基于频域特征的H.264/AVC帧内编码模式快速预测

为降低H．264／AVC帧内预测计算复杂度，提出了基于频域特征的快速帧内编码模式预测算法．首先，对要编码的4×4块进行DCT变换，得到少数的几个DCT系数．然后，通过这几个系数计算块的特征方向和方差．最后，根据特征方向和方差计算最可能的几种预测模式．实验验证了该算法在不牺牲图像质量和压缩效率的基础上将帧内预测的计算复杂度平均降低了67．5％．     

基于端到端失真优化的无线视频联合编码方案

在无线视频传输环境端到端失真分析的基础上,针对H.264视频流建立了一种新的端到端失真模型.在使用LDPC信道编码方式与该失真分析模型的基础上,提出了一种基于端到端失真优化的联合信源信道编码方案.该方案根据当前无线信道的信噪比与信道带宽等状态信息,自适应地调节信源编码量化参数和信道编码码率,使端到端失真最小.实验结果表明,与固定码率编码相比,本文提出的联合信源信道编码方案可在不增加带宽的情况下显著提高视频传输质量.     

H.26L视频正向差错跟踪技术

H.26L解码器端引入的差错掩盖技术在视频序列变化较为剧烈的情况下，效果并不十分理想，为更好地遏制差错的进一步传播，介绍了一种正向视频差错跟踪技术，这种技术只对当前帧中受到差错“污染”的宏块进行帧内编码，不会带来数据量的急剧增加，在此基础上提出了多帧分级更新策略，使得差错跟踪恢复后的数据量得以平滑地过渡，实验表明，该技术对于运动较为剧烈的图像的差错恢复效果较好。     

基于深度学习的文本情感分析并行化算法

在训练集和测试集数据量大的情况下,半监督递归自编码（semi-supervised recursive auto encoder,Semi-Supervised RAE）文本情感分析模型会出现网络训练速度缓慢和模型的测试结果输出速率缓慢等问题. 因此,提出采用并行化处理框架,在大训练集情况下,基于“分而治之”的方法,先将数据集进行分块划分并将各个数据块输入Map节点计算每个数据块的误差,利用缓冲区汇总所有的块误差,Reduce节点从缓冲区读取这些块误差以计算优化目标函数;然后,调用L-BFGS （limited-memory Broyden-Fletcher-Goldfarb-Shanno）算法调整参数,更新后的参数集再次加载到模型中,重复以上训练步骤逐步优化目标函数直至收敛,从而得到最优参数集;在测试集大的情况下,模型的初始化参数为上述步骤得到的参数集,Map节点对各句子进行编码得到其向量表示,然后暂存在缓冲区中;最后,在Reduce节点中分类器利用各语句的向量表示计算各自语句的情感标签. 实例验证表明:在标准语料库MR （movie review）下本文算法精确度为77.0%,与原始算法的精确度（77.3%）几乎相同;在大数据量训练集下,训练时间在一定程度上随着计算节点的增加而大量减少.      

无链表SPIHT图像提升小波编码的硬件算法

为了解决小波零树编码算法中,SPIHT(分级树的集合分裂)算法占用大量存储空间的问题,给出了一种无链表SPIHT图像压缩算法.在这种算法中,小波系数按照SPIHT的顺序输入到编码器中,编码器按照NLS算法编码,小波变换采用9/7提升分解算法,量化算法为均匀标量量化,图像边界延拓采用了补零延拓方法.实验结果表明,该算法图像压缩比与JPEG2000的EBCOT(基于优化截断的嵌入式块编码)算法几乎相同,编码速度大约是后者的2倍,一幅512×512的图像仅需要17.875KB额外内存.     

MIMO双向中继网络的预编码设计与功控策略

为加强MIMO双向中继系统的空间复用增益,研究一种低复杂度的发送和接收预编码矩阵.利用子空间对齐方法,将双向MIMO信道分解为多路单入单出（SISO）的子信道形式,使得两个源用户能够使用网络编码获取更好的空间复用增益.同时通过矩阵计算和转化,给出了一种优化的功率分配方案.在确定优化矩阵后,该方案能够为每个子信道独立地进行优化功率分配,并且能够得到各节点间优化功率分配的闭合表达式,从而将算法复杂度从O（n3）降低为O（n）.仿真结果表明,在典型场景下,所提方案在具有更低复杂度的优势下,系统性能接近优化的梯度下降迭代方案,优于传统单纯前向放大转发方式（AF） ,有2.99 bit/（sHz）的性能增益.      

基于Ｇabor匹配甚低码率视频编码算法的研究

提出了一种基于Ｇabor匹配的甚低比特视频编码算法，通过算法自学习优化Ｇabor匹配集提高编码效率，实验测试结果表明，在甚低码率（２４kb／s和１０kb/s）下有效保证视频编码ＰＳＮＲ和编码效率，同时从根本上解决传统编码方法在甚低码率视频编码中的块效应问题。     

分布式视频编码算法仿真软件设计

分布式视频编码具有编码简单、高效压缩等特点,近些年国内外设计实现了多种分布式视频编码算法.本文提出一种基于变换域的分布式视频编码仿真测试软件设计方案,以测试QICF、CIF以4∶2∶0采样的YUV原始视频文件,通过改变量化参数、提取位平面参数改变视频文件传输比特率,以测试该算法在不同比特率下的率失真性能,通过更改量化表和运动预测搜索半径,搜索块大小,以适用不同序列更改设置,达到最佳性能.该方案可以验证分布式视频编码中各个模块的功能,并可以提出相应模块的各种改进算法验证其性能.     

提升高精度图像弱目标视觉质量的高效压缩算法

为了实现对高精度图像进行高效压缩,同时确保重建图像的弱目标区域有较好的保真效果,提出了一种提升弱目标区域质量的基于误差优化编码的高精度图像压缩算法. 首先,使用JPEG-LS （joint photographic experts group lossless）压缩算法对图像数据进行压缩,在游程编码过程中自适应地选择需要二次编码的误差数据,并完成了基于视觉质量的非均匀量化;其次,对量化值进行数据分解,去除量化值之间的相关性,并对分解后的数据进行MQ算术编码的熵编码;图像重建时根据量化间隔重建反量化值,并设计了反量化优化和滤波优化过程;最后,将本文算法与JPEG-LS、JPEG2000 （joint photographic experts group 2000）算法进行了性能比较,结果表明:本算法能够实现高精度图像数据的高效压缩,且复杂度低,易于硬件实现,虽然引入了误差数据二次优化编码等过程,但增加编码的数据量较小,故与JPEG-LS算法的压缩速度相当,然而比JPEG2000算法的压缩速度提升4.47倍;同时有效减少了常规算法造成的信息损失,重建图像的峰值信噪比与JPEG-LS、JPEG2000相当或略低,但弱目标区域的视觉质量及保真效果更好.      

基于种群和荧光素自然感应的萤火虫算法

针对基本萤火虫算法优化多模函数时计算复杂度高和需要预先设定较多参数值的问题,提出了两种修正的萤火虫算法:基于种群的萤火虫算法(S-GSO)和基于荧光素自然感应的萤火虫算法(LNS-GSO).这种改进使学习行为更符合自然界生物的学习规律,更有利于萤火虫发现问题的所有局部最优解.通过6个标准测试函数测试,结果表明结合运用这两种修正的萤火虫算法能够取得良好的收敛性,在寻找多模函数的峰值个数上显示出较强的优势.     

Fast Mode Decision Method Based on Historical Information for H. 264/AVC

Mode decision based on rate-distortion optimization algorithm brings high compression efficiency to H.264/AVC. However, heavy computation load is also added to the encoder at the same time. In order to reduce the computation burden of mode decision, this paper presented a fast mode decision method based on mode information of the previously coded frame. Moreover, all coding modes were activated when scene change occurs and a scheme to detect scene change was proposed. The simulation results show that compared to the original encoder, the proposed method achieves a reduction of over 38% encoding time and keeps almost the same PSNR and bitrate for a wide range of quantization parameter.     

A rate-distortion based quantization level adjustment algorithm in block-based video compression

In this paper, a rate-distortion based quantization level adjustment (RDQLA) algorithm is presented. Based on the rate-distortion criterion, the quantization level adjustment algorithm effectively improves coding efficiency by adaptively optimizing quantization levels of the signals near the boundaries of quantization cells and adjusting quantization levels per block. In addition, it has no overhead and is fully compatible with the existing compression standards. The proposed algorithm can be applied in any block based image and video coding method. In particular, the algorithm has been verified on the platform of H.264. Experimental results show that the proposed algorithm improves objective and subjective performances substantially. It is shown that the proposed algorithm has a gain of several dB comparing with the newest H.264 standard for high bit rates. Foundation item: the Special Foundation of Outstanding Young Teacher of Shanghai     

An improved frame-layer bit allocation scheme for H.264/AVC rate control

In this paper, we aim at improving the video quality degradation due to high motions or scene changes. An improved frame-layer bit allocation scheme for H.264/AVC rate control is proposed. First, current frame is pre-encoded in 16×16 modes with a fixed quantization parameter (QP). The frame coding complexity is then measured based on the resulting bits and peak signal-to-ratio (PSNR) in the pre-coding stage. Finally, a bit budgetis calculated for current frame according to its coding complexity and inter-frame PSNR fluctuation, combined with the buffer status. Simulation results show that, in comparison with the H.264 adopted rate control scheme, our method is more efficient to suppress the sharp PSNR drops caused by high motions and scene changes. The visual quality variations in a sequence are also relieved.     

Error Robust H. 264 Video Transmission Schemes Based on Multi-frame

Multi-frame coding is supported by the emerging H. 264. It is important for the enhancement of both coding efficiency and error robustness. In this paper, error resilient schemes for H. 264 based on multi frame were investigated. Error robust H. 264 video transmission schemes were introduced for the applications with and without a feedback channel. The experimental results demonstrate the effectiveness of the proposed schemes.     

A High-performance Low Cost Inverse Integer Transform Architecture for AVS Video Standard

A high-performance, low cost inverse integer transform architecture for advanced video standard (AVS) video coding standard was presented. An 8×8 inverse integer transform is required in AVS video system which is compute-intensive. A hardware transform is inevitable to compute the transform for the real-time ap-plication. Compared with the 4×4 transform for H.264/AVC, the 8×8 integer transform is much more complex and the coefficient in the inverse transform matrix Ts is not inerratic as that in H.264/AVC. Dividing the Ts into matrix S8 and R8, the proposed architecture is implemented with the adders and the specific CSA-trees instead of multipliers, which are area and time consuming. The architecture obtains the data processing rate up to 8 pixels per-cycle at a low cost of area. Synthesized to TSMC 0.18 μm COMS process, the architecture attains the operating frequency of 300 MHz at cost of 34 252 gates with a 2-stage pipeline scheme. A reusable scheme is also introduced for the area optimization, which results in the operating frequency of 143 MHz at cost of only 19 758 gates.     

AN EFFICIENT BTC IMAGE COMPRESSION ALGORITHM

IntroductionBlock truncation coding( BTC) [1] is a simple andfast lossy image coding technique,which has theadvantage of being easy to implement comparedto transform coding and vector quantization[2 ] .Its simplicity,performance and channel error re-sisting capability make it attractive in the real-time image transmission.BTC is a one- bit adap-tive moment- preserving quantizer that preservescertain statistical moments of small blocks of theinput image in the quantized output.The origi-nal …     

A high-performance low cost inverse integer transform architecture for AVS video standard

A high-performance, low cost inverse integer transform architecture for advanced video standard (AVS) video coding standard was presented. An 8×8 inverse integer transform is required in AVS video system which is compute-intensive. A hardware transform is inevitable to compute the transform for the real-time ap-plication. Compared with the 4×4 transform for H.264/AVC, the 8×8 integer transform is much more complex and the coefficient in the inverse transform matrix Ts is not inerratic as that in H.264/AVC. Dividing the Ts into matrix S8 and R8, the proposed architecture is implemented with the adders and the specific CSA-trees instead of multipliers, which are area and time consuming. The architecture obtains the data processing rate up to 8 pixels per-cycle at a low cost of area. Synthesized to TSMC 0.18 μm COMS process, the architecture attains the operating frequency of 300 MHz at cost of 34 252 gates with a 2-stage pipeline scheme. A reusable scheme is also introduced for the area optimization, which results in the operating frequency of 143 MHz at cost of only 19 758 gates.     

基于自然数编码的模式理论研究

在分析基于二进制编码的遗传算法基础上,对基于自然数编码的遗传算法作了初步的探讨。论文定义了基于自然数编码的模式和模式的阶,并推导出了相应的模式理论,从而得出了该理论与基于二进制编码的模式理论相类似的结论,对于解释基于自然数编码的遗传算法的收敛性具有一定的意义。