Multiview video summarization using video partitioning and clustering

Multiview video summarization plays a crucial role in abstracting essential information form multiple videos of the same location and time. In this paper, we propose a new approach for the multiview summarization. The proposed approach uses the BIRCH clustering algorithm for the first time on the initial set of frames to get rid of the static and redundant. The work presents a new approach for shot boundary detection using frame similarity measures Jaccard and Dice. The algorithm performs effectively synchronized merging of keyframes from all camera-views to obtain the final summary. Extensive experimentation conducted on various datasets suggests that the proposed approach significantly outperforms most of the existing video summarization approaches. To state a few, a 1.5% improvement on video length reduction, 24.28% improvement in compression ratio, and 6.4% improvement in quality assessment ratio is observed on the lobby dataset.     

A single-board video codec using video image signal processors

This paper discusses a newly developed single-board video codec using Video Image Signal Processors (VISPs). The codec has both a CCITT H.261 mode and a proprietary mode. Two VISPs, one for encoding and one for decoding, are used. The board size is 210 by 295 mm, the maximum frame rate is about 7.5 f/s, and the picture size is 180 by 144 pels.     

Packet video

The 1990s will see a large increase in the use of video for many purposes, made possible through the use of new optical storage and transmission methods, as well as radio and satellite communication. HDTV, new broadcast services and 64 kbit/s ISDN video are all at an advanced stage of development. Digital methods of transmission and storage feature prominently in future scenarios of video usage. To avoid long delays in transmission, packet video techniques for storage of high-quality material on optical discs are being considered. Coded video signals can be transmitted in packet form at variable bit rate. The author discusses the principles of packet video. Packet loss, queueing delay and layered coding are discussed. The future developments are also discussed     

Constant quality video coding using video content analysis

In the literature, several rate control techniques have been proposed to aim at the optimal quality of digitally encoded video under given bit budget, channel rate and buffer size constraints. Typically, these approaches are group-of-picture (GOP) based. For longer, heterogeneous sequences, they become unacceptably complex or struggle with model mismatches. In this paper, an off-line segment-based rate control approach is proposed for controlling the distortion variation across successive shots of a video sequence when encoding with single-layer (MPEG-4 baseline, MPEG-4 AVC) and scalable (wavelet) video codecs. Consistent quality is achieved by optimally distributing the available bits among the different segments, based on efficient rate-distortion (R-D) modelling of each segment. The individual segments are defined based on shot segmentation and activity analysis techniques. The algorithm is formulated for three different distribution models: download, progressive download and streaming. The results indicate that the proposed technique improves the quality consistency significantly, while the processing overhead compared to classical two-pass variable bit-rate (VBR) encoding is limited.     

基于视频解码器的图像卡设计

介绍了BT829B的主要特性，详细叙述了快速三维扫描系统中以BT829B为核心的图像采集系统的结构。     

Attention-based video streaming

This paper considers the problem of video streaming in low bandwidth networks and presents a complete framework that is inspired by the fovea–periphery distinction of biological vision systems. First, an application specific attention function that serves to find the important small regions in a given frame is constructed a priori using a back-propagation neural network that is optimized combinatorially. Given a specific application, the respective attention function partitions each frame into foveal and periphery regions and then a spatial–temporal pre-processing algorithm encodes the foveal regions with high spatial resolution while the periphery regions are encoded with lower spatial and temporal resolution. Finally, the pre-processed video sequence is streamed using a standard streaming server. As an application, we consider the transmission of human face videos. Our experimental results indicate that even with limited amount of training, the constructed attention function is able to determine the foveal regions which have improved transmission quality while the peripheral regions have an acceptable degradation.     

Blind video quality assessment based on multilevel video perception

Blind video quality assessment (VQA) metrics predict the quality of videos without the presence of reference videos. This paper proposes a new blind VQA model based on multilevel video perception, abbreviated as MVP. The model fuses three levels of video features occurring in natural video scenes to predict video quality: natural video statistics (NVS) features, global motion features and motion temporal correlation features. They represent video scene characteristics, video motion types, and video temporal correlation variations. In the process of motion feature extraction, motion compensation filtering video enhancement is adopted to highlight the motion characteristics of videos so as to improve the perceptual correlations of the video features. The experimental results on the LIVE and CSIQ video databases show that the predicted video scores of the new model are highly correlated with human perception and have low root mean square errors. MVP obviously outperforms state-of-art blind VQA metrics, and particularly demonstrates competitive performance even compared against top-performing full reference VQA metrics.     

A framework for end-to-end video quality prediction of MPEG video

This paper proposes, describes and evaluates a novel framework for video quality prediction of MPEG-based video services, considering the perceptual degradation that is introduced by the encoding process and the provision of the encoded signal over an error-prone wireless or wire-line network. The concept of video quality prediction is considered in this work, according to which the encoding parameters of the video service and the network QoS conditions are used for performing an estimation/prediction of the video quality level at the user side, without further processing of the actual encoded and transmitted video content. The proposed prediction framework consists of two discrete models: (i) a model for predicting the video quality of an encoded signal at a pre-encoding stage by correlating the spatiotemporal content dynamics to the bit rate that satisfies a specific level of user satisfaction; and (ii) a model that predicts primarily the undecodable frames (and subsequently the perceived quality degradation caused by them) based on the monitored averaged packet loss ratio of the network. The proposed framework is experimentally tested and validated with video signals encoded according to MPEG-4 standard.     

Broadband video switching

A broadband 64×16 space-switching approach and its applicability to large-scale broadband switching systems are described. The design uses a technique that prevents the parasitic capacitances from reducing the switching speed. The switching system was implemented in 3-μm CMOS VLSI and operated in excess of 150 Mb/s. Computer simulation indicates a 1-Gb/s potential with a 1-μm CMOS implementation     

针对网络视频应用的视频转码技术综述

对视频转码的技术基础,尤其是3种典型的体系结构进行了简述和比较,然后讨论了利用从输入视频流中易于得到的信息,减少计算复杂度和改善视觉质量的方法。同时,对近年来出现的一些热点技术,如信息(包括水印和具有差错抑制特征的数据)嵌入的视频转码、H.264码流转码等进行了重点讨论。对视频转码技术的未来发展也进行了展望。     

Robust video multicast with joint network coding and video interleaving

In this work, we propose a cross-layer solution to robust video multicast in erasure networks based on random linear network coding (RLNC) in the network layer and video interleaving (VI) in the application layer, and call it the joint RLNC-VI scheme. In the RLNC implementation, we partition one video coding unit (VCU) into several priority levels using scalable properties of H.264/SVC video. Packets from the same priority level of several VCUs form one RLNC generation, and unequal protection is applied to different generations. RLNC provides redundancy for video packets in the network layer and has proved to be useful in a multicast environment. Then, we propose a new packet-level interleaving scheme, called the RLNC-facilitated interleaving scheme, where each received packet corresponds to a new constraint on source packets. As a result, it can facilitate the RLNC decoding at the destination node. Furthermore, we study the problem of optimal interleaving design, which selects the optimal interleaving degree and the optimal redundancy of each generation. The tradeoff between delay and received video quality due to the choice of different VCUs is also examined. It is shown by simulation results that the proposed RLNC-VI scheme outperforms the pure RLNC method for robust video multicast in erasure networks. This can be explained by two reasons. First, the VI scheme distributes the impact of the loss (or erasure) of one VCU into partial data loss over multiple neighboring VCUs. Second, the original video content can be easily recovered with spatial/temporal error concealment (EC) in the joint RLNC-VI scheme.     

Social video caching

We investigate online browsing of interrelated content, represented as a catalog of items of interest featuring graph dependencies. The content is served to clients via a system of decentralized proxy caches connected to cloud servers. A client selects the next item to browse from the list of recommended items, displayed on the currently browsed item׳s catalog page. A cache has a limited size to have every item selected by its browsing clients available for local access. Thus, the system pays a penalty, whenever a client selects an item that cannot be served directly from the proxy. Conversely, the system gains a reward, if a client selects an immediately available item. We aim to select the items to cache that maximize the profit earned by the system, for the given cache capacity. We design two linear-time optimization techniques for finding the desired items to cache. We enhance the operation of the system via two additional strategies. The first one dynamically tracks the items׳ selection probabilities for a client, as a function of its prior catalog access pattern and those of its community peers. The second one constructs dynamic overlays, on behalf of the clients, that are used to share the selected items directly among them. This augments the system׳s serving capacity and enhances the clients׳ browsing experience. We study the performance of the optimization techniques via numerical experiments. They exhibit efficiency gains over reference methods, by exploiting the content dependencies and correlated community-driven access patterns of the clients. We also report proxy bandwidth savings achieved by our overlay strategy over state-of-the-art methods, on content access patterns of clients with Facebook or Twitter ties.     

Digital video

Issues addressed at three Digital Systems Information Exchange meetings, held in November 1990 and March and September 1991, are summarized. The meetings explored digital developments, looked for areas of actual or potential commonality, and examined possible common frameworks for manipulating digital images. Progress in the area of digital video standards is discussed. The three main standards concern still-picture compression, video teleconferencing, and full-motion compression on digital storage media, Products being developed using standard digital video formats are surveyed. They include multimedia products, CD players, HDTV, and receivers     

No-reference pixel based video quality assessment for HEVC decoded video

This paper proposes a No-Reference (NR) Video Quality Assessment (VQA) method for videos subject to the distortion given by the High Efficiency Video Coding (HEVC) scheme. The assessment is performed without access to the bitstream. The proposed analysis is based on the transform coefficients estimated from the decoded video pixels, which is used to estimate the level of quantization. The information from this analysis is exploited to assess the video quality. HEVC transform coefficients are modeled with a joint-Cauchy probability density function in the proposed method. To generate VQA features the quantization step used in the Intra coding is estimated. We map the obtained HEVC features using an Elastic Net to predict subjective video quality scores, Mean Opinion Scores (MOS). The performance is verified on a dataset consisting of HEVC coded 4 K UHD (resolution equal to 3840 × 2160) video sequences at different bitrates and spanning a wide range of content. The results show that the quality scores computed by the proposed method are highly correlated with the mean subjective assessments.     

Digital video broadcasting

A new kind of “wireless video” is currently entering consumers' homes-digital television. The term digital video broadcasting (DVB) is used as a synonym for digital television in many countries of the world. Whereas one may tend to think that digital television means just a new, digital, form of signal representation not necessarily affecting the information content of what one has always called TV, the truth is that digital television becomes multiple-channel data broadcasting. This article reviews some of the results of the work in the worldwide DVB Project and explains some of the fundamental concepts behind the work of this group. It then concentrates on the terrestrial transmission system (DVB-T) as one example of the many transmission technologies DVB has developed over the last few years. The COFDM modulation scheme which is a key ingredient of DVB-T is described in some detail. The performance of the system is presented. The use of DVB for data broadcasting and the accompanying return channel technologies are explained     

Digital video recording

Through mutual technology transfer between consumer and professional video recorders, the last 20 years has witnessed a rapid evolution from analog to digital recording. Each new digital videotape recorder (VTR) uses different channel coding. This implies that improving the recording density involves the development of new channel-coding schemes together with evolutionary magnetic tapes and heads. As a result, professional digital VTR's offer the best features for video recording. This paper reviews state-of-the-art magnetic recording devices, signal processing techniques for digital recording. It also describes the specifications for home-use digital recorders for current television systems, and discusses disk recording technology in the future     

Inertia-based video cut detection and its integration with video coder

An encoder-dependent video cut detection algorithm is proposed. Based on the inertia property of natural videos, the proposed algorithm detects video cuts in the video-coding loop by making use of the intermediate results of video compression. Experiments show that the detection performance of the proposed algorithm is quite good as compared to previous works, and when it is integrated with the coder, not only is the content accessibility of the output code stream improved but also the compression ratio is improved. The authors regard this inertia-based algorithm as a step towards the integration of video compression and content-based video retrieval.     

基于视频修复的运动目标删除篡改行为的检测算法

将运动对象从视频中删除是视频篡改的一种常见 形式,针对删除视频运动对象这一篡改操作, 提出了基于视频修复痕迹的检测方法。运动对象删除后需采用数字视频修复技术填补由于移 除操作产生的 黑洞,使得篡改后的视频遗留有修复痕迹;通过深入分析篡改视频中遗留的修复痕迹,对篡 改后未压缩视 频采用对称帧差法检测运动对象删除区域;对压缩后的篡改视频从运动光流场的角度,由视 频帧光流方向 的不一致性进行检测。实验结果表明,本文方法不依赖于原始视频,计算复杂度低,能够有 效检测运动对象删除操作,并在空时域上对篡改区域进行定位。     

Surveillance video synopsis in the compressed domain for fast video browsing

The traditional pixel-domain based video analysis methods have taken dominated places for long. However, due to the rapidly increasing volume and resolution of surveillance video, the desirable fast and scalable browsing encounters significant challenges in terms of efficiency and flexibility. Under this circumstance, operating surveillance video in compressed domain has aroused great concern in academy and industry. In order to perform the intelligent video analysis task on the premise of preserving accuracy and controlling complexity, this paper presents a compressed-domain approach for massive surveillance video synopsis generation, labeling and browsing. The main work and achievements include: (1) a compressed-domain scheme is established to condense the compressed surveillance video and record the synopsis results; (2) a background modeling method via the Motion Vector based Local Binary Pattern (MVLBP) is introduced to extract moving objects in an efficient way; (3) an object flags based synopsis labeling method is proposed to represent the object regions as well as their display modes in a flexible way. Experimental results show that the video analysis system based on this framework can provide not only efficient synopsis generation but also flexible scalable or playback browsing.