTV operators. They utilize rather known ideas but are implemented with new extensions to original algorithms, as well as with additional, innovative modifications and solutions (not presented in the literature). With them they become reliable and efficient tools for the CCTV systems. First, generation of cylindrical panoramas is suggested in order to make long-time video content analysis of a defined area easier and faster. Using panoramas it is possible to reduce the time that is required to watch the video by a factor of hundreds or even thousands and perform an efficient compression of the video stream for the long-time storage. Second, the controlled stereovision option is discussed for quicker and more precise extraction of relevant information from the observed scene. Third, the thermo-vision is analyzed for faultless detection of pedestrians at night. Finally, a novel high dynamic range (HDR) technique is proposed, dedicated to the CCTV systems, in contrast to other typical entertainment oriented HDR approaches, for clear visualization of important and meaningful image details, otherwise invisible. We validated usefulness of the proposed techniques with many experiments presented in this paper.

In this paper we deal with the problem of digital camera identification by photographs. Identifying camera is possible by analyzing camera’s sensor artifacts that occur during the process of photo processing. The problem of digital camera identification has been popular for a long time. Recently many effective and robust algorithms for solving this problem have been proposed. However, almost all solutions are based on state-of-the-art algorithm, proposed by Lukás et al. in 2006. Core of this algorithm is to calculate the so-called sensor pattern noise based on denoising images with wavelet-based denoising filter. Such technique is very efficient, but very time consuming. In this paper we consider tracing cameras by analyzing defects of their optical systems, like vignetting and lens distortion. We show that analysis of vignetting defect allows for recognizing brand of the camera. Lens distortion can be used to distinguish images from different cameras. Experimental evaluation was carried out on 60 devices (compact cameras and smartphones) for a total number of 12 051 images, with support of the Dresden Image Database. Proposed methods do not require denoising images with wavelet-based denoising filter what has a significant influence for speed of image processing, compared with state-of-the-art algorithm.

With the progress of society and the acceleration of urbanization, the problem of multimedia assisted urban traffic is becoming increasingly apparent. Intelligent transportation system arises at the historic moment, but the road traffic information data accumulation of intelligent transportation system is quite large, and the information analysis is complex. Usually, there are huge amounts of data stored in the database of the traffic system, thus we need to analyze and manage the data with scientific methods. As the visualized tools, the multimedia based analysis for the on-time data conditions should also be integrated, therefore, how to effectively model the scenario is challenging. In this paper, we introduce and analyze the structure and related model of recurrent neural network, and apply RNN to traffic big data mining model. According to the characteristics of traffic flow, this paper analyzes the causes of the error data in the process of traffic data collection, and puts forward the corresponding processing methods. The proposed model uses the TensorFlow for development, and applies the proposed deep learning model to implement traffic data mining, and we use charts to visually show the prediction results. Experimental results show that proposed algorithm can effectively mine large traffic multimedia data and has good robustness. At the same time, the prediction accuracy has reached 97.36%.

Although videos appear to be very high-dimensional in terms of duration × frame-rate × resolution, temporal smoothness constraints ensure that the intrinsic dimensionality for videos is much lower. In this paper, we use this idea for investigating Domain Adaptation (DA) in videos, an area that remains under-explored. An approach that has worked well for the image DA is based on the subspace modeling of the source and target domains, which works under the assumption that the two domains share a latent subspace where the domain shift can be reduced or eliminated. In this paper, first we extend three subspace based image DA techniques for human action recognition and then combine it with our proposed Eclectic Domain Mixing (EDM) approach to improve the effectiveness of the DA. Further, we use discrepancy measures such as Symmetrized KL Divergence and Target Density Around Source for empirical study of the proposed EDM approach. While, this work mainly focuses on Domain Adaptation in videos, for completeness of the study, we comprehensively evaluate our approach using both object and action datasets. In this paper, we have achieved consistent improvements over chosen baselines and obtained some state-of-the-art results for the datasets.

Image fusion is the process of combining two or more related images to produce a single output image, containing more relevant information than any one of the input images. The image-fusion process depends upon: the application domain; the number of images undergoing fusion; and the type of imagery, such as whether it is multi-spectral or multi-modal. For clarity of presentation, this paper takes two important fusion methods, Stationary Wavelet Transform (SWT) and Principal Components Analysis (PCA), and applies them to a variety of imagery. Results show that in multi-modal image fusion, PCA appears to perform better for those input images that have different contrast/brightness levels. SWT appears to give better performance when the input images are multi-modal and multi-sensor. A feature of the paper are the number of objective functions employed to evaluate the SWT and PCA methods, allowing the utility of each to be judged. The reader will also find in this paper a concise guide to image fusion techniques with clear recommendations on how to evaluate them.

In the work presented in this paper, we showcase a deep learning system for sentiment analysis and emotion identification in Twitter messages. The system consists of a convolutional neural network used for extracting features from textual data and a classifier for which we experiment with several different classifying algorithms. We train the network using pre-trained word embeddings obtained by unsupervised learning on large text corpora and compare the effectiveness of the different word vectors for this task. We evaluate our system on 3-class sentiment analysis with datasets provided by the Sentiment analysis in Twitter task from the SemEval competition. Additionally, we explore the effectiveness of our approach for emotion identification, by using an automatically annotated dataset with 7 distinct emotions. Our architecture achieves comparable performances to state-of-the-art techniques in the field of sentiment analysis and improves results in the field of emotion identification on the test we use in our evaluation. Moreover, the paper presents several use case scenarios, depicting real-world usage of our architecture.

Creating copy-move forgery became even easier using a wide range of software and platforms. Many algorithms have been proposed to solve the problem, but each one of those algorithms has its own drawbacks. Researchers face many challenges in developing copy-move detection algorithms, and in this paper, we focus on two challenges. The first is the benchmark dataset, and the second involves evaluation metrics. In this paper, we investigate the available copy-move datasets and their advantages and disadvantages. In addition, we discuss the different metrics that have been used by researchers to evaluate the copy-move forgery detection (CMFD) algorithms. On that basis, we suggest the standard specifications of the appropriate copy-move dataset and the metrics that should be used to evaluate the detection algorithms. The findings of this paper will help researchers evaluate their algorithms effectively and fairly essential for developing reliable algorithms.

Human activity recognition using wearable body sensors plays a vital role in the field of pervasive computing. In this paper, we present human activity recognition framework using compressive classification of data collected from a tri-axial accelerometer sensor. Inspired by the theories of random projection, we propose a novel chaotic map for dimensionality reduction of the accelerometer raw data. This framework also involves extraction of time and frequency domain features from the compressed data. These features are used for human activity recognition using a sparse based classifier. Thus, a simultaneous dimension reduction and classification approach is presented in this paper. We experimentally validate the effectiveness of our proposed framework by recognizing 8 common daily human activities performed by 15 subjects of varying age groups. Our proposed framework achieves superior performance in terms of specificity, precision, F-score and overall accuracy.

Recently, convolutional neural networks have achieved great success in image classification. However, the traditional convolutional neural network lacks the ability to distinguish image features, especially for the low resolution images with less feature information. In the vehicle recognition task, it is inevitable to lose some feature information by convolution during the process of the low-level feature is abstracted into the high-level semantic feature. In this paper, an improved convolutional neural network model with higher robustness is proposed, we call it feature fusion convolutional neural network (FFCNN), which can not only produce more discriminative features, but also can avoid interference caused by environmental factors to some extent. Firstly, the strategy of feature fusion is used to fuse the different low-level features in the convolution neural network. Secondly, in order to prevent overfitting, we combine with the network model of sparse and data augmentation to optimize the structure of the network model. The results of the experiment show that the model proposed in this paper has higher recognition accuracy compared with the traditional vehicle recognition methods and the original convolutional neural network models.

Image segmentation is the method of partitioning an image into a group of pixels that are homogenous in some manner. The homogeneity dependents on some attributes like intensity, color etc. Segmentation being a pre-processing step in image processing have been used in the number of applications like identification of objects to medical images, satellite images and much more. The taxonomy of an image segmentation methods collectively can be divided among two categories Traditional methods and Soft Computing (SC) methods. Unlike Traditional methods, SC methods have the ability to simulate human thinking and are flexible to work with their ownership function, have been predominantly applied to the task of image segmentation. SC techniques are tolerant of partial truth, imprecision, uncertainty, and approximations. Soft Computing approaches also having advantages of providing cost-effective, high performance and steadfast solutions. In this survey paper, our emphasis is on core SC approaches like Fuzzy logic, Artificial Neural Network, and Genetic Algorithm used for image segmentation. The contribution lies in the fact to present this paper to the researchers that explore state-of-the-art elaboration of almost all dimensions associated with the image segmentation. The idea is to encapsulate various aspects like emerging topics, methods, evaluation parameters, the problem associated with different type of images, databases, segmentation applications, and other resources so that, it could be advantageous for researchers to make effort in developing new methods for segmentation. The paper accomplishes with findings and concluding remarks.

This paper proposes a selective encryption (SE) algorithm for the almost recently video compressing technique that is called High-Efficiency Video Coding (HEVC). The improvement of our HEVC SE algorithm is to secure real-time HEVC streaming using low computational overhead, same bit rate and ensuring the video format compliance. This contribution is achieved with low computational RC6 algorithm to encrypt subset of binstrings that is binarized using the bypass binary arithmetic coding mode in the context adaptive binary arithmetic coding (CABAC) process of HEVC. This binstrings subset is the non-zero discrete cosine transform (DCT) coefficients sign bits, motion vector difference (MVD) sign bits, remaining absolute values suffix of DCT, MVD absolute values suffix, Sample adaptive offset (SAO) sign bit, residual size, reference picture index and the delta Quantization Parameter (QP). Also, this paper investigates the performance evaluation of using the RC6 with its operation modes within the proposed HEVC CABAC SE. This investigation is done for choosing the best operation mode for RC6 to be suitable for being used in the HEVC real-time application. The security analysis like histogram analysis, correlation coefficients test and key sensitivity test are presented to make sure the protection of HEVC CABAC SE algorithm against brute force and statistical attacks. The performance analysis results prove that the HEVC CABAC SE is highly secure and can be used in real-time HEVC applications.

Face recognition under the influence of complex illumination is a challenging problem to be solved. The common treatments for minimizing the affection of illumination variation are illumination preprocessing and illumination insensitive extraction techniques. However, the methods proposed previously present low performances. To realize high-accuracy recognition under varying illumination, this paper proposes a novel illumination processing algorithm called REC&SIG-SVD algorithm. Above all, singular value decomposition (SVD) is utilized to obtain preliminary high-frequency and low-frequency features of the face image in logarithm domain. This study proposes Sigmoid function which satisfies the principle of diminishing marginal utility to normalize singular values, aiming at calculating effective high-frequency features. Furthermore, this paper proposes a novel illumination normalization method to process low-frequency features, which is based on retina modeling cooperate with an advanced contrast limited adaptive histogram equalization (CLAHE). Meanwhile, enhancement on high-frequency features is realized by threshold-value filtering. Last but not least, the normalized high-frequency and enhanced low-frequency features are reassembled to form the normalized face image. The comparative trials based on Yale B and CMU PIE databases are conducted for our algorithm and other similar techniques as well as deep learning methods. The experimental results demonstrate that REC&SIG-SVD algorithm shows outstanding recognition performance.

It is of great interest in spectral-spatial features classification for hyperspectral images (HSI) with high spatial resolution. This paper presents a novel Spectral-spatial classification method for improving hyperspectral image classification accuracy. Specifically, a new texture feature extraction algorithm exploits spatial texture feature from spectrum is proposed. It employs local binary patterns (LBPs) in order to extract the image texture feature with respect to spectrum information diversity (SID) to measure the differences of spectrum information. The classifier adopted in this work is support vector machine (SVM) because of its outstanding classification performances. In this paper, two real hyperspectral image datasets are used for testing the performance of the proposed method. Our experimental results from real hyperspectral images indicate that the proposed framework can enhance the classification accuracy compare to traditional alternatives.

This paper proposes a method of generating 3D building models with precise geospatial information and a photograph-based facade appearance from panoramic image sequences and digital maps. 3D building modeling research is actively being conducted in areas such as geographic information systems, virtual reality, and augmented reality. However, the generation of realistic 3D models from a ground-level viewpoint is still extremely costly in terms of labor of modeling experts, and collection of data. We have developed a method for 3D building modeling with high-resolution photograph-based appearance information using panoramic images captured at ground level with a mobile mapping system, and geospatial information obtained from a digital map. The proposed method includes 1) pre-processing for tilt correction and base 3D model generation, 2) geo-registration of panoramic images with minimal user input, and 3) building height and shape estimation. This paper presents the proposed method and the quantitative performance measure obtained from a developed test modeling system. In addition, modeling results from an experimental dataset are also presented.

In this paper, a method has been proposed for enhancement of underwater images commonly suffering from low contrast and degraded shading quality. The entirety of the image is changed when we move to capture of images, from air to the water. During capturing some absorption, reflection and scattering effects are induced in the form of contrast, quality and noise as the images look hazy or blurred. This makes one shading to overwhelm the image. For use of underwater resources and overcome these factors the enhancement of the images is required. So, in this paper, we proposed a strategy for underwater image enhancement using Contrast-Limited Adaptive Histogram Equalization (CLAHE) and Percentile methodologies. Finally, these two methodologies are blended for improving the outcomes. Two parameters, namely, Root Mean Squared Error (RMSE) and entropy have been considered for comparing the experimental results of the proposed methodology with the state-of-the-art works. It has been noticed that the proposed system performs better than already existing techniques for underwater image enhancement.

This paper addresses the demand for an intelligent and rapid classification system of skin cancer using contemporary highly-efficient deep convolutional neural network. In this paper, we mainly focus on the task of classifying the skin cancer using ECOC SVM, and deep convolutional neural network. RGB images of the skin cancers are collected from the Internet. Some collected images have noises such as other organs, and tools. These images are cropped to reduce the noise for better results. In this paper, an existing, and pre-trained AlexNet convolutional neural network model is used in extracting features. A ECOC SVM clasifier is utilized in classification the skin cancer. The results are obtained by executing a proposed algorithm with a total of 3753 images, which include four kinds of skin cancers images. The implementation result shows that maximum values of the average accuracy, sensitivity, and specificity are 95.1 (squamous cell carcinoma), 98.9 (actinic keratosis), 94.17 (squamous cell carcinoma), respectively. Minimum values of the average in these measures are 91.8 (basal cell carcinoma), 96.9 (Squamous cell carcinoma), and 90.74 (melanoma), respectively.

The P2P networks play a crucial role in communication and data transfer. During recent years the importance of P2P network is increasing due to its scalability and an easy cost-efficient model. From the last few years, a P2P network has also been playing a significant role in real-time data transmission. The features offered by P2P network for video on demand/ live video streaming are getting popular day by day. The key factors which affect the performance of the P2P network are Scheduling and Overlay. Scheduling schemes like push-based, pull-based scheduling schemes that can be further classified into FCFS, rarity based and deadline based scheduling schemes are not sufficient to fulfill the demand of P2P network. So in this paper, we have defined two new scheduling schemes which overcome the shortcomings of the existing schemes. These are priority based scheduling schemes which calculate the priority by combining the different factors which directly affect the performance of a network. This approach is a combination of priority based chunk selection and bandwidth aware peer selection mechanism, which improves the delay between the peers. A comparative analysis of the new priority-based scheduling scheme is done with previous approaches. The simulation results are shown in this paper verify the approach and depict that performance of network and quality of video receiver is improved with the use of priority based scheduling scheme. Performance analysis of scheduling schemes is evaluated using parameters like startup delay, end-to-end delay, playback delay, and frame redundancy, and for the quality of video the distortion, frame loss ratio of video parameter is used.

With the prevalence of surveillance video, surveillance data can be used in a wide variety of applications where moving object detection, object recognition and pedestrian tracking has become a significant field of research. Especially for pedestrian tracking, it has become an urgent problem to be solved. This paper proposes a novel method based on convolutional neural network called Matching-Siamese network for pedestrian tracking. First, the pedestrians are detected from surveillance videos through Faster-R-CNN and are numbered sequentially. Second, Matching-Siamese network is designed by modifying the structure of the traditional Siamese network to calculate the similarity of two input images. Third, using the image similarity determines whether the probe image of the target pedestrian and each pedestrian images are of the same identity or not. Finally, we track the target pedestrian in all videos by using the identity of probe image and pedestrian images. The results in this paper show that the proposed method outperforms most popular algorithms in terms of accuracy, overlap rate and computational efficiency, especially in the circumstances of object disappearing and reappearing. In addition, our method could use a latest probe pedestrian image to accomplish its tracking in videos ranging from randomly selected time and regions well.

This paper presents a novel approach of image noise removal via integrating geometric morphological patch grouping and adaptive principal component analysis (PCA) transform domain choosing. Image noise removal based on PCA has acquired much attention and success because of the essential difference: the energy of signal concentrates on the small subset of PCA transformed dataset, while the energy of noise evenly spreads over the whole data set. In this paper, the noisy image will be firstly decomposed into overlap patches that contain different content and structure information. However, some of them potentially have similar geometric morphology. So, their gradient map is utilized to compute the dominant orientation of gradient field to group these geometric morphology patches. Such a grouping procedure guarantees that only similar patches are used to perform hard thresholding on the coefficients to remove the noise. Furthermore, as the result and effect of feature extraction are different in different transform domain, a proper one could be adaptively chosen for different types. Finally, a comprehensive empirical evaluation of the proposed method is carried out in terms of accuracy and visuality, and the results reveal that our method appears to be competitive with the state-of-the-art noise removal methods.

The application of human motion monitoring technology based on wearable inertial sensors has achieved great success in the last ten years. But now the research is mainly focused on isolated motion recognition, and there is scarce research on recognition of human motion sequences. In this paper a novel monitoring framework of human motion sequences is proposed based on wearable inertial sensors. The monitoring framework is composed of data acquisition, segmentation, and recognition stages; the main work of this paper is the last two parts. At the segmentation stage, SVD is used to perform pre-segmentation of motion sequence and its purpose is to reduce time in the segmentation process as much as possible. Then a novel similarity measure named MSHsim is proposed to accomplish the fine segmentation. At the recognition stage an HMM is used to recognize the motion sequence. We use four inertial sensors to collect the human motion data. Experiments are implemented to evaluate the performance of the proposed monitoring framework, and from the experiment results, it can be seen that the proposed method may achieve better performance compared to other methods.

In this paper, we propose an efficient no-reference image quality assessment (NR-IQA) method dubbed Center-Surround based Blind Image Quality Assessment (CS-BIQA). Our proposed method employs the Difference of Gaussian (DoG) model to decompose images into several frequency bands, considering the center-surround effect and multi-channel attribute of human visual system (HVS). The integrated natural scene statistics (NSS) features can be further derived from all DoG bands. After that, regression models between the integrated features and associated subjective assessment scores are learned on the training dataset. Subsequently, the learned models are used to predict the quality scores of test images. The main contribution of this paper is twofold. Firstly, the empirical distributions of DoG bands of images are proven to be a Gaussian-like distribution. And thus, the NSS features can be employed to represent the perceptual quality of images. Secondly, different types of distortions are observed to affect different frequency components of images. So, the integrated features extracted from multi-frequency bands are employed in CS-BIQA to achieve stronger distinguishable capability of image quality. Excessive experiments are conducted to indicate that our proposed CS-BIQA metric can represent the perceptual characteristics of HVS. The results on popular IQA databases demonstrate that the CS-BIQA metric is competitive with the state-of-the-art relevant IQA metrics. Furthermore, our proposed method has very low computational complexity, making it more suitable for real-time applications.

In data mining and knowledge discovery applications, outlier detection is a fundamental problem for robust machine learning and anomaly discovery. There are many successful outlier detection methods, including Local Outlier Factor (LOF), Angle-Based Outlier Factor (ABOF), Local Projection Score (LPS), etc. In this paper, we assume that outliers lie in lower density region and they are at relatively larger distance from any points with a higher local density. In order to identify such outliers quantitatively, the paper proposed a decision graph based outlier detection (DGOD) method. The DGOD method works by firstly calculating the decision graph score (DGS) for each sample, where the DGS is defined as ratio between discriminant distance and local density, next ranking samples according to their DGS values, and finally, returning samples with top-r largest DGS values as outliers. Experimental results on synthetic and real-world datasets have confirmed its effectiveness on outlier detection problems, and it is a general and effective information detection method, which is robust to data shape and dimensionality.

Video scene analysis is a recent research topic due to its vital importance in many applications such as real-time vehicle activity tracking, pedestrian detection, surveillance, and robotics. Despite its popularity, the video scene analysis is still an open challenging task and require more accurate algorithms. However, the advances in deep learning algorithms for video scene analysis have been emerged in last few years for solving the problem of real-time processing. In this paper, a review of the recent developments in deep learning and video scene analysis problems is presented. In addition, this paper also briefly describes the most recent used datasets along with their limitations. Moreover, this review provides a detailed overview of the particular challenges existed in real-time video scene analysis that has been contributed towards activity recognition, scene interpretation, and video description/captioning. Finally, the paper summarizes the future trends and challenges in video scene analysis tasks and our insights are provided to inspire further research efforts.

Opinion leaders in online social networks are important for various fields such as public opinion propagation, marketing management, administrative science and even politics. There are often many kinds of relationships in an online social network. Detecting and identifying opinion leaders depending on any one kind of relationship is inaccurate. In this paper, node importance analysis in multi-relationship online social networks was proposed by signalling based on Multi-subnet Composited Complex Networks Model, and considering the characteristics of multiple relationships which would interrelate with each other. Through node importance under multiple relationships, the novel opinion leader detecting algorithm in multi-relationship online social networks is proposed and approved to be efficient by experiments described in this paper.

In order to solve the shortcomings of traditional identity authentication technology, such as low security, low efficiency, a mobile terminal identity authentication scheme based on cloud computing environment is proposed in this paper. In addition, the two-dimensional code technology is used for identity authentication in the cloud computing environment, and the QR coding technology is also used. The dynamic authentication of the mobile terminal is realized by using the two-dimensional code as the information transmission carrier. According to the security analysis, the scheme has simple structure and no need to use the third party equipment, which has high security and adaptability. Finally, the two fusion of two-dimensional code proposed in this paper provides a new way of thinking for the identity authentication based on the cloud environment, and also promotes the development of the Internet of things.

The problem we are trying to solve in this paper is the authenticity identification of speaker digital recording data. As an important basis of the judicial identification, it is crucial to ensure the authenticity of digital recording data. A large number of solutions have been proposed to address the problem. However, classic methods are usually based on logical symbol rather than the physical detection of energy or phase, and these solutions show drawbacks in terms of identification inefficiency, algorithm instability and heavy time overhead. In this paper, we propose to utilize the quantum theory to address the problem. Any tampering operation for digital recording data can lead to the change of charge in the memory, and it can utilize the subtle change to implement the identification. First, we analyze the quantum nature of storage and investigate to extract the transmittance of speech signal as the characteristic value through quantum tunneling theory. Second, aiming at the characteristics of speech signal, we utilize the transmittance to define the rotation angle step function and propose an improved quantum genetic algorithm to detect the change of phase. The proposed method achieves the authenticity identification based on phase detection. The results obtained in this research include the problem can be addressed by phase detection solution based on quantum genetic algorithm, and it shows performance benefits compared with existing solutions by simulation experiment. It is not only theoretically but also practically feasible to realize authenticity identification of speak digital recording data.

In recent years, the use of mobile and cloud Computing is rapidly growing. In other word, mobile and cloud Computing play an important role in everybody’s life today. It carries rich computational resources to mobile users, network operators, as well as cloud computing providers. Moreover, the explosion of multimedia big data (image, video, 3D, etc.) in mobile and cloud computing have created unprecedented opportunities and fundamental security and privacy challenges as they are not just big in volume, but also unstructured and multi-modal. Therefore, the papers of this special issue address variety of security and privacy issues of multimedia big data in mobile and cloud computing and security of other aspects of mobile and cloud networks and also emphasizes many open questions. We anticipate that papers of this special issue will open new entrance for further research and technology improvements in this important area.

Electrocardiographic (ECG) signals often consist of unwanted noises and speckles. In order to remove the noises, various image processing filters are used in various studies. In this paper, FIR and IIR filters are initially used to remove the linear and nonlinear delay present in the input ECG signal. In addition, filters are used to remove unwanted frequency components from the input ECG signal. Linear Discriminant Analysis (LDA) is used to reduce the features present in the input ECG signal. Support Vector Machines (SVM) is widely used for pattern recognition. However, traditional SVM method does not applicable to compute different characteristics of the features of data sets. In this paper, we use SVM model with a weighted kernel function method to classify more features from the input ECG signal. SVM model with a weighted kernel function method is significantly identifies the Q wave, R wave and S wave in the input ECG signal to classify the heartbeat level such as Left Bundle Branch Block (LBBB), Right Bundle Branch Block (RBBB), Premature Ventricular Contraction (PVC) and Premature Atrial Contractions (PACs). The performance of the proposed Linear Discriminant Analysis (LDA) with enhanced kernel based Support Vector Machine (SVM) method is comparatively analyzed with other machine learning approaches such as Linear Discriminant Analysis (LDA) with multilayer perceptron (MLP), Linear Discriminant Analysis (LDA) with Support Vector Machine (SVM), and Principal Component Analysis (PCA) with Support Vector Machine (SVM). The calculated RMSE, MAPE, MAE, R2 and Q2 for the proposed Linear Discriminant Analysis (LDA) with enhanced kernel based Support Vector Machine (SVM) method is low when compared with other approaches such as LDA with MLP, and PCA with SVM and LDA with SVM. Finally, Sensitivity, Specificity and Mean Square Error (MSE) are calculated to prove the effectiveness of the proposed Linear Discriminant Analysis (LDA) with an enhanced kernel based Support Vector Machine (SVM) method.

Image haze removal techniques are extensively used in several outdoor applications. Lack of sufficient knowledge that is required to restore hazy images, the existing techniques usually use various attributes and assign constant values to these attributes. Unsuitable assignment to these attributes does not provide desired dehazing results. The primary objective of this review paper is to provide a structured outline of some well-known haze removal techniques. This paper also focuses on the methods which can assign optimal values to image dehazing attributes. The review has revealed that the meta-heuristic techniques can attain the optimistic haze removal parameters and also concurrently develops an optimistic objective function to estimate the depth map efficiently. Finally, this paper describes the various issues and challenges of image dehazing techniques, which are required to be further studied.

Wireless visual sensor networks (WVSNs) are composed of a large number of visual sensor nodes covering a specific geographical region. This paper addresses the target detection problem within WVSNs where visual sensor nodes are left unattended for long-term deployment. As battery energy is a critical issue it is always challenging to maximize the network’s lifetime. In order to reduce energy consumption, nodes undergo cycles of active-sleep periods that save their battery energy by switching sensor nodes ON and OFF, according to predefined duty cycles. Moreover, adaptive compressive sensing is expected to dynamically reduce the size of transmitted data through the wireless channel, saving communication bandwidth and consequently saving energy. This paper derives for the first time an analytical framework for selecting node’s duty cycles and dynamically choosing the appropriate compression rates for the captured images and videos based on their sparsity nature. This reduces energy waste by reaching the maximum compression rate for each dataset without compromising the probability of detection. Experiments were conducted on different standard datasets resembling different scenes; indoor and outdoor, for single and multiple targets detection. Moreover, datasets were chosen with different sparsity levels to investigate the effect of sparsity on the compression rates. Results showed that by selecting duty cycles and dynamically choosing the appropriate compression rates, the desired performance of detection can be achieved with adaptive CS and at the same time saving energy, where the proposed framework results in an 70% on average energy saving as compared to transmitting the captured image without CS.

In secret image sharing (SIS) schemes, a secret image is shared among a set of n images called stego-images. Each stego-image is preserved by a participant. In the recovery stage, at least k out of n stego-images are required to obtain the secret image, while k − 1 cannot reveal the secret in the sense of perfect secrecy. Hence, SIS guarantees long-term security. However, as the longer the stego-images remain stored, the higher is the probability of being vulnerable against steganalysis. To resolve this issue, this paper proposes the use of proactive secret sharing in an SIS scheme (P-SIS). P-SIS allows the stego-images to be renewed frequently while these are stored, without changing both cover and secret images. However, direct implementation of a proactive SIS requires more embedding rate (ER), causing high steganalysis accuracy detection and loss of quality in the stego-images. Our proposal addresses this issue and presents the combination of a (k, L, n)-threshold ramp secret sharing scheme and least significant bit matching (LSBM) steganography to reduce the steganalysis accuracy detection. The results of the evaluation show effectiveness of the proposal in terms of good quality of the stego-images, accurate recovery of the secret, and reduce the ER. Note that, despite the extensive research of SIS presented until now, only a few previous work is found on steganalysis in SIS. Not only constructing P-SIS scheme, but we also experimented the tolerance of the proposed P-SIS scheme against stganalysis in this paper. As a result, it is shown that the proposed scheme can withstand steganalysis based on machine learning (i.e., based on subtractive pixel adjacency matrix, SPAM).

An elementary visual unit – the viseme is concerned in the paper in the context of preparing the feature vector as a main visual input component of Audio-Visual Speech Recognition systems. The aim of the presented research is a review of various approaches to the problem, the implementation of algorithms proposed in the literature and a comparative research on their effectiveness. In the course of the study an optimal feature vector construction and an appropriate selection of the classifier were sought. The experimental research was conducted on the basis of a spoken corpus in which speech was represented both acoustically and visually. The extracted features represented three types: geometrical, textural and mixed ones. The features were processed employing the classification algorithms based on Hidden Markov Models and Sequential Minimal Optimization. Tests were carried out employing the processed video material recorded with English native speakers who read specially prepared list of commands. The obtained results are discussed in the paper.

Dimensionality reduction has been attracted extensive attention in machine learning. It usually includes two types: feature selection and subspace learning. Previously, many researchers have demonstrated that the dimensionality reduction is meaningful for real applications. Unfortunately, a large mass of these works utilize the feature selection and subspace learning independently. This paper explores a novel supervised feature selection algorithm by considering the subspace learning. Specifically, this paper employs an l 2,1−norm and an l 2,p −norm regularizers, respectively, to conduct sample denoising and feature selection via exploring the correlation structure of data. Then this paper uses two constraints (i.e. hypergraph and low-rank) to consider the local structure and the global structure among the data, respectively. Finally, this paper uses the optimizing framework to iteratively optimize each parameter while fixing the other parameter until the algorithm converges. A lot of experiments show that our new supervised feature selection method can get great results on the eighteen public data sets.

This paper presents a computationally efficient joint imperceptible image watermarking and joint photographic experts group (JPEG) compression scheme. In recent times, the transmission and storage of digital documents/information over the unsecured channel are enormous concerns and nearly all of the digital documents are compressed before they are stored or transmitted to save the bandwidth requirements. There are many similar computational operations performed during watermarking and compression which lead to computational redundancy and time delay. This demands development of joint watermarking and compression scheme for various multimedia contents. In this paper, we propose a technique for image watermarking during JPEG compression to address the optimal trade-off between major performance parameters including embedding and compression rates, robustness and embedding alterations against different known signal processing attacks. The performance of the proposed technique is extensively evaluated in the form of peak signal to noise ratio (PSNR), correlation, compression ratio and execution time for different discrete cosine transform (DCT) blocks and watermark sizes. Embedding is done on DCT coefficients using additive watermarking.

This paper presents two novel directional patterns, a Maximum Response-based Directional Texture Pattern (MRDTP) and a Maximum Response-based Directional Number Pattern (MRDNP), for recognizing the facial emotions in constrained as well as unconstrained situations. The intensity information obtained from the maximum of the edge responses, after applying eight Kirsch masks, is used for the calculation of facial features in MRDTP. In MRDNP, instead of intensity information, the direction number of the maximum response is used. After dividing MRDNP and MRDTP code images into grids, feature vectors are created from the concatenated histograms obtained from the grids. This paper also proposes an effective Generalized Supervised Dimension Reduction System (GSDRS) and uses Extreme Learning Machine with Radial Basis Function (ELM-RBF) classifier for rapid and efficient classification of emotions. Both the proposed patterns are more effective than the existing ones in removing random noise and providing good structural information using prominent edges which help to achieve high classification accuracy when tested with seven datasets.

Contrast enhancement is a very important issue in image processing, pattern recognition and computer vision. Fuzzy logic based techniques perform enhancement using more detailed information of grayness of an image. However, these methods do not perform well on images taken in uncontrolled environment which pose different challenges such as illumination variation, perspective distortion and viewpoint variation. In this paper, we have worked to devise a more robust image enhancement method using fuzzy logic. We propose a novel multi scale entropy based measurement performed using fuzzy logic image processing and utilize it to define and enhance the contrast. For this purpose, we present a mathematical formula to calculate contrast using an adaptive amplification constant. Our approach uses both the local and global entropy information. We have experimented our algorithm on images from Crowd Counting UCF dataset, which contains very dense crowds and complex texture that stands in line with the challenges targeted in this paper. The results show an improved quality than original dataset images and prove that our method enhances the images with a more dynamic ranged contrast as well as better visual results.

Image pattern recognition in the field of big data has gained increasing importance and attention from researchers and practitioners in many domains of science and technology. This paper focuses on the usage of image pattern recognition for big data applications. In this context, the taxonomy of image pattern recognition and big data is revealed. The applications of image pattern recognition for big data, including multimedia, biometrics, and biology/biomedical, are also highlighted. Moreover, the significance of using pattern-based feature reduction in big data is discussed, and machine-learning techniques in pattern recognition applications are presented. A comparison based on the objectives of the approaches is presented to underline the taxonomy. This paper provides a novel review in exploring image recognition approaches for big data, which can be used in future research.

In this paper, a new algorithm for single image super resolution using coupled wavelet and spatial domain dictionary pairs is proposed. The standard deviation parameter, which is approximately scale invariant for low and high resolution patch pairs is employed for clustering. A pair of online coupled dictionaries is learned for each cluster using a low resolution image. The standard deviation measure of a low resolution patch is used to select the appropriate cluster dictionary pair for reconstructing the high resolution counterpart. Experimental results show that the performance of the proposed algorithm is superior to the existing methods in terms of objective and subjective quality measures. The objective image quality is measured in terms of PSNR and SSIM. This paper also proposed an extended algorithm, based on selective sparse representation over a set of coupled dictionary pair. The extended algorithm applies the coupled dictionary based sparse framework for patches having high standard deviation. Whereas, low complexity patch collaging method is used to super resolve low standard deviation valued patches. It is found empirically that a large percentage of patches have low standard deviation values. Moreover, the selective approach significantly reduces the computational complexity without losing the overall reconstruction quality.

In contrast to other services on the Internet, streaming media service needs to fetch data from local disks more frequently, since it always lasts long and the bit rate is quite high. In addition, because of the much slower reading/writing speed of disk than random access memory (RAM), adopting advisable RAM caching policy can efficiently reduce disk I/O. In this paper, we study the problem of reducing disk I/O by using a novel approach. We first provide a new popularity estimate algorithm. Then a formal optimization problem about average disk I/O is presented, and a suboptimal caching algorithm for a special case of the problem is given. Furthermore, a partially observable Markov decision process (POMDP) model is constructed for the caching problem. Based on the model, popularity is taken advantage of to predict clients’ randomized behaviors, data replacing decisions are made when the defined observations occur, and the impact of caching actions on disk performance for future infinite steps is assessed. The method of event-based optimization is applied in search of the optimal stochastic policy. Disk I/O, as the long-run average performance measure, is optimized by applying the policy-gradient algorithm. The simulation results illustrate that data required by clients could be better predicted and lower disk I/O could be achieved by using the model proposed in this paper.

Content-based image retrieval (CBIR) retrieves images from image database based on the visual similarity of query image. For the implementation of CBIR, feature extraction plays a significant role, where colour feature is quite remarkable. But, due to achromatic surfaces or unevenly colored, the role of texture is also important. This paper introduced an efficient and fast CBIR system, which is based on the combination of computationally light weighted colour and texture features viz. chromaticity moment, colour percentile, and local binary pattern. For searching, this paper proposes inverse variance based varying weighted similarity measure (low for high variance feature and high for low variance feature), which reduces the effect of redundancy by assigning the priority to each feature, and effectively retrieves relevant images. In addition, this paper also proposes query image classification and retrieval model by filtering out irrelevant class images using Random Forests (RF) classifier, which recovers the class of a query image based on distinct learning (supervised) of various decision trees. This successful ensemble classification of query images reduces the semantic gap, searching space, and enhances the retrieval performance. Extensive experimental analyses on benchmark databases confirm the usefulness and effectiveness of this work.

Image compression leads to minimize the storage-requirement of an image by reducing the size of the image. This paper presents a bit allocation strategy based on psychovisual threshold in image compression considering a similar idea of audio coding. In the audio coding, a dynamic bit allocation to each signal is related to the concept of variable block coding and bit allocation is performed on either a short block or long block of sample signals. Similarity, in our technique, more bits are assigned to a local block with visually-significant low frequency order, and fewer, with visually-insignificant high frequency order. This paper presents a bit allocation strategy based on psychovisual threshold in image compression. A psychovisual threshold is developed by minimizing the visual impact on the image quality degradation in image frequency coding. This paper investigates the error generated by the discrete cosine transform and sets the maximum acceptable error as a psychovisual threshold. The average reconstruction error per pixel on frequency order is utilized to prescribe a set of bit allocations which provide a significant improvement on the quality of image reconstruction at relatively low bit rates. The experimental results show that our dynamic bit-allocation technique in image compression manages to overcome artifact images in the image output. The proposed bit allocation strategy improves the quality of image reconstruction by about 20% compared to JPEG compression. This bit allocation strategy is designed to replace the traditional role of the quantization process in image compression.

A mixed reality system is a set of interlinked real-time sensors’ and actuators’ data presented within a virtual world with graphics and texts. This paper describes a methodology for the implementation of mixed reality systems that interconnect real-world IoT systems with 3D virtual worlds. In particular, this paper presents the steps for integrating X3DoM graphics with real-time data and discusses the associated technologies from a developer’s point of view. Such a mixed reality 3D system, interconnecting real-time sensors and actuators from a real-world wastewater management system, with all the corresponding components and controls was developed. This mixed reality 3D world was created using open source web technologies that provide users, in this case the building tenants, with the ability to monitor and control the water consumption of their homes at anytime and from anywhere. It provides not only a comprehensive monitoring 3D model of the real building management system but also an easy to use controlling platform, where the commands are executed in the real building infrastructure.

Taking into account of the illumination characteristics of nighttime imaging, a new method for nighttime image dehazing is proposed in this paper. In the first place, based on the color transfer theory, the illumination level of nighttime hazy image can be artificially enhanced through flexibly selecting the reference image. In contrast to the classical model of color transfer with the strategy of overall to overall transfer, the modified model focuses on the different characteristics of various regions in the original image, and it works well even though the nighttime image is interfered by various artificial light sources. In the second place, the enhancement dehazing method based on the theory of guided image filtering is adopted since the key parameters of dehazing method using the atmospheric degradation model are difficult to obtain in the conditions of nighttime imaging. In addition, the key model parameters of guided image filter are selected according to the boundary information of original image rather than the original image itself, which makes it more advantageous for dehazing image taken on the hazy night. The experimental results show that the proposed method has better performance than the classical daytime dehazing methods. Additionally, our method exhibits superior effect compared to the well-known nighttime dehazing method in the aspects of suppressing color distortion and background illumination controlling. The evaluations of the experimental results are established on both the subjective and objective aspects, so the conclusion in this paper is more convincing.

In this paper, we report an effective algorithm for removing both fog and haze from a single image. Existing algorithms based on atmospheric degeneration model generally lead to non-definite solutions for the haze and thick fog images, though they are very efficient for thin fog images. In general, as the algorithms based on vision enhancement cannot automatically adjust weight coefficient for the different structure images, the excessive or inadequate enhancement may emerge. In this paper an original degradation image is primarily segmented into the sky and non-sky regions, and then the main boundaries of non-sky region are extracted using L0 smoothing filter. So our vision enhancement algorithm automatically adjusts weight coefficient according to various structure images. At the stage of vision enhancement, guided image filter famous for its excellent boundary preservation is adopted. As for haze image, the color channel information scattered by haze particles can be obtained in the sky region to make an effective color correction. Both the subjective and objective evaluations of experimental results demonstrate that the proposed algorithm has more outstanding recovery effect for haze and thick fog images. Moreover, the proposed algorithm can judge fog or haze image, which is a by-product of this research.