Search

There is a concerted effort to build intelligent sea and numerous artificial intelligence technologies have been explored. At present, more and more people are engaged in the research of deep reinforcement learning algorithm, and its mainstream application is in the field of games. Reinforcement learning has conquered chess belonging to complete information game, and Texas poker belonging to incomplete information games. And it reached or even surpassed the highest player level of mankind in E-sports games with huge state space and complex action space. However, reinforcement learning algorithm still has great challenges in fields such as automatic driving. The main reason is that the training of reinforcement learning needs to build an environment for interacting with agents. However, it is very difficult to construct realistic simulation scenes, and there is no guarantee that we will not encounter the state that the agent has not seen. Therefore, it is necessary to explore the simulation scene first. Based on this, this paper mainly studies reinforcement learning in simulation scenario. There are huge challenges in migrating them to real scenario applications, especially in sea missions. Aiming at the heterogeneous multi-agent game confrontation scenario, this paper proposes a sea battlefield game confrontation decision algorithm based on multi-agent deep deterministic policy gradient. The algorithm combines long short-term memory and actor-critic, which not only realizes the convergence of the algorithm in huge state space and action space, but also solves the problem of sparse real rewards. At the same time, imitation learning is integrated into the decision algorithm, which not only improves the convergence speed of the algorithm, but also greatly improves the effectiveness of the algorithm. The results show that the algorithm can deal with a variety of different tactical sea battlefield scenarios, make flexible decisions according to the changes of the enemy, and the average winning rate is close to 90%. [ABSTRACT FROM AUTHOR]

In the field of data exchange, the security of information is the critical aspect. This void in the security helped the evolution of steganography to address the problem of security, especially when highly critical information exchange is the priority. Better and advanced level of security has been attained using the techniques of steganography with cryptography. However, it has been observed that steganographic techniques principled on computationally intelligent or inspired by nature algorithms are gaining much attention. The paper proposes a method for image steganography in transform domain using hybridization of two computationally intelligent algorithms, i.e., firefly optimization algorithm and ant colony optimization algorithm with Huffman encoding to address the gap of high payload. Comparative analysis of the proposed steganographic tool with other existing methods and analysis of the values obtained against some key performance terms like PSNR (peak signal to noise ratio), MSE (mean square error), BER (bit error rate) and SSI (structural similarity index) determines the proposed algorithm as a powerful method. The proposed tool achieves the PSNR of 68.06 dB and the embedding capacity of 985,682 bits. [ABSTRACT FROM AUTHOR]

In recent years, human pose estimation has been widely used in human-computer interaction, augmented reality, video surveillance, and many other fields, but the task of pose estimation still faces many challenges. To address the large number of parameters and complicated calculation in the current mainstream human pose estimation network, this paper proposes a lightweight pose estimation network (Lightweight Polarized Network, referred to as LPNet) based on a polarized self-attention mechanism. First, ghost convolution is used to reduce the number of parameters of the feature extraction network; second, by introducing the polarized self-attention module, the pixel-level regression task can be better solved, the lack of extracted features due to the decrease in the number of parameters can be reduced, and the accuracy of the regression of human keypoints can be improved; finally, a new coordinate decoding method is designed to reduce the error in the heatmap decoding process and improve the accuracy of keypoint regression. The method proposed in this paper was evaluated on the human keypoint detection datasets COCO and MPII, and compared with the current mainstream methods. The experimental results show that the proposed method greatly reduces the number of parameters of the model while ensuring a small loss in accuracy. [ABSTRACT FROM AUTHOR]

Applying ensemble learning to facial expression recognition is an important research field nowadays, but all may not be better than many, the redundant learners in the classifier pool may hinder the ensemble system's performance, so ensemble pruning is needed. Ensemble pruning selects the most suitable subset of classifiers to classify test samples according to the classifier competence. However, the noisy and redundant samples in the validation set will often adversely affect the evaluation of the classifier, making it impossible to select the most suitable classifier. In this paper, a novel ensemble pruning algorithm based on clustering soft label optimization and sorting for facial expression recognition is proposed. First, to increase classifier evaluation objectivity, the novel method uses the clustering optimization model to perform prototype selection and classifier clustering simultaneously. Then the accuracy-based ordering is employed to remove the redundant or poor quality learners, and keep a balance between diversity and accuracy of the ensemble system. Experimental results show that the proposed method outperforms or competes with some state-of-the-art methods on several typical facial expression datasets. [ABSTRACT FROM AUTHOR]

Aiming at the security and efficiency problems in the process of image transmission, an image compression–encryption scheme based on 2D compressive sensing and hyperchaotic system is proposed in this paper. First, we construct a hyperchaotic system with more complex chaotic behavior, which is used to construct the measurement matrix of compressive sensing. Then, two-dimensional compressive sensing is used to compress the image. Compared with one-dimensional sensing, it achieves faster execution efficiency and better image reconstruction quality. Finally, to improve the encryption security, we use the multiplicative inverse operation in the finite domain to diffuse the cipher image after compressive sensing. The experimental simulation results show that the algorithm in this paper has higher execution efficiency, better image reconstruction quality, great security and robustness. [ABSTRACT FROM AUTHOR]

Deep learning has become the main solution for face recognition applications due to its high accuracy and robustness. In recent years, a batch of research on lightweight convolutional neural networks (CNNs) has emerged, bringing new ideas to the economic application of face recognition systems. In this paper, a lightweight face recognition algorithm is proposed to reduce the number of parameters and calculations of the face feature extraction network. The most important part of our approach lies in designing a novel inverted residual shuffle unit (IR-Shuffle). After being trained by ArcFace loss on a GPU workstation, our model built on improved IR-Shuffle blocks of size 1.45 MB achieves an accuracy of 98.65%. In terms of running time, our model is 5 ms faster than the current fastest MobileFaceNet, with only about 0.5% drop in accuracy. The proposed algorithm is implemented and optimized on the Jetson Nano embedded platform, and accurate and real-time deployment of the face recognition system is realized. The system takes 37 ms to perform the complete face detection and recognition and is robust to complex backgrounds and ambient light changes. Experimental results show that our system is of practical application value. [ABSTRACT FROM AUTHOR]

In this paper, an effective method is proposed for restoring underwater image based on color correction and image sharpening. The main purpose of this method is to improve the visibility of underwater images. Traditional methods generally adopt color balancing method to restore the images. However, we found that color balancing has a poor effectiveness on underwater image when the values of blue channel are greater than other two channels, specially, red channel is great of small than blue channel. Therefore, we propose a hybrid method for color correction process based on a principle that exploits the relationship of three channels (red channel, green channel, and blue channel). On the other words, color balancing will be employed to restore the images when the values of red channel approximate to blue channel, while the DCP-base method will be used if otherwise. To enhance the sharpness of underwater image, we employed a sharpening process based on Maximum a Posteriori (MAP) method when color correction has finished. The proposed method also been validated through carrying out experiments on several underwater image datasets which are provided by previous researchers. Our validation has proved that the proposed method has a better performance than the state-of-the-art methods. [ABSTRACT FROM AUTHOR]

In dark or poorly lit environments, it is often difficult for the naked eye to distinguish low-light-level images because of low brightness, low contrast and noise, and it is difficult to perform subsequent image processing (such as video surveillance and target detection). To solve these problems, this paper proposes a low-light-level image enhancement algorithm based on deep learning. First, the low-light-level image is segmented into several super-pixels, and the noise level of each super-pixel is estimated by the ratio of the local standard deviation to the local gradient. Then, the image is inverted and smoothed by a BM3D filter, and the structural filter adaptive method is used to obtain complete images without noise but with the correct texture. Finally, the noise-free image and texture-complete images are applied to the integrated network, which can not only enhance the contrast but also effectively prevent the over-enhancement of the contrast. The experimental results show that this method is superior to traditional methods in terms of both subjective and objective evaluation, and the peak signal–noise ratio and improved structural similarity are 31.64 dB and 91.2%, respectively. [ABSTRACT FROM AUTHOR]

Now, all satellite channels handle high-quality data, including HD video files (1 K, 2 K, and 4 K). High-quality compression software is required to archive and transmit these files accurately and efficiently. HEVC is a standard development technique for high-quality video compression which can produce double compression while preserving video quality. On the other hand, the use of a robust digital video watermark algorithm can solve copyright issues for these types of categories. Therefore, in this paper, an efficient compression-based secure digital watermarking is proposed. In this paper, initially, the video are converted into number of frames. Then, we de-compose the frame using dual-tree complex wavelet transform. After that, to increase the security of the system, the embedding positions are optimally selected using adaptive cuttlefish optimization algorithm. Then, the secrete images are encrypted using elliptic curve cryptography algorithm. After that, the encrypted images are converted into binary bit. Then, the binary bits are embedded into the selected position of video frame. After the encryption process, to decreases the size of encrypted document, H.265 video encoding approach is applied. This method is, effectively, reduce the size of the image, without affecting quality of the image. Finally, the compressed image is transmitted to the receiver. The reverse process of embedding is used for extraction process. The performance of proposed approach is analyzed in terms of different metrics and effectiveness compared with different methods. [ABSTRACT FROM AUTHOR]

To improve the image-processing accuracy and speed of laser three-dimensional imaging system, effectively filter the noise in the image, and effectively optimize the processing speed and image accuracy, this paper proposes an improved adaptive mean-shift image-filtering algorithm based on the traditional mean-shift-filtering algorithm. First, this paper introduces the traditional mean-shift-filtering algorithm, and improves it on the basis of the traditional algorithm. Experiments show that the mean square deviation of the pixels in the area of selection can be used as a feather of image noise and as a control parameter to adjust the size of bandwidth matrix h adaptively, so as to achieve the optimization of accuracy and speed. When the mean square error of the area of selection is large, it shows that the noise is large, and h is increased, so that more pixels are involved in the mean calculation, thus greatly improving the accuracy of calculation. When the mean square error of the area of selection is small, it shows that the noise is small, and then, h is reduced appropriately. A small number of pixels are selected to participate in the mean calculation, which can improve the speed of the algorithm. According to the size of the broadband matrix h, the appropriate pixel values are selected to participate in the process of calculating the mean values, so as to improve the accuracy of the results. Finally, the improved algorithm is verified by comparative experiments. The experimental results show that the improved algorithm can effectively filter the noise in the image and improve the image clarity. Experiments show that the algorithm has good edge-preserving and denoising characteristics. [ABSTRACT FROM AUTHOR]

With the improvement of SAR resolution, super-resolution SAR imaging is more and more widely used in the all-time and all-weather video surveillance and remote sensing imaging. One implementation of super-resolution SAR is that radar works in the spotlight mode. In the case with highly squinted angle and acceleration, the azimuth space variance and the coupling between range and azimuth will become serious in super-resolution imaging. Thus, an azimuth frequency nonlinear chirp scaling algorithm is proposed to solve this problem. Based on the acceleration model, the accurate 2-D spectrum is performed by adopting the method of series reversion. The space-variance of missile-borne SAR in curved flight path is analyzed and an azimuth polynomial phase filter is constructed to make the coefficients of the perturbation function has sufficient flexibility to eliminate the spatial-variant couple terms between range and azimuth. In addition, the gradient operation is used to expand the space-variant coefficients of azimuth modulation term, and the perturbation function is applied to eliminate the space-variant in azimuth direction. The proposed focusing method can process the missile-borne SAR data obtained in spotlight mode in highly squinted angle with high efficiency. The simulation results verify the effectiveness of the proposed imaging approach. The integration of the research in this paper and the deep learning will further pave the way of super-resolution SAR imaging applications in disaster monitoring, security and surveillance. [ABSTRACT FROM AUTHOR]

To improve the imperceptibility without decreasing the steganographic capacity when steganography by a modification to an audio is carried out, this paper proposes an optimal matching path algorithm to reduce the modification to achieve it. A steganography operation unit is defined, which is composed of every N consecutive audio frames and each frame includes M LPC parameters. All these parameters in the same steganography operation unit will form an M × N matrix. A traversal path is a sequence of all these parameters in the matrix. Each parameter in the matrix will be converted into a one-bit binary digit based on two subcodebooks divided by the complementary neighbor vertex (CNV) method. Thus, every traversal path will have a corresponding binary string named PathBinStr. Steganography is performed by computing the Hamming distance between the binary secret and the corresponding PathBinStr and selecting the traversal path with minimal hamming distance expected value to hide the secret via the CNV embedding method. The experimental results show that the proposed steganography algorithm has better performance in audio quality and the modified bit rate decreases from 50 to 37.4% when only 6 traversal paths are predefined, and its processing time can satisfy the real-time requirement of VoIP services. Also, it does not change the audio size, and the theoretical analysis and experiments show its undetectability is good. [ABSTRACT FROM AUTHOR]

Self-embedding fragile watermarking algorithms can perform detection of the manipulated areas as well as recovery of these detected areas. Most of the self-embedding fragile watermarking algorithms are performed block-wise authentication approach. However, entire block is marked as tampered in case one of the pixels in the block is detected as manipulated. This situation decreases the accuracy rate of the authentication process especially against pixel-based attacks such as salt and paper noise adding. Therefore, we present a novel pixel-wise authentication-based self-embedding fragile watermarking method for manipulation detection and recovery. In this proposed method, reference image is divided into four main blocks. Then, each main block is subdivided into 2 × 4 or 4 × 2 sized blocks according to block type determined using recovery quality. Recovery bits of each main block generated from sub-blocks are spreaded into the two main blocks in the other half of the image. For each pixel, two authentication bits are generated using the six most significant bits of the pixel with two pixel position bits and then embedded into the first and second least significant bits of the pixel. In experimental results, pixel-based attacks are applied to the images to demonstrate the success of the presented method. Also, performance of the presented method has been evaluated by applying different size of cropping attacks to the watermarked images. Experimental results show that the proposed method satisfactory detect and recover the manipulated areas. [ABSTRACT FROM AUTHOR]

There has been less attention towards the research on face recognition with partial occlusion. Facial accessories such as masks, sunglasses, and caps, etc., cause partial occlusion which results in a significant performance drop of the face recognition system. In this paper, a novel compressed string matching algorithm based on run-length encoding (CSM-RL) is proposed to solve the partial occlusion problem. In this, the face image is represented by a string sequence that is then compressed using run-length encoding. The proposed CSM-RL algorithm performs string matching between query face and gallery face string sequences by computing the edit distance between string sequences, finally, classifies query face based on the minimum edit distance. The proposed method does not require a classifier and has less time complexity, thus it is more suitable for real-world face recognition applications. The proposed method performs better than the state-of-the-art methods even limited sample images per person are available in the gallery. Extensive experimental results on benchmark face datasets such as AR and Extended Yale-B prove that the proposed algorithm exhibits significant performance improvement both in terms of speed and recognition accuracy for the recognition of partially occluded faces. [ABSTRACT FROM AUTHOR]

Multi-exposure image fusion is an effective method for depicting high dynamic range of the target scene in a single image. However, there are still some problems remaining: the preserving of global contrast, the preserving of the local details in saturated regions, and the existence of halo artifacts. To solve these problems, this paper proposes a new multi-exposure image fusion algorithm with detail enhancement. Firstly, the well-exposedness evaluation function, the chromatic information evaluation function and the local detail preserved function are used to measure weight maps. Then, an improved multi-exposure fusion framework based on pyramid decomposition is proposed to further enhance the details. The experimental results demonstrate that the proposed algorithm can preserve more details than the state-of-the-art. In the view of appreciation, our approach could produce a more realistic brightness distribution of target scene as well as avoid halo artifacts. [ABSTRACT FROM AUTHOR]

Color image quantization is a significant procedure of reducing the huge range of color values of a digital color image into a limited range. In this paper, an automated clustering of pixels and color quantization algorithm is proposed. The ideal number of representative colors is unknown beforehand in most color quantization algorithms. This is an important handicap in most practical cases. The proposed color quantization approach (PPCS) is able to automatically estimate an appropriate number of colors in a quantized palette. Hence, PPCS requires no number of representative colors to be set in advance. This algorithm has two main steps to follow: color palette design and pixel mapping. The color palette is generated by the combination of the entire peaks of all color component histograms. Such that, all color component histogram was smoothed in order to remove unreliable peaks. Next, unreliable colors will be removed from the palette. Then, each pixel in the image will be assigned to the cluster (unit color in the palette) which has the least Euclidean distance. To evaluate the ability of the PPCS, 22 images from Berkeley segmentation dataset have been randomly selected and tested with PPCS and also by two well-known quantization algorithms. The numerical evaluations have been carried out by using computation time, PSNR, MSE, and SSIM performance criteria. Both visual and numerical evaluations reveal that the proposed method presents promising quantization results. Such that, PPCS is ranked first, second, first and first according to PSNR, MSE, SSIM and computation time, respectively. [ABSTRACT FROM AUTHOR]

Due to the low compression performance of traditional compression models, we have developed a new HOA based Fractional KCA with 2D-DWT for improving the multispectral image quality. In this paper, we present a novel multispectral image compression method for improving the complexity by maintaining quality reconstruction and also reducing the size of the storage of multispectral images. Initially, Karhunen–Loeve transform (KLT) is used to remove the spatial redundancies. In the second stage, 2D DWT is used to eliminate the intraband spatial redundancies. In the third stage, Fractional KCA (FKCA) is applied to improve the post-transformation process. FKCA is connected to the band of all wavelet sub-bands to minimize the spatial redundancy between intra sub-bands. Finally, the Hybrid Optimal algorithm (HOA) based FKCA is used to eliminate the residual and information redundancy among the neighboring bands. The experimental analysis of proposed 2D-DWT based Fractional KCA shows that the model improves the performance of compression data in terms of PSNR, MSSI, and VIF. Also, the multispectral image dataset shows the proposed compression model outperforms the existing compression models such as FKLT + PCA, ADWT + OADL, and DWT + DCT [ABSTRACT FROM AUTHOR]

Most state-of-the-art multiple-object tracking (MOT) methods adopt the tracking-by-detection (TBD) paradigm, which is a two-step procedure including the detection module and the tracking module. In these methods, the tracking performance heavily depends on initial detections and data association. In this paper, we present an online MOT method by introducing a single-object tracking (SOT) based on correlation filter. Our contributions lie in twofold: (a) we use the KCF-based SOT in learning of discriminative target appearance relying on handcrafted and deep features and (b) we employ the predicted result to refine the detection mistakes in a new way. Furthermore, we introduce normalize APCE score as an occlusion indicator of tracklet confidence, and build a candidate target hypotheses set to improve the association performance. Both approaches are found beneficial to eliminate the track errors caused by the inability of association algorithm. The experimental results, both qualitative and quantitative on three benchmark datasets, demonstrate that our tracking algorithm achieves comparable or even better results than competitor approaches. [ABSTRACT FROM AUTHOR]

In this paper, we propose a novel visual object tracking method using a part based appearance model. First, a local kernel feature is developed to encode edge information of patches. Next, bounding box of the target is divided into multiple parts. Then, each part uses correlation filter based tracking to predict position in the next frame. The matrix cosine similarity is utilized to measure reliabilities of the patches. Finally, optimal target location is predicted via maximizing the likelihood, which is obtained by adaptively fusing the reliable patches locations. Experimental results illustrate that our algorithm outperforms state-of-the-art tracking methods significantly in terms of accuracy and robustness. [ABSTRACT FROM AUTHOR]

The interest of users towards three-dimensional (3D) video is gaining momentum due to the recent breakthroughs in 3D video entertainment, education, network, etc. technologies. In order to speed up the advancement of these technologies, monitoring quality of experience of the 3D video, which focuses on end user's point of view rather than service-oriented provisions, becomes a central concept among the researchers. Thanks to the stereoscopic viewing ability of human visual system (HVS), the depth perception evaluation of the 3D video can be considered as one of the most critical parts of this central concept. Due to the lack of efficiently and widely utilized objective metrics in literature, the depth perception assessment can currently only be ensured by cost and time-wise troublesome subjective measurements. Therefore, a no-reference objective metric, which is highly effective especially for on the fly depth perception assessment, is developed in this paper. Three proposed algorithms (i.e., Z direction motion, structural average depth and depth deviation) significant for the HVS to perceive the depth of the 3D video are integrated together while developing the proposed metric. Considering the outcomes of the proposed metric, it can be clearly stated that the provision of better 3D video experience to the end users can be accelerated in a timely fashion for the Future Internet multimedia services. [ABSTRACT FROM AUTHOR]

The traditional iterative closest point (ICP) algorithm could register two point sets well, but it is easily affected by local dissimilarity. To deal with this problem, this paper proposes an isotropic scaling ICP algorithm with corner point constraint. First of all, because the corner points can preserve the similarity of the whole shapes, an objective function based on least square error is proposed under the guidance of the corner points. Second, a new ICP algorithm is proposed to complete the isotropic scaling registration. At each iterative step of this new algorithm, the correspondence is built based on the closest point searching, and then a closed-form solution of the transformation is computed. This new algorithm converges monotonically to a local minimum from any given initial scaling transformation. To obtain the expected minimum, the traditional scaling ICP algorithm is applied to compute the initial transformation. The experimental results demonstrate that our algorithm can prevent the influence of the local dissimilarity and improve the registration precision compared with the traditional ICP algorithm. [ABSTRACT FROM AUTHOR]

Visual tracking is an essential technique in computer vision. Even though the notable improvement has been achieved during last few years, tracking multiple objects still remains as a challenging task. In this paper, a novel method for tracking multiple players in soccer videos, which include severe occlusions between players and nonlinear motions by their complex interactions, is introduced. Specifically, we first extract moving objects (i.e., players) by refining results of background subtraction via the edge information obtained from the frame differencing result. Then, we conduct multiscale sampling in foreground regions, which are spatially close to each tracked player, and subsequently computing the dissimilarity between sampled image blocks and each tracked player. Based on the best-matched case, the state of each tracked player (e.g., center position, color, etc.) is consistently updated using the online interpolation scheme. Experimental results in various soccer videos show the efficiency and robustness of our method compared to previous approaches introduced in the literature. [ABSTRACT FROM AUTHOR]

Adaptive streaming allows for dynamic adaptation of the bitrate to varying network conditions, to guarantee the best user experience. Adaptive bitrate algorithms face a significant challenge in correctly estimating the throughput, as the throughput varies widely over time. The current throughput estimation methods cannot distinguish between throughput fluctuations of different amplitude and frequency. In this paper, we propose a throughput estimation method that accurately estimates the throughput based on previous throughput samples. It is robust to short term and small fluctuations, and sensitive to large fluctuations in throughput. Furthermore, we propose a rate adaptive algorithm for video on demand (VoD) services that selects the quality of the video based on the estimated throughput and playback buffer occupancy. The objective of the rate adaptive algorithms is to guarantee high video quality to improve user experience. The proposed algorithm dynamically adjusts the quality level of the video stream. The proposed method selects high quality video segments, while minimizing the risk of playback interruption. Furthermore, the proposed method minimizes the frequency of video rate changes. We show that the algorithm smoothly switches the video rate to improve user experience. Furthermore, we determine that it efficiently utilizes network resources to achieve a high video rate; competing HTTP clients achieve equitable video rates. We also confirm that variations in the playback buffer size and segment duration do not affect the performance of the proposed algorithm. [ABSTRACT FROM AUTHOR]

Despite their popularity, approaches based on salient point descriptors have yet to be proven effective for content-based image retrieval. In this paper, we show how the WINDSURF library can be effectively exploited to assess a fair comparison among the existing alternative approaches based on salient points, which can be contrasted on aspects of both effectiveness and efficiency. Our extensive experimental evaluation, performed on four different image benchmarks, indeed, shows that techniques based on salient point descriptors have effectiveness not better than other existing techniques and are less amenable to be indexed, and thus, their efficiency remains questionable. [ABSTRACT FROM AUTHOR]

The transformation between a planar image and a planar pattern is commonly described by a homography. However, a homography cannot express the exact relation between a spherical panoramic image and a planar pattern. In this paper, we propose a new method of parameter estimation for locating planar patterns in spherical panoramic images. First, we estimate transformation parameters which express how much the target object is scaled, translated and rotated in the spherical panoramic image. Then, we project the planar pattern onto the center of the spherical panoramic image. The center-aligned pattern is used to estimate the transformation parameters. The parameters express the geometric relation between pixels in the spherical panoramic image and pixels in the planar pattern. Our proposed estimation process decreases the error of pattern matching caused by the heavy spherical distortion. Experimental results show that the failure rate of matching is significantly reduced. [ABSTRACT FROM AUTHOR]

Lying-pose human detection is an active research field of computer vision in recent years. It has a good theoretical significance and furthermore many applications, such as victim detection or home service robot. But the study on lying-pose human detection in low-altitude overlooking images have many unsolved problems owing to multiple poses, arbitrary orientation, in-plane rotation, perspective distortion, and time-consuming. In this paper, the proposed framework of human lying-pose detection is optimization and machine learning algorithms inspired by processes of neurobiology suggest and human vision system to select possible object locations. First, the proposed model effectively utilizes binarized normed gradient features to obtain the objectness rapidly based on the vision saliency. Further, deep-learning techniques based on the convolution neural network are trained for learning rich feature hierarchies, in order to obtain the object of lying-pose human from objectness estimation, unlike the classical sliding-window algorithm. Eventually, employed pyramid mean-shift algorithm and rotation-angle recovery method to find position and direction of human lying-pose. The experimental results show that our method is rapid and efficient, and that it achieves state-of-the-art results with our XMULP dataset. [ABSTRACT FROM AUTHOR]

This paper concentrates on the problem of image reconstruction from compressed sensing (CS) measurements in multi-view compressed imaging systems, where each view is acquired independently by CS technique. In order to take advantage of both the inter-view correlation and the spatial prior information in multi-view image sets, a weighted total variation (TV) regularized model, which combines the TV norm of a target view and the TV norm of the corresponding residual, is proposed. To efficiently solve the weighted TV regularization constrained problem, novel algorithms are presented for both the anisotropy TV and the isotropy TV cases. Given the multi-view CS measurements, a sliding window-based recovery framework is also developed to work with the weighted TV-based reconstruction algorithms and produce high-quality results. We show by experiments that the proposed methods greatly outperform the straight forward reconstruction which applies view by view image reconstruction independently, and also have significant advantages over other benchmark methods. [ABSTRACT FROM AUTHOR]

Multiple description coding (MDC) is one of the promising solutions for live video delivery over lossy networks. In this paper, we present a review of MDC techniques based on their application domain and we explain their functionality, with the objective of giving enough insight to designers to decide which MDC scheme is best suited for their specific application based on requirements such as standard compatibility, redundancy tunability, complexity, and extendibility to n-description coding. The focus is mainly on video sources but image-based algorithms applicable to video are considered as well. We also cover the well-known and important problem of drift and solutions to avoid it. [ABSTRACT FROM AUTHOR]

This paper presents a novel method for open-loop watermarking of H.264/AVC bitstreams. Existing watermarking algorithms designed for previous encoders, such as MPEG-2 cannot be directly applied to H.264/AVC, as H.264/AVC implements numerous new features that were not considered in previous coders. In contrast to previous watermarking techniques for H.264/AVC bitstreams, which embed the information after the reconstruction loop and perform drift compensation, we propose a completely new intra-drift-free watermarking algorithm. The major design goals of this novel H.264/AVC watermarking algorithm are runtime-efficiency, high perceptual quality, (almost) no bit-rate increase and robustness to re-compression. The watermark is extremely runtime-efficiently embedded in the compressed domain after the reconstruction loop, i.e., all prediction results are reused. Nevertheless, intra-drift is avoided, as the watermark is embedded in such a way that the pixels used for the prediction are kept unchanged. Thus, there is no drift as the pixels being used in the intra-prediction process of H.264/AVC are not modified. For watermark detection, we use a two-stage cross-correlation. Our simulation results confirm that the proposed technique is robust against re-encoding and shows a negligible impact on both the bit-rate and the visual quality. [ABSTRACT FROM AUTHOR]

In this paper, an adaptive steganography algorithm based on block complexity and matrix embedding is proposed. The matrix embedding constructed by [8,3], [8,4] [8,5] [8,6] and [8,7] linear codes is taken as the basic embedding strategy. Each block with 2 × 4 pixels is used as a cover unit and its block complexity is computed by neighboring pixel difference. The embedding strategy sets are defined for seven kinds of blocks with different complexity. The corresponding embedding strategies are determined by resolving the embedding risk minimization problem. The adaption manners guarantee that message bits are mainly embedded into the regions with higher complexity values. Experiments are performed and the results show that the proposed method can provide a moderate capacity with lower distortion and have higher resistance ability against the chosen steganalytic algorithms. [ABSTRACT FROM AUTHOR]

With advanced mobile devices, the mobile applications of the high-definition display attract a lot of attentions nowadays. The existing image super-resolution methods are computationally inefficient for the high-definition display on the mobile devices. In this paper, we point out that the above critical issue deteriorates the display quality of the high-definition mobile devices. We propose an efficient and effective algorithm to reconstruct the high-resolution images for the mobile devices. Our algorithm outperforms previous approaches in not only smaller running time but also the higher quality of the super-resolution image reconstruction for the mobile devices. [ABSTRACT FROM AUTHOR]

In games, the goals and interests of players are key factors in their behavior. However, techniques used by networked games to cope with infrequent updates and message loss, such as dead reckoning, estimate a player's movements based mainly on previous observations. The estimations are typically made by using dynamics of motion, taking only inertia and some external factors (e.g., gravity, wind) into account while completely ignoring the player's goals (e.g., chasing other players or collecting objects). This paper proposes AntReckoning: a dead reckoning algorithm, inspired from ant colonies, which models the players' interests to predict their movements. AntReckoning incorporates a player's interest in specific locations, objects, and avatars in the equations of motion in the form of attraction forces. In practice, these points of interest generate pheromones, which spread and fade in the game world, and are a source of attraction. To motivate and validate our approach we collected traces from Quake III. We conducted specific experiments that demonstrate the effect of game-related goals, map features, objects, and other players on the mobility of avatars. Our simulations using traces from Quake III and World of Warcraft show that AntReckoning improves the accuracy by up to 44 % over traditional dead reckoning techniques and can decrease the upload bandwidth by up to 32 %. [ABSTRACT FROM AUTHOR]

Lag compensation algorithms used in networked games require programmers to manage the complexities of dealing with both time and shared state. This can make implementing lag compensation techniques challenging. The difficulties in expressing these algorithms limit experimentation with different algorithms and inhibit programmers from exploring the space of the algorithms and testing their effects. The solution is to have a programming model that is better able to deal with time. In this paper, we present such a programming model, timelines. Timelines dramatically reduce the time and effort required to implement lag compensation techniques by allowing for the explicit treatment of time. The timelines model has been implemented as part of the Janus toolkit. [ABSTRACT FROM AUTHOR]

Fully annotated image dataset is required for supervised learning. However, the image labeling process is laborious and monotonous. In this paper, we focus on automatic image labeling for a class-specified image dataset. We propose a weakly supervised approach to localize objects in a class of unlabelled images without using any manually labeled examples. Firstly, an image is segmented based on a multiple segmentation algorithm. Secondly, the segmented regions are mined based on the commonality and saliency to discovery the category pattern in the image. Thirdly, objects are localized based on the weakly supervised learning algorithm. To prove the effectiveness of the proposed approach, we experimentally evaluate the performance of our approach on 12 object classes of the Caltech101 dataset and 2 landmark classes collected from the Internet. The experimental results demonstrate that our approach is effective and accurate to automatically label images. [ABSTRACT FROM AUTHOR]

In video transcoding, accuracy and efficiency of macroblock mode decision are critical issues at the re-encoder side due to the changes in frame size, frame rate, and bit rate. In this paper, a fast macroblock mode decision scheme based on support vector machines is proposed for H.264/AVC baseline profile video transcoder. Features including motion vectors, residual data, pre-encoded macroblock modes, and quantization parameters are extracted from incoming bitstream in both of training stage and classification stage. Feature extraction methods are investigated for spatial resolution transcoder, temporal resolution transcoder, and bit-rate transcoder. After off-line training and simplification of support vectors, the obtained support vector machine classifier can determine macroblock mode in the re-encoder accurately. Extensive experiments are carried out on different types of transcoders and results show that the proposed method can save about 80% in computational complexity compared to full mode search algorithm implemented in the latest H.264/AVC reference software (JM17.1), while maximum peak signal-to-noise ratio is degraded by 0.2-1.1 dB depending on different sequences and bit rate. [ABSTRACT FROM AUTHOR]

In this paper, the main objective is to find an optimal rate allocation strategy that can maximize the total-weighted quality of experience (QoE) associated with multiple video sources over error-prone multi-hop wireless networks based on the particle swarm optimization (PSO) technique which belongs to the family of swarm intelligence algorithms. In video transmission over such wireless networks, many network-based (packet loss, delay etc.) and source-based (encoding quantization level etc.) parameters can impair the perceived video quality. The main contributions of the proposed work are twofold. At first, an optimal bandwidth allocation framework is being developed based on PSO in which by incorporating an accurate video quality metric, the total weighted quality of experience of some competing video sources is being optimized. Second, these optimal rates have been used for differentiated QoE enforcement between multiple competing scalable video sources. The resulting optimal rates can be used as rate-feedbacks for on-line rate adaptation of a moderate scalable video encoder such as H.264/MPEG4 AVC. The aforementioned weight parameters are selected based on the importance of each video sequence's quality and can be associated with some previous service level agreement-based prices. A strong motivation for differentiated quality enforcement is that video sources need to be encoded differently for different resolutions to cater to diverse devices from mobile displays to HDTV displays. Some numerical analysis have been presented to validate the theoretical results and to verify the claims. [ABSTRACT FROM AUTHOR]

To reduce tedious work in cartoon animation, some computer-assisted systems including automatic Inbetweening and cartoon reusing systems have been proposed. In existing automatic Inbetweening systems, accurate correspondence construction, which is a prerequisite for Inbetweening, cannot be achieved. For cartoon reusing systems, the lack of efficient similarity estimation method and reusing mechanism makes it impractical for the users. The semi-supervised graph-based cartoon reusing approach proposed in this paper aims at generating smooth cartoons from the existing data. In this approach, the similarity between cartoon frames can be accurately evaluated by calculating the distance based on local shape context, which is expected to be rotation and scaling invariant. By the semi-supervised algorithm, given an initial frame, the most similar cartoon frames in the cartoon library are selected as candidates of the next frame. The smooth cartoons can be generated by carrying out the algorithm repeatedly to select new cartoon frames after the cartoonists specifying the motion path in a background image. Experimental results of the candidate frame selection in our cartoon dataset suggest the effectiveness of the proposed local shape context for similarity evaluation. The other experiments show the excellent performance on cartoon generation of our approach. [ABSTRACT FROM AUTHOR]

udio classification is an essential task in multimedia content analysis, which is a prerequisite to a variety of tasks such as segmentation, indexing and retrieval. This paper describes our study on multi-class audio classification on broadcast news, a popular multimedia repository with rich audio types. Motivated by the tonal regulations of music, we propose two pitch-density-based features, namely average pitch-density (APD) and relative tonal power density (RTPD). We use an SVM binary tree (SVM-BT) to hierarchically classify an audio clip into five classes: pure speech, music, environment sound, speech with music and speech with environment sound. Since SVM is a binary classifier, we use the SVM-BT architecture to realize coarse-to-fine multi-class classification with high accuracy and efficiency. Experiments show that the proposed one-dimensional APD and RTPD features are able to achieve comparable accuracy with popular high-dimensional features in speech/music discrimination, and the SVM-BT approach demonstrates superior performance in multi-class audio classification. With the help of the pitch-density-based features, we can achieve a high average accuracy of 94.2% in the five-class audio classification task. [ABSTRACT FROM AUTHOR]

This paper presents an efficient approach for copy detection in large archives containing several hundred hours of videos, called ViCopT for Video Copy Tracking. Our video content indexing method consists in computing trends of behaviors of points of interest and then to assign them a label of behavior. Two methods are proposed to assign the labels: one uses heuristic tresholds and the other one uses a clustering algorithm. Such an indexing approach has several interesting properties: it provides a rich, compact and generic description, while labels of behavior provide a high-level description of the video content. A dedicated online retrieval method for copy detection is described, compared and evaluated on a large video database (1,000 h). This evaluation is done on a framework proposed for video copy detection: ViCopT displays excellent robustness to various severe signal transformations and the ability to accurately identify copies from highly similar videos. Other evaluation focuses on the flexibility of ViCopT due to the asymmetry of the video description. This allow the system to be highly scalable and very flexible regarding the situation faced: searching for copies on the internet or monitoring TV stream. [ABSTRACT FROM AUTHOR]

Nowadays, users not only want to watch digital videos on a PC but also use mobile devices to play back multimedia content. Video adaptation algorithms enable the visualization of recorded videos on mobile devices. Based on our Mobile Business system, we have developed a video adaptation application called Mobile Cinema which considers several characteristic features of mobile devices like screen resolution, bit rate or color depth of a display. The functionality of this system is high and it would be much easier to understand the system if the analysis and the design of the algorithms and the communication interfaces are based on predefined processes which we call canonical processes for media production. The definition of these processes would also make the comparison of two video adaptation applications or the exchange of modules between different systems much easier. In this paper, we describe the functionality and requirements of our existing video adaptation application and map them to the canonical processes for media production. [ABSTRACT FROM AUTHOR]

Image preview is a convenient way to browse large or multiple images on small displays. However, current signal-level image resampling algorithms may remove many features of interest in the preview image. In this paper, we propose perceptual image preview which retains more perceptual features such that users can inspect features of interest by viewing the preview image only and without zooming in. This technology has two components, structure enhancement and perceptual feature visualization. Structure enhancement enhances the image structure while suppressing subtle details using a gradient modulation method, thus making the succedent perceptual features more apparent. For perceptual feature visualization, features of interest detected in the picture is visualized on the structure enhanced preview image. We demonstrate with two examples of most commonly used image quality features, image blur and noise. The effectiveness of the proposed method is validated by experimental results. [ABSTRACT FROM AUTHOR]

Infrastructured networks typically employ centralized approaches for group management and information provisioning. In contrast to that, in multi-hop ad hoc networks each node acts as a router as well as sender and receiver. In pure ad hoc networks, no Internet access is available. An additional challenge is to deal with mobility that causes network partitioning and re-organizations. Technically, these problems can be tackled by providing additional uplinks to a backbone network. Those can be used to access resources in the Internet as well as to inter-link multiple ad hoc network partitions, creating a hybrid wireless network. In this paper, we present HyMN, a prototypically implemented hybrid wireless network system optimized for multimedia content providing. Within the ad hoc network, adequate devices are elected to maintain uplinks to a backbone, which can provide for instance multimedia news from certain sports events like Football Championships, Olympic Games and alike. In order to efficiently manage the ad hoc communicating devices, a weighted clustering algorithm is employed. [ABSTRACT FROM AUTHOR]

In this paper, we propose a novel lightweight approximate authentication algorithm that provides efficient protection for wireless video streaming where attacks on the stream are possible, but classical integrity protection algorithms are impractical because bit errors occur naturally. The benefits of the proposed algorithm over other algorithms are fast execution and small message authentication code size. Moreover, the approximate authentication supports error resilient video decoding by dropping seriously damaged packets, thus improving the perceived quality of the video stream. The performance of the algorithm is demonstrated via numerical analysis, simulations and measurements over modeled and real wireless channels. [ABSTRACT FROM AUTHOR]

Abstract  In this paper, we propose a fast mode decision algorithm using a filter bank of Kalman filters for H.264/ AVC. For the highest coding efficiency in H.264/AVC, a macroblock can be coded with seven different block sizes for motion compensation in an inter mode and various spatial prediction modes in an intra mode. The conventional encoder employs a complex technique for mode decision based on a rate-distortion (RD) cost of all possible modes. Hence, for the purpose of selecting the best block mode with the minimum RD cost, the conventional procedure requires much computational burden and a very complex encoding structure. In order to reduce the complexity, we propose a fast algorithm for mode decision based on Kalman filtering to estimate RD cost of a specific block mode. Furthermore, we propose an optimized structure of H.264/AVC encoder to implement the proposed algorithm. Without considerable performance degradation, using SIMD technology, the computer simulation shows that the proposed methods are dramatically faster than the original JM 9.6 encoder. [ABSTRACT FROM AUTHOR]

Abstract  Several types of Content Distribution Networks are being deployed over the Internet today, based on different architectures to meet their requirements (e.g., scalability, efficiency and resiliency). Peer-to-peer (P2P) based Content Distribution Networks are promising approaches that have several advantages. Structured P2P networks, for instance, take a proactive approach and provide efficient routing mechanisms. Nevertheless, their maintenance can increase considerably in highly dynamic P2P environments. In order to address this issue, a two-tier architecture called Omicron that combines a structured overlay network with a clustering mechanism is suggested in a hybrid scheme. In this paper, we examine several sampling algorithms utilized in the aforementioned hybrid network that collect local information in order to apply a selective join procedure. Additionally, we apply the sampling algorithms on Chord in order to evaluate sampling as a general information gathering mechanism. The algorithms are based mostly on random walks inside the overlay networks. The aim of the selective join procedure is to provide a well balanced and stable overlay infrastructure that can easily overcome the unreliable behavior of the autonomous peers that constitute the network. The sampling algorithms are evaluated using simulation experiments as well as probabilistic analysis where several properties related to the graph structure are revealed. [ABSTRACT FROM AUTHOR]

Abstract  Gradient-based algorithms for global motion estimation are effective in many image-processing tasks. However, when analytical estimation of derivatives of objective function is not possible, linear search based algorithms such as Powell perform better than the gradient-based ones. In this paper we propose global motion estimation algorithm that exploits linear search based algorithm, particularly Powell, instead of commonly used gradient-based one. We also introduce a new approach for extracting global motion parameters called Two Step Powell-based GME. Using this approach we further improve the Powell-based GME. The proposed Powell-based GME outperforms Gauss–Newton algorithm (gradient-based) in terms of PSNR. The proposed Two Step Powell GME algorithm outperforms Powell-based GME in terms of PSNR and computational time. [ABSTRACT FROM AUTHOR]

Abstract  Grouping photos of the same event together is extremely useful for the management of personal photo collections. However, most methods cannot be applied to the problem of online event detection in embedded devices because they do not consider hardware constraints or a user’s photo-taking behavior. In this paper, we propose an efficient and effective event detection algorithm for managing personal photo collections in camera phones or digital cameras. The proposed algorithm fuses time and location information, which is deemed the most important information for personal photo management, and works in real time in embedded devices. We model event occurrences in a user’s photo-taking behavior as a Poisson process by imposing certain constraints on calculating the elapsed time. Location information is incorporated into event detection when confidence in a decision based on the Poisson process is not high enough. The algorithm is user-centric because it provides the unique capabilities of accepting and adjusting to user feedback. Our experiment results show that the proposed event detection method has the potential to support emerging multimedia applications in embedded devices. [ABSTRACT FROM AUTHOR]