Search

The rapid development of short video platforms poses new challenges for traditional recommendation systems. Recommender systems typically depend on two types of user behavior feedback to construct user interest profiles: explicit feedback (interactive behavior), which significantly influences users' short-term interests, and implicit feedback (viewing time), which substantially affects their long-term interests. However, the previous model fails to distinguish between these two feedback methods, leading it to predict only the overall preferences of users based on extensive historical behavior sequences. Consequently, it cannot differentiate between users' long-term and short-term interests, resulting in low accuracy in describing users' interest states and predicting the evolution of their interests. This paper introduces a video recommendation model called CAT-MF Rec (Cross Attention Transformer-Mixed Feedback Recommendation) designed to differentiate between explicit and implicit user feedback within the DIEN (Deep Interest Evolution Network) framework. This study emphasizes the separate learning of the two types of behavioral feedback, effectively integrating them through the cross-attention mechanism. Additionally, it leverages the long sequence dependence capabilities of Transformer technology to accurately construct user interest profiles and predict the evolution of user interests. Experimental results indicate that CAT-MF Rec significantly outperforms existing recommendation methods across various performance indicators. This advancement offers new theoretical and practical insights for the development of video recommendations, particularly in addressing complex and dynamic user behavior patterns. [ABSTRACT FROM AUTHOR]

Short video applications are popular nowadays. Typically, a short video may be played for a few seconds or swiped away by users at any time. Due to this uncertain user behavior, video chunks should be preloaded to ensure a smooth viewing process for users. In other words, the short video preloading scheme is crucial to the quality of experience (QoE) of users and revenue. Specifically, the short video preloading scheme should determine which video chunk to download, when to download it, and at what bitrate. Existing schemes either fail to consider all these factors together or find it hard to make the best decision. In this work, we argue that the selection of downloaded video chunks is the foremost task due to the uncertain user behaviors and the corresponding huge number of possible playing sequences of video chunks. As a step further, we propose the playing time-aware chunk selection (PaCs) scheme, which downloads the video chunk with the smallest expected playing time first. After the selection of the video chunk, the bitrate is selected according to the classic MPC algorithm and then whether the downloading process is paused or executed is discussed. In total, PaCs can improve the score consisting of the QoE of downloaded video chunks and the utilized network bandwidth under different conditions. Simulations confirm that PaCs achieves a higher score than the existing scheme Dashlet proposed in NSDI 2024. [ABSTRACT FROM AUTHOR]

Our paper presents an adaptive multiscale retinex algorithm and a new wavelet energy metric to improve low-light video captured on mobile devices. Initially, we extract RGB frames from the video and convert them to hue-saturation-value (HSV) format, preserving the hue channel to prevent common RGB colour shifting issues. Saturation channel enhancement is achieved through histogram equalization (HE), extending the dynamic range. The adaptive retinex algorithm enhances the value channel, quantified by our new wavelet energy metric. Combining the modified value and saturation channels improves the contrast of the reconstructed image. As a final step, we transform the HSV video back to RGB and restore naturalness using a modified colour restoration technique. The proposed approach has been tested on over 300 images and videos. It is evident from the experimental results presented that the proposed method lowers noise and halo artifacts more effectively than existing methods. [ABSTRACT FROM AUTHOR]

The anomaly or abnormality detection in crowded scenes helps in identifying the violence and protecting the people from severe damage. Thus, there is a need to detect the anomalies with the classifier for learning information along with the usage of huge architectures. A new anomaly detection model is implemented in this model. The collected data is fed to optimal ensemble pattern extraction scheme through techniques like Local binary patterns (LBP), Local Gradient Pattern (LGP), and Local Tetra Pattern (LTrP). The weights are tuned by a new hybrid Spiral Search-based Black Widow Glowworm Swarm Optimization (SS-BWGSO) for getting the optimal ensemble patterns. Next, anomaly frame classification is carried out by optimized VGG16+ResNet technique, where the hyperparameters of VGG16 and ResNet are tuned by SS-BWGSO algorithm. Finally, anomaly detection is performed by the YOLOV3 classifier. Throughout the result analysis the higher performance of the designed technique is observed over the classical methods. [ABSTRACT FROM AUTHOR]

In this article, we introduce a nonconvex tensor recovery approach, which employs the powerful ket augmentation technique to expand a low order tensor into a high‐order one so that we can exploit the advantage of tensor train (TT) decomposition tailored for high‐order tensors. Moreover, we define a new nonconvex surrogate function to approximate the tensor rank, and develop an auto‐weighted mechanism to adjust the weights of the resulting high‐order tensor's TT ranks. To make our approach robust, we add two mode‐unfolding regularization terms to enhance the model for the purpose of exploring spatio‐temporal continuity and self‐similarity of the underlying tensors. Also, we propose an implementable algorithm to solve the proposed optimization model in the sense that each subproblem enjoys a closed‐form solution. A series of numerical results demonstrate that our approach works well on recovering color images and videos. [ABSTRACT FROM AUTHOR]

This paper proposes an efficient encoding framework based on Scalable High Efficiency Video Coding (SHVC) technology, which supports both low- and high-resolution 2D videos as well as stereo 3D (S3D) video simultaneously. Previous studies have introduced Cross-View SHVC, which encodes two videos with different viewpoints and resolutions using a Cross-View SHVC encoder, where the low-resolution video is encoded as the base layer and the other video as the enhancement layer. This encoder provides resolution diversity and allows the decoder to combine the two videos, enabling 3D video services. Even when 3D videos are composed of left and right videos with different resolutions, the viewer tends to perceive the quality based on the higher-resolution video due to the binocular suppression effect, where the brain prioritizes the high-quality image and suppresses the lower-quality one. However, recent experiments have shown that when the disparity between resolutions exceeds a certain threshold, it can lead to a subjective degradation of the perceived 3D video quality. To address this issue, a conditional replenishment algorithm has been studied, which replaces some blocks of the video using a disparity-compensated left-view image based on rate–distortion cost. This conditional replenishment algorithm (also known as VEI technology) effectively reduces the quality difference between the base layer and enhancement layer videos. However, the algorithm alone cannot fully compensate for the quality difference between the left and right videos. In this paper, we propose a novel encoding framework to solve the asymmetry issue between the left and right videos in 3D video services and achieve symmetrical video quality. The proposed framework focuses on improving the quality of the right-view video by combining the conditional replenishment algorithm with Cross-View SHVC. Specifically, the framework leverages the non-HEVC option of the SHVC encoder, using a VEI (Video Enhancement Information) restored image as the base layer to provide higher-quality prediction signals and reduce encoding complexity. Experimental results using animation and live-action UHD sequences show that the proposed method achieves BD-RATE reductions of 57.78% and 45.10% compared with HEVC and SHVC codecs, respectively. [ABSTRACT FROM AUTHOR]

In this paper, we propose a definition of block tensors and the real representation of tensors. Equipped with the simplification method, i.e., the real representation along with the M-P inverse, we demonstrate the conditions that are necessary and sufficient for the system of dual split quaternion tensor equations (A ∗ N X , X ∗ S C) = (B , D) , when its solution exists. Furthermore, the general expression of the solution is also provided when the solution of the system exists, and we use a numerical example to validate it in the last section. To the best of our knowledge, this is the first time that the aforementioned tensor system has been examined on dual split quaternion algebra. Additionally, we provide its equivalent conditions when its Hermitian solution X = X ∗ and η -Hermitian solutions X = X η ∗ exist. Subsequently, we discuss two special dual split quaternion tensor equations. Last but not least, we propose an application for encrypting and decrypting two color videos, and we validate this algorithm through a specific example. [ABSTRACT FROM AUTHOR]

Electrophoretic displays (EPDs) are popular in consumer electronics, like e‐readers and tags, for their paper‐like visuals and minimal power consumption. However, EPDs struggle to rapidly switch to target grayscales due to the inherent limitations in the response speed of their display particles. Video halftoning technology achieves a smoother video grayscale display, but it also inevitably reduces the image quality and affects the viewing experience. Therefore, this paper proposes an EPD driving method based on image semantic segmentation. The method could distinguish the dynamic region and the static region accurately. For static areas, high‐definition driving waveforms were used to ensure the image display accuracy. For the dynamic area, dynamic target and dynamic text were processed by different algorithms, respectively, and the speed‐driving waveforms were used to maintain the fluency and clarity of the video display. With the method we proposed, the display quality for EPD in complex video playback was optimized significantly. Compared to DBS refresh techniques, the SSIM had increased from 4% to 82% and the PSNR had increased from 2.56 dB to 10.89 dB with our method, significantly enhancing the video playback performance of EPDs. [ABSTRACT FROM AUTHOR]

Temporal grounding involves identifying the target moment based on the provided sentence in an untrimmed video. In weakly supervised temporal grounding studies, existing temporal sentence grounding methods face challenges in (1) learning semantic alignment between the candidate window and language query and (2) identifying accurate temporal boundaries during the grounding process. In this work, we propose a reinforcement learning (RL)-based multi-policy movement framework (MMF) for weakly supervised temporal sentence grounding. We imitate the behavior of human beings when grounding specified content in a video, starting from a coarse location and then identifying fine-grained temporal boundaries. The RL-based framework initially sets a series of candidate windows and learns to adjust them step-by-step by maximizing the rewards, indicating the semantic alignment between the current window and the query. To better learn the alignment, we propose a Gaussian-based Dual-Alignment Module (GDAM) which combines the strengths of both scoring-based and reconstruction-based alignment methods, addressing the issues of negative sample bias and language bias. We also employ the multi-policy movement strategy (MMS) which grounds the temporal position in a coarse-to-fine manner. Extensive experiments demonstrate that our proposed method outperforms existing weakly supervised algorithms, achieving state-of-the-art performance on the Charades-STA and ActivityNet Captions datasets. [ABSTRACT FROM AUTHOR]

More than 80% of video coding times are consumed by motion estimation calculations, which are the most complex aspect of the process. This method eliminates temporal redundancies in a video sequence to achieve maximum compression. Numerous efforts have been made to bring calculations closer to real time, yielding fruitful results. This study proposes a fast subpixel motion estimation algorithm for video encoding with fewer search points. This method employs the capabilities of human visual systems (HVSs), physical motion characteristics of real‐world objects, and special image information from successive frames. The number of search points (NSP) using the statistical data of the movement of the blocks in the frames of video sequences is reduced to apply fewer calculations to the system while maintaining the quality of images. Therefore, it is possible to approach fast and real‐time calculations instead of time‐consuming algorithms by accurately modeling this algorithm. [ABSTRACT FROM AUTHOR]

To investigate the point where inappropriate defensive movements can no longer be inhibited and to validate suitable stimulus material for constructing a basketball-specific anticipation-response-inhibition task, two experiments were conducted. In Experiment 1, participants without basketball expertise (N = 25) watched a video of a basketball jump shot and were asked to release the space bar at the point when the ball leaves the player's fingertips (go-trials). In 25% of all trials, the video was stopped prematurely and participants should withhold their finger-lift response (stop-trials). A staircase-tracking algorithm was used to adjust the point-in-time when the jump shot was stopped in a way that participants' inhibition rate was at 50% (reflecting the so called "point-of-no-return", PNR). In Experiment 2, the stimulus material was adapted so that stop-trials simulated a pump fake. The PNR in Experiment 1 was located 187 ms and in Experiment 2 177 ms before the point of ball release. Precision performance benefit from practice across blocks and participants delayed their responses after stop-trials in a subsequent go-trial, which reflects strategic post-stop-trial adjustments. Based on the comparable results of previous studies, the given stimulus material is suitable for investigating response inhibition skills in dynamic sport-specific environments. [ABSTRACT FROM AUTHOR]

Copyright of Journal of Chongqing University of Posts & Telecommunications (Natural Science Edition) is the property of Chongqing University of Posts & Telecommunications and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)

The current paradigm of joint detection and tracking still requires a large amount of instance-level trajectory annotation, which incurs high annotation costs. Moreover, treating embedding training as a classification problem would lead to difficulties in model fitting. In this paper, we propose a new self-supervised multi-object tracking based on the real-time joint detection and embedding (JDE) framework, which we termed as self-supervised multi-object tracking (SS-MOT). In SS-MOT, the short-term temporal correlations between objects within and across adjacent video frames are both considered as self-supervised constraints, where the distances between different objects are enlarged while the distances between same object of adjacent frames are brought closer. In addition, short trajectories are formed by matching pairs of adjacent frames using a matching algorithm, and these matched pairs are treated as positive samples. The distances between positive samples are then minimized for futher the feature representation of the same object. Therefore, our method can be trained on videos without instance-level annotations. We apply our approach to state-of-the-art JDE models, such as FairMOT, Cstrack, and SiamMOT, and achieve comparable results to these supevised methods on the widely used MOT17 and MOT20 challenges. [ABSTRACT FROM AUTHOR]

Coverless steganography‏ ‏has received a lot of attention lately because it is a ‎technique that ‎completely defies ‎steganalysis detection by not modifying the ‎carriers.‎‏ ‏The majority of ‎currently used coverless steganography ‎algorithms, ‎however, use images as their carriers, and ‎there aren't many studies on coverless ‎video steganography. Actually, video is a more trustworthy and educational medium. In addition, most of covert ‎video ‎steganography techniques in ‎use today hide information in a specific video frame. These ‎techniques ‎are ‎insufficiently robust and do not take into account the distinct ‎sequential ‎characteristics of video carriers that set ‎them apart from images. This ‎research ‎proposes a coverless steganography scheme depend on the integer ‎wavelet ‎transform ‎‎(IWT) of video. At first, a new method is proposed to generate hash ‎sequences based on ‎IWT ‎coefficients. We take each frame in the video and divide ‎it into sub matrices(blocks). ‎After that, the coefficients of ‎each block are converted ‎into a bit sequence. Then, a video ‎index structure was created to expedite the ‎process of ‎finding matching blocks in the video. ‎The procedure of information ‎embedding involves segmenting the confidential ‎info‏ ‏into ‎binary segments (8 ‎bits) and choosing the blocks based on the video index structure whose ‎hash ‎sequence ‎matches the confidential information segment. Finally, all the chosen ‎blocks ‎and ancillary information are ‎transmitted to the recipient. The suggested ‎method outperforms ‎the most recent coverless steganography algorithms ‎in terms ‎of capacity, resilience, and safety, ‎according to experimental findings and analysis.‎ [ABSTRACT FROM AUTHOR]

We introduce a deep learning-driven framework for creating an adaptably applicable importance map (A2R-Map) that can be integrated with existing image and video retargeting operators. A conventional retargeting algorithm uses a heuristic approach to seek an off-the-self algorithm used into their retargeting system. The extracted importance map of the image does not match the characteristics of the input image; therefore, it affects the retargeting results and limits the performance of the retargeting method. Our designed framework attempts to minimize the artifacts/distortions caused by inappropriate energy, e.g., the shrunk phenomenon in warping-based results and carving-through-object distortion in the seam carving-based approach. Our proposed framework focuses on capturing sensitive distortion regions and activating their energy to solve this challenge. We verify the effectiveness of our proposed scheme by plugging it in three typical retargeting methods: seam carving-based, warping-based for image, and video retargeting. Extensive experiments and evaluations are conducted on two widely used databases. On the one hand, A2R-Map significantly reduces the time of importance map generation in retargeting systems to ∼ 9 times compared to the baseline saliency map. On the other hand, our A2R-Map achieves improvement over the baseline methods with an average of 11% and 9% in terms of image and video quality, respectively. The experimental results and evaluations demonstrate that our strategy for A2R-Map substantially outperforms the previous works and significantly boosts the visual quality of video/image retargeting. [ABSTRACT FROM AUTHOR]

Multi-object tracking (MOT) aims at estimating locations and identities of objects in videos. Many modern multiple-object tracking systems follow the tracking-by-detection paradigm, consisting of a detector followed by a method for associating detections into tracks. Tracking by associating detections through motion-based similarity heuristics is the basic way. Motion models aim at utilizing motion information to estimate future locations, playing an important role in enhancing the performance of association. Recently, a large-scale dataset, DanceTrack, where objects have uniform appearance and diverse motion patterns, was proposed. With existing hand-crafted motion models, it is hard to achieve decent results on DanceTrack because of the lack of prior knowledge. In this work, we present a motion-based algorithm named TLtrack, which adopts a hybrid strategy to make motion estimates based on confidence scores. For high confidence score detections, TLtrack employs transformers to predict its locations. For low confidence score detections, a simple linear model that estimates locations through trajectory historical information is used. TLtrack can not only consider the historical information of the trajectory, but also analyze the latest movements. Our experimental results on the DanceTrack dataset show that our method achieves the best performance compared with other motion models. [ABSTRACT FROM AUTHOR]

Data-driven modal analysis methods provide a powerful way to decompose data into a sum of modes. The spatiotemporal Koopman decomposition (STKD) enables the computation of modes defined by global frequencies and growth rates in various spatial dimensions and time. The method is an extension of the dynamic mode decomposition (DMD) and higher-order dynamic mode decomposition (HODMD) that represents the data as a sum of standing and traveling, possibly growing or decaying, waves. In this paper, the STKD with HODMD is applied to schlieren video highlighting second mode instability waves traveling down the length of a 3-degree half-angle cone and a 7-degree half-angle cone, both at a freestream Mach number of 6. The HODMD is able to compute dominant modes and frequencies that align with those from associated experimental measurements of unsteady pressure fluctuations, and whose mode shapes clearly show the intensifying wavepacket structure of the waves. The STKD algorithm is used to compute streamwise wavenumbers, spatial growth rates, and wave speeds. The spatial growth rates from the STKD and the magnitudes of the HODMD mode shapes are used to compute the N-factor for waves of several frequencies. Overall, the STKD with HODMD is shown to be a useful tool for extracting spatiotemporal disturbance growth from a schlieren video. [ABSTRACT FROM AUTHOR]

Classic video compression methods usually suffer from long encode time and requires large memories, making it hard to deploy on edge devices; thus, video compressive sensing, which requires less resources during encoding, is receiving more attention. We propose a robust mixed-rate ROI-aware video compressive sensing algorithm for transmission line surveillance video compression. The proposed method compresses foreground targets and background frames separately and uses reversible neural network to reconstruct original frames. The result on transmission line surveillance video data shows that the proposed compressive sensing method can achieve 26.47, 34.71 PSNR and 0.6839, 0.9320 SSIM higher than existing methods on 1.5% and 15% measurement rates, and the proposed ROI extraction net can precisely retrieve regions under high noise levels. This research not only demonstrates the potential for a more efficient video compression technique in resource-constrained environments, but also lays a foundation for future advancements in video compressive sensing techniques and their applications in various real-time surveillance systems. [ABSTRACT FROM AUTHOR]

In this paper, we develop an effective iterative algorithm to solve a generalized Sylvester tensor equation over quaternions which includes several well-studied matrix/tensor equations as special cases. We discuss the convergence of this algorithm within a finite number of iterations, assuming negligible round-off errors for any initial tensor. Moreover, we demonstrate the unique minimal Frobenius norm solution achievable by selecting specific types of initial tensors. Additionally, numerical examples are presented to illustrate the practicality and validity of our proposed algorithm. These examples include demonstrating the algorithm's effectiveness in addressing three-dimensional microscopic heat transport and color video restoration problems. [ABSTRACT FROM AUTHOR]

Action segmentation consists of temporally segmenting a video and labeling each segmented interval with a specific action label. In this work, we propose a novel action segmentation method that requires no initial video analysis and no annotated data. Our proposal involves extracting features from videos using several pre-trained deep-learning models, including spatiotemporal and self-supervised methods. Data is then transformed using a positional encoder, and finally, a clustering algorithm is applied, where each produced cluster presumably corresponds to a different single and distinguishable action. For self-supervised features, we explored DINO, and for spatiotemporal features, we investigated I3D and SlowFast methods. Moreover, two different clustering algorithms (FINCH and KMeans) were investigated, and we also explored how varying the length of video snippets that generate the feature vectors affected the quality of the segmentation. Experiments show that our method produces competitive results on the Breakfast and INRIA Instructional Videos dataset benchmarks. Our best result was produced using a composition of self-supervised features generated by DINO, FINCH clustering, and positional encoding. [ABSTRACT FROM AUTHOR]

This paper studies an LSTM-based adaptive transmission method for a 360-degree video and proposes a non-uniform encoding transmission strategy based on LSTM. Our goal is to maximize the user's video experience by dynamically dividing the 360-degree video into tiles of different numbers and sizes, and selecting different bitrates for each tile. This aims to reduce buffering events and video jitter. To determine the optimal number and size of tiles at the current moment, we constructed a dual-layer stacked LSTM network model. This model predicts, in real-time, the number, size, and bitrate of the tiles needed for the next moment of the 360-degree video based on the distance between the user's eyes and the screen. In our experiments, we used an exhaustive algorithm to calculate the optimal tile division and bitrate selection scheme for a 360-degree video under different network conditions, and used this dataset to train our prediction model. Finally, by comparing with other advanced algorithms, we demonstrated the superiority of our proposed method. [ABSTRACT FROM AUTHOR]

Existing cost allocation methods for adaptive video steganography mainly focus on specific transform coefficients, resulting in lower capacity. Moreover, distortion drift is a significant challenge for steganography in HEVC videos. Therefore, this paper proposed a cost allocation method that combined the intra-frame and inter-frame processes of HEVC video coding to achieve high-capacity, low-distortion transmission in high-performance adaptive video steganography. Firstly, the method investigated the discrete sine transform features in HEVC video coding, analyzing the error propagation patterns of these coefficients under disturbance. During the embedding processed, it conducted a detailed analysis on the intra-block distortion, interblock distortion, and inter-frame distortion caused by modifying transform coefficients. The algorithm also took into account the differentiation in inter-block distortion resulting from steganography in different blocks, leading to the classification of blocks. This paper maximized the utilization of all non-zero transform coefficients, allocating distinct distortion costs for various carrier coefficients. The covert information was then embedded into frames that minimally impacted video quality. Experimental results indicate that, compared to existing HEVC video coefficient domain steganography methods, the proposed algorithm demonstrates advantages in terms of video bitrate, video quality, and embedding capacity. [ABSTRACT FROM AUTHOR]

The paper presents a deep-learning model that may be used to calculate saliency maps for video content. Classic saliency algorithms take into account only spatial information obtained from an image to generate a saliency map showing regions of higher importance—hence, regions where people are more likely to turn their gaze and attention. Algorithms for video stimuli add temporal information about the movements of objects on a frame-to-frame basis, resulting in more complex three-dimensional architectures. The paper analyses existing models and proposes a model based on one of them. The model's performance is compared with the literature using four widely accepted measures (AUC, NSS, SIM, and CC). It is comparable and, in many cases, even better than already published models. Additionally, because of some improvements in the architecture, it is significantly faster in terms of the number of frames processed per second. [ABSTRACT FROM AUTHOR]

Copyright of Chinese Journal of Liquid Crystal & Displays is the property of Chinese Journal of Liquid Crystal & Displays and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)

To select videos to compose a change detection dataset, we can consider the videos' difficulty level. We need to use difficulty maps, which store values representing the pixels' difficulty level, to estimate these levels. The problem is that ground truth is needed to generate a difficulty map, and generating the ground truth requires manual attribution of labels to the pixels of the frames. Identifying the difficulty level of a video before producing its ground truth allows researchers to obtain the difficulty level, select the videos considering this information, and, subsequently, generate ground truths only for the videos with different difficulty levels. Datasets containing videos with different difficulty levels can evaluate an algorithm more adequately. In this research, we developed a method to generate difficulty maps of a video without using its ground truth. Our method uses the videos and the ground truths from the CDNet 2014 dataset to generate difficulty maps to train a pix2pix neural network. The results showed that the trained network could generate difficulty maps similar to those generated by the traditional approach. [ABSTRACT FROM AUTHOR]

This article is part of a recent line of research that emerged after the diagnosis and overcoming of breast cancer by one of the individuals signing this academic text. Using Social Network Analysis, it addresses the playlist of videos about breast cancer stored and offered by YouTube when a user seeks to view content about the disease. The aim is to identify the most relevant videos and the concerns of those who publish and comment on the topic. The results highlight the significance of content shared by medical professionals and health-related organizations, as well as the importance of personal stories as sources of hope and support for those affected by the illness. Regarding video comments, in addition to medical aspects and personal testimonies, the significance of faith and prayer as aids in overcoming breast cancer is also emphasized. Lastly, the authors wish to express solidarity through this work with all those individuals who battle against the disease day by day. [ABSTRACT FROM AUTHOR]

The classical moving target segmentation (MTS) algorithm in a video can segment the moving targets out by calculating frame by frame, but the algorithm encounters a real-time problem as the data increases. Recently, the benefits of quantum computing in video processing have been demonstrated, but it is still scarce for MTS. In this paper, a quantum moving target segmentation algorithm for grayscale video based on background difference method is proposed, which can simultaneously model the background of all frames and perform background difference to segment the moving targets. In addition, a feasible quantum subtractor is designed to perform the background difference operation. Then, several quantum units, including quantum cyclic shift transformation, quantum background modeling, quantum background difference, and quantum binarization, are designed in detail to establish the complete quantum circuit. For a video containing 2 m frames (every frame is a 2 n × 2 n image with q grayscale levels), the complexity of our algorithm is O (n + q) . This is an exponential speedup over the classical algorithm and also outperforms the existing quantum algorithms. Finally, the experiment on IBM Q demonstrates the feasibility of our algorithm in this noisy intermediate-scale quantum (NISQ) era. [ABSTRACT FROM AUTHOR]

Deepfakes are digital audio, video, or images manipulated using machine learning algorithms. These manipulated media files can convincingly depict individuals doing or saying things they never actually did. Deepfakes pose significant risks to our lives, including national security, financial markets, and personal privacy. The ability to create convincing deep fakes can also harm individuals' reputations and can be used to spread disinformation and fake news. As such, there is a growing need for reliable and accurate methods to detect deep fakes and prevent their harmful effects. In this paper, a hybrid convolutional neural network (CNN) and recurrent neural network (RNN) with a particle swarm optimization (PSO) algorithm is utilized to demonstrate a deep learning strategy for detecting deepfake videos. High accuracy, sensitivity, specificity, and F1 score were attained by the proposed approach when tested on two publicly available datasets: Celeb-DF and the Deepfake Detection Challenge Dataset (DFDC). Specifically, the proposed method achieved an average accuracy of 97.26% on Celeb-DF and an average accuracy of 94.2% on DFDC. The results were compared to other state-of-the-art methods and showed that the proposed method outperformed many. The proposed method can effectively detect deepfake videos, which is essential for identifying and preventing the spread of manipulated content online. [ABSTRACT FROM AUTHOR]

The visual microphone is a technique for remote sound recovery that extracts sound information from tiny pixel‐scale vibrations in a video. Despite having demonstrated success in sound recovery, the impact of various visual enhancement and color conversion algorithms applied on the video before the sound recovery process has not been explored. Thus, it is important to investigate these effects have on the recovered sound quality, as the vibrations are so small the effects play an important role. This work experimented with different color to grayscale conversions and visual enhancement algorithms on 576 videos, and found that the recovered sound quality is indeed greatly affected by the choice of algorithms. The best conversion algorithms were found to be the average of the red, green and blue color channels and the perceptual lightness in the CIELAB color space, improving the recovered sound quality by up to 23.22%. Furthermore, visual enhancement techniques such as gamma correction have been found to corrupt vibration information, leading to a 22.47% drop in recovered sound quality in one of the tested videos. Therefore, it is advisable to avoid or minimize the use of visual enhancement techniques for remote sound recovery to prevent the elimination of useful subtle vibrations. [ABSTRACT FROM AUTHOR]

In an age where preserving knowledge and information from books and documents is crucial, traditional manual scanning methods are tedious and error-prone. It involves a lot of human intervention and, as a result, sometimes results in erroneous digitization, which makes the downstream tasks, such as optical character recognition, difficult. Therefore, innovative techniques are required to be proposed that not only reduce human effort in terms of digitization but also give highly accurate results over the recently proposed state-of-the-art techniques. We proposed a novel computer vision-based algorithm that combines Gray-Level Co-occurrence Matrix (GLCM) features with Thepade's 10-ary texture features (TSBTC) for video frame classification. This hybrid approach significantly enhances frame selection accuracy, ensures high-quality digitization, and accommodates multiple languages and document types. We also proposed a dataset of 54,000 diverse images to demonstrate our algorithm's effectiveness in real- world scenarios and compare it to existing methods, making a valuable contribution to document digitization. The proposed dataset can be utilized for several document image analysis tasks. [ABSTRACT FROM AUTHOR]

Prior research has identified a variety of task-dependent networks that form through inter-regional phase-locking of oscillatory activity that are neural correlates of specific behaviors. Despite ample knowledge of task-specific functional networks, general rules governing global phase relations have not been investigated. To discover such general rules, we focused on phase modularity, measured as the degree to which global phase relations in EEG comprised distinct synchronized clusters interacting with one another at large phase lags. Synchronized clusters were detected with a standard community-detection algorithm, and the degree of phase modularity was quantified by the index q. Notably, we found that the mechanism controlling phase modularity is remarkably simple. A network comprising anterior-posterior long-distance connectivity coherently shifted phase relations from low-angles (|Δθ| < π/4) in low-modularity states (bottom 5% in q) to high-angles (|Δθ| > 3π/4) in high-modularity states (top 5% in q), accounting for fluctuations in phase modularity. This anterior-posterior network may play a fundamental functional role as (1) it controls phase modularity across a broad range of frequencies (3–50 Hz examined) in different behavioral conditions (resting with the eyes closed or watching a silent nature video) and (2) neural interactions (measured as power correlations) in beta-to-gamma bands were consistently elevated in high-modularity states. These results may motivate future investigations into the functional roles of phase modularity as well as the anterior-posterior network that controls it. [ABSTRACT FROM AUTHOR]

Object detection algorithms play an important role in detecting people in surveillance videos. In recent years, with the rapid development of deep learning, the performance of object detection has improved by leaps and bounds, and the scheme of object detection by the YOLOV7 algorithm has also been born. Traditional object detection methods often fail to achieve a balance between speed and accuracy. To address these issues, in this research, an improved YOLOv7 algorithm performance is proposed to get the best speed-to-accuracy balance compared to state-of-the-art object detection within recorded videos using an effective compression method. This method calculates the difference between frames of video, and by using the zero difference approach by removing the duplicate frames from the recorded video and choosing only the meaningful frames based on many variables, including frame size, frame details, and the distance of the frames, influence the choice of a meaningful frame, and this will reduce the size of the video by eliminating the frames comparable to those chosen. Additionally, any other datasets or pre-trained weights have not been used; YOLOv7 has been exclusively trained on the MS COCO dataset from scratch. In order to ensure the effectiveness of this approach, numerous detection systems are used in this work. Additionally, positive performance results to reduce the processing time required for object detection have been attained. [ABSTRACT FROM AUTHOR]

This paper proposes a video security transmission enhancement algorithm based on block permutation and dynamic multiplexing sequences encryption based on 4D autonomous hyperchaotic system. Firstly, we employ the block permutation encryption and diffusion confusion encryption module, which is based on dynamic multiplexing chaotic sequences, to encrypt the plaintext and obtain the ciphertext. Subsequently, the hash value of this round's ciphertext is utilized to generate the chaotic key, produced by the multiplexing sequence of this round after mathematical processing. Then, the key is used to generate the chaotic sequence to confuse the N-th of the multiplexed sequence, and the next round of multiplexed sequence is obtained. If the current round of chaotic sequence has been completely confused, the chaotic sequence is re-generated to generate a new multiplex sequence by using the key generated by the current round key and the initial key. Finally, the above steps are repeated for the encryption of each frame of the video. Compared with the traditional permutation coding algorithm, it increases the difficulty of estimation or recognition while ensuring efficiency, and effectively improves the avalanche effect of the algorithm. Through frame by frame ciphertext closed-loop feedback, it has the ability to resist known plaintext attack and selected plaintext attack. The results show that the scheme has high security and significant diffusion characteristics, and can effectively resist various common cryptographic attacks. [ABSTRACT FROM AUTHOR]

Instead of viewing TikTok as a platform, in this article we borrow Dutch film theorist Patricia Pisters's concept of neuro-images to approach TikTok as a cultural form that is deeply participatory, platform contingent, and algorithmically engraved. In the co-production between algorithms and users, TikTok becomes an enormous database and generates personalised narratives about individuals and the world onto and through its 'brain-screen' interfaces, which simulate our conscious and unconscious mind, and actualise the idea of creativity based on repetition. TikTok thus enables a quasi-automated cinema, whose non-stopping filming of everyday lives does not seek to reduce desires and tastes into a singular and coherent structure, but instead uncovers, releases and contains them in its vast database and interfaces, leading to a fluid and modulating categorisation of identities. It is within this quasi-automated, deeply participatory digital cinema that TikTok constitutes neuro-images, producing a distinctive experience of time, and unpredictable and unstable futures. [ABSTRACT FROM AUTHOR]

In the mobile network environment, the accuracy of related image matching algorithms is affected by factors such as bandwidth uncertainty and channel interference, resulting in significant limitations in image feature matching. This article designs a high-precision matching algorithm for multi-image segmentation of micro animation videos in mobile network environments. Fully denoise micro animation video images using 2D High Density Discrete Wavelet Transform (HD-DWT), and apply fixed block count segmentation to process micro animation video images; Using Harris algorithm to complete image corner detection and obtain corner features of sub images; In the K-means clustering algorithm, SIFT feature vectors are divided into clusters and paired with the nearest neighbor cluster in another sub image to form a sub image matching pair, completing block based sub image matching; Combine all sub image matching results to obtain video image matching results, and use the Improved Random Sampling Consistency (RANCAS) algorithm to remove incorrect matching during the matching process, improving matching accuracy. The experimental results show that the designed algorithm can effectively reduce image noise, improve image quality, and generate a large number of matching pairs in mobile network environments. After the application of the designed algorithm, the production effect of micro animated videos in mobile networks can be significantly improved. [ABSTRACT FROM AUTHOR]

Screen content video has been widely used in teleconferencing, cloud gaming, screen sharing, etc. In order to improve the compression efficiency of screen content video, the international standards organization developed the extended screen content coding standard (HEVC-SCC). However, high compression rates may also cause serious information loss when transmission error occurs, degrading the quality of received video. To tackle this problem, this paper proposes an adaptive intra refresh strategy for screen content video transmission. Firstly, the intra refresh optimization problem is mathematically modeled considering error propagation and error concealment. Next, the optimization problem is simplified based on the quantization mechanism, and the optimal solution of the optimization problem is obtained using the offline training method. Then, the buffer control is designed to reduce the transmission delay. Finally, experimental results show that compared with the original HEVC-SCC encoding method, the proposed algorithm could averagely improve the rate-distortion performance by 14.12%, and significantly improve the subjective quality of the received video. [ABSTRACT FROM AUTHOR]

In this paper, the comparison of the power requirements during real-time processing of video sequences in embedded systems was investigated. During the experimental tests, four modules were tested: Raspberry Pi 4B, NVIDIA Jetson Nano, NVIDIA Jetson Xavier AGX, and NVIDIA Jetson Orin AGX. The processing speed and energy consumption have been checked, depending on input frame size resolution and the particular power mode. Two vision algorithms for detecting lines located in airport areas were tested. The results show that the power modes of the NVIDIA Jetson modules have sufficient computing resources to effectively detect lines based on the camera image, such as Jetson Xavier in mode MAXN or Jetson Orin in mode MAXN, with a resolution of 1920 × 1080 pixels and a power consumption of about 19 W for 24 FPS for both algorithms tested. [ABSTRACT FROM AUTHOR]

Copyright of Journal of Computer Engineering & Applications is the property of Beijing Journal of Computer Engineering & Applications Journal Co Ltd. and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)

The new generation of 3D video is an international frontier research hotspot. However, the large amount of data and high complexity are core problems to be solved urgently in 3D video coding. The latest generation of video coding standard versatile video coding (VVC) adopts the quad-tree with nested multi-type tree (QTMT) partition structure, and the coding efficiency is much higher than other coding standards. However, the current research work undertaken for VVC is less for 3D video. In light of this context, we propose a fast coding unit (CU) decision algorithm based on the gray level co-occurrence matrix (GLCM) and Extra trees for the characteristics of the depth map in 3D video. In the first stage, we introduce an edge detection algorithm using GLCM to classify the CU in the depth map into smooth and complex edge blocks based on the extracted features. Subsequently, the extracted features from the CUs, classified as complex edge blocks in the first stage, are fed into the constructed Extra trees model to make a fast decision on the partition type of that CU and avoid calculating unnecessary rate-distortion cost. Experimental results show that the overall algorithm can effectively reduce the coding time by 36.27–51.98%, while the Bjøntegaard delta bit rate (BDBR) is only increased by 0.24% on average which is negligible, all reflecting the superior performance of our method. Moreover, our algorithm can effectively ensure video quality while saving much encoding time compared with other algorithms. [ABSTRACT FROM AUTHOR]

In the context of safety and security, the ability to track and identify faces in hazy conditions presents a significant challenge. The deleterious effects of haze on video quality, such as the diminution of detail, reduction in contrast, distortion of color, and complications in depth estimation, impede effective facial recognition. Additionally, the complexity of live video tracking is exacerbated by factors such as occlusion, positional variations, and lighting changes. Despite these challenges, video sequences offer an abundance of information, surpassing static images in terms of potential data extraction. In this study, a dual approach strategy is employed to detect and track faces in hazy conditions. The Kanade-Lucas-Tomasi (KLT) algorithm, celebrated for its adept feature tracking capabilities, is deployed to execute face tracking. The effectiveness of this algorithm lies in its ability to accurately trace points across successive image frames, a crucial aspect of reliable face tracking. Concurrently, the Viola-Jones algorithm is utilized for face detection. The algorithm harnesses Haar-like features to efficiently discern faces in real-time, effectively overcoming the challenge of identifying faces within video frames. To further enhance the quality of the video, the dark channel prior (DCP) image dehazing technique is employed. This technique improves visibility by increasing contrast and color saturation, whilst concurrently identifying and eliminating air haze from the video frames. [ABSTRACT FROM AUTHOR]

The critical challenge of video object counting is to avoid counting the same object multiple times in different frames. By comparing the appearance and motion feature information of the detection results, the authors use the multi-object tracking method to assign an independent ID number to each object. From the time the ID tag is obtained until the end of the video, each object is counted only once. However, even minor amounts of image noise can cause irreversible changes in feature information, resulting in severe tracking drifts. This paper introduces the concept of scene awareness and addresses unreasonable ID assignment caused by unreliable feature matching in the context of region division. Through the macro analysis of the scene, the authors define the region (called the transition region) where the number of objects can increase or decrease and require that all ID assignments for new objects and ID deletions for existing objects take place only in the transition region. Because the actual number of objects in the non-transition region is constant, they rematch unmatched objects with existing IDs in the region (called ID relocation) because changes in object ID are caused by feature matching failure. In this paper, the authors create algorithms for dynamically generating transition regions, detecting object increases and decreases, and relocating object IDs. Experimental results show that the method effectively improves the accuracy of video object counting. [ABSTRACT FROM AUTHOR]

Weakly supervised temporal action localization (WTAL) aims to locate the start and end boundaries of action instances and identify the corresponding actions. Although the existing methods have made great progress, there are still problems of incomplete localization and missing detection of shorter motions. To this end, this paper proposed a localization method of feature mining and region enhancement (FMRE) Firstly it calculated the similarity score between video segments through the base branch, and aggregated the context information with this score to obtain a more differentiated segment classification score, further realizing the complete positioning of the action. Then, it added a enhance branch to dynamically up-sample action proposals with a shorter duration in the initial localization along the temporal dimension, and then utilized the multi-head self-attention mechanism to explicitly model the temporal structure between action proposals, which facilitated action localization with temporal dependencies and prevented missing detection of short actions. Finally, it constructed pseudo-labels of mutual supervision between the two branches to gradually improve the quality of action proposals during the training process. The algorithm achieves mAP of 70.3% and 40.7% detection performances on the THUMOS14 and ActivityNet1.3 datasets respectively, which proves the effectiveness of the proposed algorithm. [ABSTRACT FROM AUTHOR]

As many new coding techniques and coding structures have been introduced to further improve the coding efficiency of depth maps in 3D video extensions, the coding complexity has been greatly increased. Fast algorithms are now needed to improve coding unit (CU) depth decisions as well as the coding pattern decision based on the coding. This paper presents an innovative machine learning-based approach aimed at mitigating the complexity associated with in-frame coding algorithms. We build different clustering models for different CU sizes to cluster CUs of the same size to decide their CU sizes. This is achieved by augmenting ensemble clustering through the expedited propagation of clustering similarities, considering CU with the same or similar texture complexity the same as for CU depth selection, which is informed by a comprehensive analysis of the original texture and its neighboring elements. The experimental findings demonstrate that the proposed scheme yields a substantial average reduction of 44.24% in the coding time. Remarkably, the corresponding Bjøntegaard delta bit rate (BDBR) increment observed for the synthetic view is a mere 0.26%. [ABSTRACT FROM AUTHOR]

Video Snapshot compressive imaging (SCI) is a promising technique to capture high-speed videos, which transforms the imaging speed from the detector to mask modulating and only needs a single measurement to capture multiple frames. The algorithm to reconstruct high-speed frames from the measurement plays a vital role in SCI. In this paper, we consider the promising reconstruction algorithm framework, namely plug-and-play (PnP), which is flexible to the encoding process comparing with other deep learning networks. One drawback of existing PnP algorithms is that they use a pre-trained denoising network as a plugged prior while the training data of the network might be different from the task in real applications. Towards this end, in this work, we propose the online PnP algorithm which can adaptively update the network's parameters within the PnP iteration; this makes the denoising network more applicable to the desired data in the SCI reconstruction. Furthermore, for color video imaging, RGB frames need to be recovered from Bayer pattern or named demosaicing in the camera pipeline. To address this challenge, we design a two-stage reconstruction framework to optimize these two coupled ill-posed problems and introduce a deep demosaicing prior specifically for video demosaicing in SCI. Extensive results on both simulation and real datasets verify the superiority of our adaptive deep PnP algorithm. Code is available at https://github.com/xyvirtualgroup/AdaptivePnP_SCI. [ABSTRACT FROM AUTHOR]

Depending on the context of interest, an anomaly is defined differently. In the case when a video event isn’t expected to take place in the video, it is seen as anomaly. It can be difficult to describe uncommon events in complicated scenes, but this problem is frequently resolved by using high-dimensional features as well as descriptors. There is a difficulty in creating reliable model to be trained with these descriptors because it needs a huge number of training samples and is computationally complex. Spatiotemporal changes or trajectories are typically represented by features that are extracted. The presented work presents numerous investigations to address the issue of abnormal video detection from crowded video and its methodology. Through the use of low-level features, like global features, local features, and feature features. For the most accurate detection and identification of anomalous behavior in videos, and attempting to compare the various techniques, this work uses a more crowded and difficult dataset and require light weight for diagnosing anomalies in objects through recording and tracking movements as well as extracting features; thus, these features should be strong and differentiate objects. After reviewing previous works, this work noticed that there is more need for accuracy in video modeling and decreased time, and since attempted to work on real-time and outdoor scenes. [ABSTRACT FROM AUTHOR]

Different techniques are being applied for automated vehicle counting from video footage, which is a significant subject of interest to many researchers. In this context, the You Only Look Once (YOLO) object detection model, which has been developed recently, has emerged as a promising tool. In terms of accuracy and flexible interval counting, the adequacy of existing research on employing the model for vehicle counting from video footage is unlikely sufficient. The present study endeavors to develop computer algorithms for automated traffic counting from pre-recorded videos using the YOLO model with flexible interval counting. The study involves the development of algorithms aimed at detecting, tracking, and counting vehicles from pre-recorded videos. The YOLO model was applied in TensorFlow API with the assistance of OpenCV. The developed algorithms implement the YOLO model for counting vehicles in two-way directions in an efficient way. The accuracy of the automated counting was evaluated compared to the manual counts, and was found to be about 90 percent. The accuracy comparison also shows that the error of automated counting consistently occurs due to undercounting from unsuitable videos. In addition, a benefit–cost (B/C) analysis shows that implementing the automated counting method returns 1.76 times the investment. [ABSTRACT FROM AUTHOR]

The majority of human activity recognition (HAR) algorithms have been developed for RGB cameras. However, RGB cameras have limitations including illumination sensitivity, motion blur, and high redundancy between frames. Neuromorphic event cameras are an emerging alternative that output asynchronous events capturing brightness changes, enabling advantages like high dynamic range, low latency, and data efficiency. Although event cameras show promise for HAR, there is a lack of large-scale eventbased HAR datasets recorded in real-world conditions. In this paper, we present HARDVS, a novel neuromorphic dataset for HAR containing over 100,000 video clips of 300 everyday human activities recorded with a DAVIS346 event camera. The videos reflect challenges like varying viewpoints, illuminations, speeds, backgrounds, and occlusion. We benchmark popular HAR algorithms on HARDVS and propose a novel spatial-temporal feature learning framework (ESTF) that transforms the event stream into spatial-temporal tokens and fuses the representations using transformers. ESTF achieves state-of-the-art results on HARDVS and other neuromorphic datasets. The large-scale, realistic HARDVS dataset captured with commodity event cameras, comprehensive benchmarks, and effective ESTF framework presented will facilitate future event-based HAR research. [ABSTRACT FROM AUTHOR]

Traditional content-based video retrieval algorithms usually only used the low-level features of video images, and the content description is not enough, which leads to the unsatisfactory Retrieval results. This paper studied how to combine the low-level features and semantic features of video, improved the existing index structure, and designed an efficient sports video retrieval algorithm. The experimental results show that the proposed algorithm can meet the real-time requirements and improve the accuracy and recall rate compared with the existing methods. In addition, the proposed sports video index combined with semantic features is better than the existing index in both query time and query results. [ABSTRACT FROM AUTHOR]

As a cutting-edge research direction in the field of radar imaging, video SAR has the capability of high-resolution and persistent imaging at any time and under any weather. Video SAR requires high computational efficiency of the imaging algorithm, and PFA has become the preferred imaging algorithm because of its applicability to the spotlight mode and relatively high computational efficiency. However, traditional PFA also has problems, such as low efficiency and limited scene size. To address the above problems, a generalized persistent polar format algorithm, called GPPFA, is proposed for airborne video SAR imaging that is applicable to the persistent imaging requirements of airborne video SAR under multitasking conditions. Firstly, the wavenumber domain resampling characteristics of video SAR PFA are analyzed, and a generalized resampling method is proposed to obtain higher efficiency. Secondly, for the problem of scene size limitation caused by wavefront curvature error, an efficient compensation method applicable to different scene sizes is proposed. GPPFA is capable of video SAR imaging at different wavebands, different slant ranges, and arbitrary scene sizes. Point target and extended target experiments verify the effectiveness and efficiency of the proposed method. [ABSTRACT FROM AUTHOR]

On YouTube, self-styled algorithmic experts claim to know how algorithms "actually work." However, their knowledge is largely speculative. Developing recent work that pays attention to algorithmic expertise, I argue that algorithmic lore videos are "market devices" which are economically productive for the platform in four ways: they (1) legitimize platform narratives of algorithmic objectivity, (2) teach creators how to calculate the value of content and format it according to platform metrics, (3) encourage creators to build and govern audiences, and (4) justify continued content production even when it does not pay off for creators. Thus, despite claiming to describe "how the algorithm actually works," algorithmic lore videos are performative; by teaching creators how to understand and act in the platform economy they do important work to bring the platform's "visibility markets"—its labor market of content creators engaging in content production and its goods market of content to be watched—into being. [ABSTRACT FROM AUTHOR]