The exponential growth of video content presents substantial challenges in summarizing and retrieving relevant information, particularly in multi-video scenarios involving heterogeneous sources. This paper presents an unsupervised, graph-based centrality framework for multi-video summarization. Segment representations are extracted using 3D Convolutional Neural Networks (3DCNNs) to capture both spatial and temporal features. We introduce three novel ranking algorithms — Weighted Degree Centrality (WDC), V-Rank, and VL-Rank — extending classical methods such as Degree Centrality, PageRank, and LexRank. These algorithms incorporate visual saliency, motion, and semantic similarity to ensure relevance, diversity, and structural representativeness. The framework comprises four stages: segmentation, graph construction, ranking, and selection. We provide a detailed computational analysis, including time complexity and convergence behavior. VL-Rank achieves significantly faster convergence than PageRank through a normalized propagation scheme, while WDC offers a highly efficient, non-iterative alternative. Evaluations on the Tour20 dataset demonstrate that the proposed methods outperform state-of-the-art approaches, with WDC achieving a mean F1 score of 0.741 compared to 0.680 for the Multi-Stream baseline. The framework is both effective and scalable, making it suitable for large-scale or real-time applications.

The proliferation of low-cost DRAM chipsets now begins to allow for the consideration of substantially-increased decoded picture buffers in advanced video coding standards such as HEVC, VVC, and Google VP9. At the same time, the increasing demand for rapid scene changes and multiple scene repetitions in entertainment or broadcast content indicates that extending the frame referencing interval to tens of minutes or even the entire video sequence may offer coding gains, as long as one is able to identify frame similarity in a computationally- and memory-efficient manner. Motivated by these observations, we propose a “stitching” method that defines a reference buffer and a reference frame selection algorithm. Our proposal extends the referencing interval of inter-frame video coding to the entire length of video sequences. Our reference frame selection algorithm uses well-established feature descriptor methods that describe frame structural elements in a compact and semantically-rich manner. We propose to combine such compact descriptors with a similarity scoring mechanism in order to select the frames to be “stitched” to reference picture buffers of advanced inter-frame encoders like HEVC, VVC, and VP9 without breaking standard compliance. Our evaluation on synthetic and real-world video sequences with the HEVC and VVC reference encoders shows that our method offers significant rate gains, with complexity and memory requirements that remain manageable for practical encoders and decoders.

Advanced video classification systems decode video frames to derive texture and motion representations for ingestion and analysis by spatio-temporal deep convolutional neural networks (CNNs). However, when considering visual Internet-of-Things applications, surveillance systems, and semantic crawlers of large video repositories, the video capture and the CNN-based semantic analysis parts do not tend to be co-located. This necessitates the transport of compressed video over networks and incurs significant overhead in bandwidth and energy consumption, thereby significantly undermining the deployment potential of such systems. In this paper, we investigate the trade-off between the encoding bitrate and the achievable accuracy of CNN-based video classification models that directly ingest AVC/H.264 and HEVC encoded videos. Instead of retaining entire compressed video bitstreams and applying complex optical flow calculations prior to CNN processing, we only retain motion vector and select texture information at significantly reduced bitrates and apply no additional processing prior to CNN ingestion. Based on three CNN architectures and two action recognition datasets, we achieve 11%-94% savings in bitrate with marginal effect on classification accuracy. A model-based selection between multiple CNNs increases these savings further to the point where, if up to 7% loss of accuracy can be tolerated, video classification can take place with as little as 3 kb/s for the transport of the required compressed video information to the system implementing the CNN models.

Advanced video classification systems decode video frames to derive the necessary texture and motion representations for ingestion and analysis by spatio-temporal deep convolutional neural networks (CNNs). However, when considering visual Internet -of- Things applications, surveillance systems and semantic crawlers of large video repositories, the compressed video content and the CNN-based semantic analysis parts do not tend to be co-located. This necessitates the transport of compressed video over networks and incurs significant overhead in bandwidth and energy consumption, thereby significantly undermining the deployment potential of such systems. In this paper, we investigate the trade-off between the encoding bitrate and the achievable accuracy of CNN-based video classification that ingests AVC/H.264 encoded videos. Instead of entire compressed video bitstreams, we only retain motion vector and selected texture information at significantly reduced bitrates. Based on two CNN architectures and two action recognition datasets, we achieve 38%-59% saving in bitrate with marginal impact in classification accuracy. A simple rate-based selection between the two CNNs shows that even further bitrate savings are possible with graceful degradation in accuracy. This may allow for rate/accuracy-optimized CNN-based video classification over networks.

In this paper, we propose an optimal strategy for the transmission of scalable video over packet-based multiple-input multiple-output (MIMO) systems. The scalable extension of H.264/AVC that provides a combined temporal, quality and spatial scalability is used. For given channel conditions, we develop a method for the estimation of the distortion of the received video and propose different error concealment schemes. We show the accuracy of our distortion estimation algorithm in comparison with simulated wireless video transmission with packet errors. In the proposed MIMO system, we employ orthogonal space-time block codes (O-STBC) that guarantee independent transmission of different symbols within the block code. In the proposed constrained bandwidth allocation framework, we use the estimated end-to-end decoder distortion to optimally select the application layer parameters, i.e., quantization parameter (QP) and group of pictures (GOP) size, and physical layer parameters, i.e., rate-compatible turbo (RCPT) code rate and symbol constellation. Results show the substantial performance gain by using different symbol constellations across the scalable layers as compared to a fixed constellation.

We propose a low-delay low-complexity end-to-end video transmission system that integrates the latest scalable H.264 codec and the full-rate full-diversity quasi-orthogonal space-time block codes. At the encoder of the scalable H.264 codec, we have developed a low-delay low-complexity method for the estimation of the video distortion at the receiver for given channel conditions. The distortion estimation algorithm is validated using experimental results. In this work, we show clearly the advantage of using the quasi-orthogonal space-time block codes (STBC), which provide a higher rate compared to the orthogonal space-time block codes in a wireless video transmission system. This is in spite of the use of a non-optimal signal constellation structure. The performance results are obtained after addressing the bandwidth allocation problem that employs the scalable decoder distortion estimation algorithm for the optimal selection of the application layer and physical layer parameters.

We propose an efficient strategy for the transmission of scalable video over multiple-input multiple-output (MIMO) wireless systems. In this paper, we use the latest scalable H.264 codec (SVC), which provides combined temporal, quality and spatial scalability. At the transmitter, we estimate the decoded video distortion for given channel conditions taking into account the effects of quantization, packet loss and error concealment. The proposed scalable decoder distortion algorithm offers low delay and low complexity. The performance of this method is validated using experimental results. In our proposed system, we use a MIMO system with orthogonal space-time block codes (O-STBC) that provides spatial diversity and guarantees independent transmission of different symbols within the block code. The bandwidth constrained allocation problem considered here is simplified and solved for one O-STBC symbol at a time. Furthermore, we take the advantage of the hierarchical structure of SVC to attain the optimal solution for each group of pictures (GOP) of the video sequence. We incorporate the estimated decoder distortion to optimally select the application layer parameter, i.e., quantization parameter (QP), and physical layer parameters, i.e., channel coding rate and modulation type for wireless video transmission.

We propose an optimal strategy for the transmission of scalable video over multiple-input multiple-output (MIMO) wireless systems. In this paper, we use the latest scalable H.264 codec which provides combined temporal, quality and spatial scalability. At the transmitter, we have developed a method for the estimation of the video distortion at the receiver for given channel conditions. The accuracy of this method is validated using experimental results. In our proposed system, we use a MIMO system with orthogonal space-time block codes (O-STBC) that provides spatial diversity and guarantees independent transmission of different symbols within the block code. Rate-compatible punctured convolutional (RCPC) codes are used for unequal error protection of the scalable layers. In the constrained bandwidth allocation framework considered here, we use the estimated decoder distortion to optimally select the application layer parameter, i.e. quantization parameter (QP), and physical layer parameters, i.e. RCPC and modulation.