针对通用视频编码(versatile video coding,VVC)在编码单元(coding unit,CU)划分中引入了多类型树划分结构导致编码复杂度增加的问题,提出了一种基于CU子块方向特性与空间复杂度的快速划分算法。首先利用CU整体的纹理复杂度对当前CU进...针对通用视频编码(versatile video coding,VVC)在编码单元(coding unit,CU)划分中引入了多类型树划分结构导致编码复杂度增加的问题,提出了一种基于CU子块方向特性与空间复杂度的快速划分算法。首先利用CU整体的纹理复杂度对当前CU进行分类,筛选出不划分CU;然后利用子块不同划分方向的特性差异提前决策CU划分方向;最后利用CU中间区域与边缘区域的复杂度差异特征判断是否跳过三叉树(ternary tree,TT)划分,进一步减少候选列表划分模式数量。实验结果表明,与官方测试平台VTM10.0相比,编码器在平均输出比特率增加1.12%的代价下,编码时间减少了40.25%,说明该算法在通用视频编码中能以较小的质量损失实现更短的编码时间。展开更多
多功能视频编码(versatile video coding,VVC)作为最新一代的视频编码标准,通过引入多种高效的编码工具进一步提升了视频编码性能。然而,VVC标准引入了四叉树加多类型树(quadtree plus multi-type tree,QTMT)划分结构,并将帧内预测模式...多功能视频编码(versatile video coding,VVC)作为最新一代的视频编码标准,通过引入多种高效的编码工具进一步提升了视频编码性能。然而,VVC标准引入了四叉树加多类型树(quadtree plus multi-type tree,QTMT)划分结构,并将帧内预测模式从35种扩展到67种,导致编码复杂度急剧上升。为降低VVC的帧内编码复杂度,首先,提出了一种基于图神经网络的帧内编码单元(coding unit,CU)划分快速算法,该算法利用高效的图神经网络模型直接预测CU的最优划分模式,从而跳过冗余的CU划分遍历。其次,提出了一种基于空间相关性和纹理特征的帧内模式选择快速算法,该算法利用平均方向方差和Sobel梯度算子确定纹理方向,并跳过部分角度预测模式,同时结合预测模式间的相关性精简率失真模式列表。实验结果表明,该算法能够在BDBR(bjontegaard delta bit rate)上升2.29%的代价下,节省64.04%的编码时间。展开更多
The Base Station (BS) or access point is the building block of wireless networks, so, we propose exploiting it together with the Network Coding (NC) principle. NC suffers from the complexity of the decoding processes,...The Base Station (BS) or access point is the building block of wireless networks, so, we propose exploiting it together with the Network Coding (NC) principle. NC suffers from the complexity of the decoding processes, i.e., complicated Jordan Gaussian Elimination (JGE) processes. So, this paper proposes a deterministic NC algorithm to reduce the number of sequential network decoding steps, and hence minimizing the complexity of JGE process resulting to better time delay and processing time. We propose an algorithm that combines higher number of the transmitted packets resulting to better data-rate but worse Bet Error Rate (BER). However, using such strong Forward error correction channel code, which is Partial Unit Memory Turbo Code (PUMTC) results to minimize the losses in the BER to a very acceptable lever, in fact, in Decode-and-Forward (DF) BS, the BER can be regarded as minimum. Simulation results, for both Amplify-and-Forward (AF) and DF BS schemes using PUMTC based on (8, 4, 3, 8) component codes, confirm that using PUMTC mitigates the problem of noise aggregation resulting from applying NC in the proposed schemes.展开更多
The BEAMnrc code was used for the simulation of the Theratron Equinox-80 telecobalt machine. The phase space of radiation beam was generated at treatment distance of 80 cm for various field sizes. The phase spaces in ...The BEAMnrc code was used for the simulation of the Theratron Equinox-80 telecobalt machine. The phase space of radiation beam was generated at treatment distance of 80 cm for various field sizes. The phase spaces in air were analyzed by BEAMdp data processing program. The electron energy fluence with respect to photon energy was 0.09% and 0.34% for field size of 05 × 05 and 35 × 35 cm<sup>2</sup> respectively and it was maximum at the central axis which gradually decreases beyond this. The profiles for photon fluence were in symmetry for all the fields. The full width at half maximum of profiles in photon energy fluence shows good agreement with the field size. The photon energy fluence was flat till the field size of 27 cm<sup>2</sup>, after which it decreases gradually till the edge in larger field sizes. The air-kerma output factor from the simulation was in good agreement with measured value. We analyzed the dose data scored in the voxels in a large water phantom by simulation using dosxyznrc code. The percentage depth dose for all field sizes was in good agreement with the BJR supplement 25 and the data supplied by the manufacturer of machine. Significant deviation of about 20% in isodose line near the edge of the profile was observed for 35 × 35 cm<sup>2</sup> field size. The penumbra widths of all field sizes were comparable except for 35 × 35 cm<sup>2</sup>, which has a penumbra width of 4.1 cm at 10 cm depth. The significant under dose near the edge as compared to central axis for larger field sizes may be the indication for its careful use in treatment.展开更多
文摘针对通用视频编码(versatile video coding,VVC)在编码单元(coding unit,CU)划分中引入了多类型树划分结构导致编码复杂度增加的问题,提出了一种基于CU子块方向特性与空间复杂度的快速划分算法。首先利用CU整体的纹理复杂度对当前CU进行分类,筛选出不划分CU;然后利用子块不同划分方向的特性差异提前决策CU划分方向;最后利用CU中间区域与边缘区域的复杂度差异特征判断是否跳过三叉树(ternary tree,TT)划分,进一步减少候选列表划分模式数量。实验结果表明,与官方测试平台VTM10.0相比,编码器在平均输出比特率增加1.12%的代价下,编码时间减少了40.25%,说明该算法在通用视频编码中能以较小的质量损失实现更短的编码时间。
文摘多功能视频编码(versatile video coding,VVC)作为最新一代的视频编码标准,通过引入多种高效的编码工具进一步提升了视频编码性能。然而,VVC标准引入了四叉树加多类型树(quadtree plus multi-type tree,QTMT)划分结构,并将帧内预测模式从35种扩展到67种,导致编码复杂度急剧上升。为降低VVC的帧内编码复杂度,首先,提出了一种基于图神经网络的帧内编码单元(coding unit,CU)划分快速算法,该算法利用高效的图神经网络模型直接预测CU的最优划分模式,从而跳过冗余的CU划分遍历。其次,提出了一种基于空间相关性和纹理特征的帧内模式选择快速算法,该算法利用平均方向方差和Sobel梯度算子确定纹理方向,并跳过部分角度预测模式,同时结合预测模式间的相关性精简率失真模式列表。实验结果表明,该算法能够在BDBR(bjontegaard delta bit rate)上升2.29%的代价下,节省64.04%的编码时间。
文摘The Base Station (BS) or access point is the building block of wireless networks, so, we propose exploiting it together with the Network Coding (NC) principle. NC suffers from the complexity of the decoding processes, i.e., complicated Jordan Gaussian Elimination (JGE) processes. So, this paper proposes a deterministic NC algorithm to reduce the number of sequential network decoding steps, and hence minimizing the complexity of JGE process resulting to better time delay and processing time. We propose an algorithm that combines higher number of the transmitted packets resulting to better data-rate but worse Bet Error Rate (BER). However, using such strong Forward error correction channel code, which is Partial Unit Memory Turbo Code (PUMTC) results to minimize the losses in the BER to a very acceptable lever, in fact, in Decode-and-Forward (DF) BS, the BER can be regarded as minimum. Simulation results, for both Amplify-and-Forward (AF) and DF BS schemes using PUMTC based on (8, 4, 3, 8) component codes, confirm that using PUMTC mitigates the problem of noise aggregation resulting from applying NC in the proposed schemes.
文摘The BEAMnrc code was used for the simulation of the Theratron Equinox-80 telecobalt machine. The phase space of radiation beam was generated at treatment distance of 80 cm for various field sizes. The phase spaces in air were analyzed by BEAMdp data processing program. The electron energy fluence with respect to photon energy was 0.09% and 0.34% for field size of 05 × 05 and 35 × 35 cm<sup>2</sup> respectively and it was maximum at the central axis which gradually decreases beyond this. The profiles for photon fluence were in symmetry for all the fields. The full width at half maximum of profiles in photon energy fluence shows good agreement with the field size. The photon energy fluence was flat till the field size of 27 cm<sup>2</sup>, after which it decreases gradually till the edge in larger field sizes. The air-kerma output factor from the simulation was in good agreement with measured value. We analyzed the dose data scored in the voxels in a large water phantom by simulation using dosxyznrc code. The percentage depth dose for all field sizes was in good agreement with the BJR supplement 25 and the data supplied by the manufacturer of machine. Significant deviation of about 20% in isodose line near the edge of the profile was observed for 35 × 35 cm<sup>2</sup> field size. The penumbra widths of all field sizes were comparable except for 35 × 35 cm<sup>2</sup>, which has a penumbra width of 4.1 cm at 10 cm depth. The significant under dose near the edge as compared to central axis for larger field sizes may be the indication for its careful use in treatment.
