分布式压缩感知视频编码中CS帧的二次修正准则研究

分布式视频编码是新的视频编码体系,与传统的视频编码体系相比,具有编码端相对简单、解码端相对复杂的特点。此外,压缩感知突破了奈奎斯特采样定理,降低了信号的采样率。将压缩感知理论应与分布式视频编码相结合,使编码端复杂度降低。在一些分布式压缩视频编码研究中,CS帧是由边信息和发送端传送的信息联合重建的,由于不同CS帧的边信息的预测准确度不同,导致不同CS帧恢复质量相差较大。为了解决这个问题,本文对CS帧的二次修正准则的进行研究,首先从理论上推导出方差作为修正准则的可行性,并在实验中加以验证。实验可知,本文提出的方法在一定程度上改善了这些帧的重建质量。      

一种用于帧内篡改检测的压缩感知视频水印实现算法

 针对视频水印在帧内篡改检测方面定位精度的不足,通过压缩感知对MPEG-4(Moving Picture Experts Group-4)视频内容的特征表示,提出一种新的视频水印生成方法及其帧内篡改检测算法.该算法由压缩感知DCT(Discrete Cosine Transform)测量矩阵对I-VOP(Intra-Video Object Plane)图像提取U、V特征参数,生成基于内容的压缩感知视频水印数据并嵌入到图像Y分量的DCT中高频系数中实现帧内篡改检测.实验结果表明,与Hash视频水印算法比较,压缩感知视频水印数据具有更好的恢复能力,且水印算法对视频帧内篡改定位精度更高.     

基于压缩感知的自适应帧图像分组视频编解码器研究

目前基于压缩感知的视频编解码器在对视频帧图像进行处理时大多采用固定分组的形式,视频中出现场景快速变化时,其重构视频帧的效果较差。提出一种基于压缩感知的自适应帧图像分组视频编解码器,无论场景变化快慢,均能获到较好的视频帧重构效果。实验结果表明,视频场景快速变换时该编解码器的视频帧重构效果较好。     

鲁棒性压缩感知重构技术及其在智能视频监控中的应用研究

在视频监控中需要进行鲁棒性压缩感知重构,降低视频丢包和时延等因素的影响,提出一种基于隐马尔科夫模型的鲁棒性压缩感知重构技术并应用在智能视频监控中。首先利用隐马尔科夫模型对智能视频监控系统中的视频帧序列进行频域特征点奇偶分裂处理;然后计算视频监控编码的标量量化码率分配系数,建立视频压缩感知重构的时空方向树,采用隐马尔科夫模型进行解码重建,实现视频帧的鲁棒性压缩感知重构;最后进行仿真测试。结果表明,采用该方法进行监控视频的压缩感知重构,能有效降低丢包率和传输时延,在智能视频监控中具有较好的应用价值。     

基于自适应卡尔曼的时域增强算法研究

在图像序列进行压缩感知重构的过程中,基于运动补偿的分块压缩感知重建算法利用了帧间残差图像的稀疏特性,有效提高了重构视频的质量。但该算法仅在空域对图像进行了维纳滤波,帧间存在抖动现象,视频主观质量较差。文章将自适应卡尔曼滤波算法应用到分块压缩感知重建算法的重建过程中,可以有效地去除视频帧间的噪声,使得图像的主观质量得到了改善。     

使用能量匹配的监控视频自适应速率压缩感知

获取信号稀疏度对压缩感知(CS)性能的提升有重大意义,但在采样端不进行完整信号数字化采集和存储的情况下,对信号稀疏度进行估计比较困难。现有方法在稀疏度估计性能和计算复杂度方面难以取得较好的平衡。针对采样端对信号特性未知的监控视频应用,该文提出一种新的使用能量匹配的自适应速率压缩感知方法(ARCS-EM),通过观测一个恒定低速率的压缩感知观测结果来对当前帧实际稀疏度进行估计,然后根据估计结果决定当前帧应执行的压缩感知测量数,再进行补充测量得到当前帧的优化压缩感知采样结果。实验结果表明,该方法可以较好地适应视频中前景稀疏度的变化,为每帧图像分配适当的压缩感知测量速率,在不显著提高采样端计算复杂度的前提下,有效提高重建视频的质量。     

帧间自适应压缩感知算法在视频编码中的应用

阐述了压缩感知的理论框架,分析了视频信号帧间相关性特点,提出了一种帧间自适应压缩感知的视频编码算法。本方法中,利用视频差值信号的特点建立自适应感知模型,自适应的选择稀疏域和重构域对信号进行压缩感知恢复,在空域稀疏度较强的情况下选择空域作为稀疏域和重构域,在空域稀疏度较差的情况下选择小波域作为稀疏域和重构域。用测试视频进行了仿真分析,结果表明该算法能够取得较好的效果。     

压缩感知的等效二维稀疏变换

目前的压缩感知研究尚不能真正实现基于二维稀疏变换的影像采集和重构。通过对二维压缩感知和稀疏变换的理论分析和数学推导,将基于一维稀疏变换的二维压缩感知模型等价转换成适用于二维稀疏变换的二维压缩感知模型。从而在测量过程不变的前提下,基于一维线阵推扫数据采集方式实现了基于二维稀疏变换的压缩感知影像采集和重构。实验验证了等效二维稀疏变换的正确性。     

基于张量字典学习的压缩感知视频重构

传统的基于矩阵形式的视频图像重构算法,由于其二维表达矩阵的局限性,在重构过程中降低了相邻帧图像之间的关联性以及图像的重构质量。为了克服该问题,提出一种基于张量字典学习的压缩感知视频重构算法。把视频图像的二维空间特性和一维时间特性映射到三阶张量上,保持了图像的时间特性,增强了图像前后帧之间的相关性。同时在重构视频图像块的过程中,相对于二维矩阵字典,原子的稀疏表达有着更高的自由度,进而提高了重构质量。对张量的计算在傅里叶域中进行,减少了算术运算的次数,缩短了重构时间。通过实验数据以及视觉直观证明,提出的算法重构图像的峰值信噪比较传统方法提高了2～4 dB。     

基于视频自然统计特性的无参考移动终端视频质量评价

针对无线网络中压缩编码及无线丢包等因素对移动终端视频的降质影响,在分析视频相邻帧差信号空-时感知统计特性的基础上,该文提出一种基于视频自然统计特性的无参考移动终端视频质量评价(NMVQA)算法。进行视频帧差空-时自然统计规律分析,确定移动终端视频失真类型对视频相邻帧差系数统计特性的影响;计算水平、垂直、主对角线和副对角线方向的帧差相邻系数乘积分布参数的时域统计特性;以多尺度帧差相邻系数的时域统计特性相关程度来衡量移动终端视频失真程度。在LIVE移动视频数据库上的实验结果表明,该文所提算法的结果与主观评价具有较好的一致性,能够准确反映人类对视频失真程度的视觉感知效果,可为实时在线调节信源码率和无线信道参数提供参考依据。     

Nonlocally Adaptive Pattern Classification Based Compressed Sensing for Video Recovery

Traditional video coding algorithms usually undergo several coding steps for each frame before transmission, which reduces the efficiency at the encoder. The compressed sensing (CS) as an innovative method in signal processing can make the encoder much easier than ever before, with which each frame only needs to multiply a projection matrix at the encoder, if the frame is sparse in a transform domain. Frames in a video usually exhibit sparsity in different parts on different bases; however, existing compressed sensing reconstruction methods usually recover a frame in a fixed set of bases for the entirety of the frame. Therefore, the frames cannot be recovered faithfully by the conventional CS reconstruction methods from a small number of measurements. In this paper, in order to rectify the flaw, we construct an initial estimation frame by motion estimation from neighboring frames and through the observation of the current frame. Then, nonlocally adaptive sparse signal presentation facilitation by a 2D piecewise autoregressive (AR) model is integrated into the reconstruction. The piecewise AR model is generated from the pattern classification of subimages of the initial estimation frame and its neighboring frames. An iterative procedure is proposed to recover a new estimated frame and its AR model alternatively, until the termination threshold is satisfied. The experimental results demonstrating the capabilities of the proposed method are presented.     

帧测量率自适应分配的分布式视频压缩感知

传统分布式视频压缩感知通常对所有的非关键帧采用相同的测量率进行测量,这种测量方式并未考虑到不同帧之间的相关性具有差异性,造成帧组的重构质量不高。针对以上问题,本文首先建立一种对非关键帧进行测量率分配的模型。然后根据所建立的分配模型,利用帧间相关性提出一种快速的自适应测量率分配算法。在该算法中,以当前帧对帧组重构的贡献率为参考标准来决定该帧的测量率。实验结果表明,在相同压缩比的条件下,将本文所提出的自适应测量率分配算法应用到分布式视频压缩感知中能有效提升视频重构质量。      

基于压缩采样理论的改进型视频编解码器

压缩采样或压缩传感是数据采样并同时压缩的新理论、新方法,是现在理论研究的热点.通过对CS视频编码的研究,提出了一种基于CS理论的改进型视频编码方法,所提的改进型方法结合使用SRM和GPSR算法,比原算法更加注重利用视频本身的特性和更为合理的观测数分配方法.     

Video Event Filtering in Consumer Domain

A real-time content Altering system for live broadcasts in TV terminals enables a TV viewer to obtain desired video events from multiple channel broadcasts. In order to achieve stable and reliable event filtering for multiple video inputs, real-time filtering requirements such as frame sampling rate per channel, number of input channels, and buffer condition should be considered to meet the limited capacity of the terminal. In this paper, we propose a method of selecting those requirements by modeling a filtering system using a D/M/l queue. By minimizing the queueing time of input video frames, the proposed model can maximize filtering capacity for multiple video inputs in real-time. To verify the proposed model and analysis, we perform experiments on soccer videos with the proposed model. The experimental results show that the proposed model allows for more efficient choice of requirements within limited resources.     

A video error concealment method using data hiding based on compressed sensing over lossy channel

The video error concealment with data hiding (VECDH) method aims to conceal video errors due to transmission according to the auxiliary data directly extracted from the received video file. It has the property that can well reduce the error propagated between spatially/temporally correlated macro-blocks. It is required that, the embedded information at the sender side should well capture/reflect the video characteristics. Moreover, the retrieved data should be capable of correcting video errors. The existing VECDH algorithms often embed the required information into the corresponding video frames to gain the transparency. However, at the receiver side, the reconstruction process may loss important information, which could result in a seriously distorted video. To improve the concealment performance, we propose an efficient VECDH algorithm based on compressed sensing (CS) in this paper. For the proposed method, the frame features to be embedded in every video frame are generated from the frame residuals CS measurements and scrambled with other frame features as marked data. The marked data is embedded into the corresponding frames by modulating color-triples for its least impacts on the carriers. For the receiver, the extracted data is used to reconstruct residuals to conceal errors. Error positions are located using the set theory. Since the CS has the ability to sample a signal within a lower sampling rate than the Shannon–Nyquist rate, the original signal could be reconstructed very well in theory. This indicates that the proposed method could benefit from the CS, and therefore keep better error concealment behavior. The experimental results show that the PSNR values gain about 10 dB averagely and the proposed scheme in this paper improves the video quality significantly comparing with the exiting VECDH schemes.     

二维稀疏信号的联合压缩感知方法研究

传统的压缩感知理论主要考虑一维稀疏信号的感知和重构。当待处理信号是二维(2 dimension, 2D)或多维时,若直接将信号向量化处理,会造成感知矩阵维度急剧变大,使得存储和后续的重构复杂度大大增加,同时重构性能下降。为实现对2D信号的高效感知和快速重构,本文首先构建一个针对2D信号的模拟信息转换(Analog-to-Information Conversion, AIC)感知框架,通过行、列同时感知的策略实现量测值获取,以达到降低量测值存储维度的目的;其次针对压缩采样后的量测数据,提出一种2D快速迭代收缩阈值算法(2D Fast Iterative Shrinkage-Thresholding Algorithm, 2D-FISTA),并对该算法的基本迭代格式、收敛条件、参数选择以及算法收敛速度等问题进行了详细分析。仿真结果表明,所研究的算法可直接处理2D信号,具有重构速度快和存储量低等优势。      

基于压缩传感和EMD距离的视频镜头关键帧提取

关键帧提取是视频内容分析与检索技术的核心问题。提出一种基于压缩传感和EMD距离的关键帧提取方法,首先构造一个符合有限等距性质的稀疏矩阵,将帧高维特征投影到低维空间,然后通过计算帧低维特征之间的调节余弦相似度完成子镜头分割。在各子镜头中,利用EMD距离计算帧与子镜头中心的差异,并选择差异最小值所对应的帧作为该子镜头的关键帧。实验结果表明,该方法提取的关键帧能够对视频内容进行准确的描述。     

Transcoding with Resolution Conversion Using Super-Resolution and Irregular Sampling

In transcoding, quantization and other techniques could result in lower video output quality. To address this problem a novel super-resolution (SR) algorithm based on irregular sampling (IS) is presented in this paper. The high-resolution (HR) frame is obtained as an interpolation of one or more previous frames; the resulting interpolated frame has samples non-uniformly spaced in the areas where movement happened. To reconstruct the irregular sampled frame we use a well-known irregular sampling algorithm modified to perform in 2-D space. Moreover, because SR algorithms are in general computationally expensive, we also present a hardware feasibility study. The proposed solution does not target any specific application but we have specifically tested the algorithm in a transcoding environment. In particular, we have applied it to VC-1 to H.264 transcoding and applied down/up sampling. Experimental results show that the proposed algorithm improves video quality significantly.