2007-08-02_WO2007087613A_发明申请_WO2007087613A2 PARALLEL DECODING OF INTRA-ENCODED VIDEO_+++C_E_F_I_J_L_b_g_h+++.docx

对比文件名称:2007-08-02_WO2007087613A_发明申请_WO2007087613A2 PARALLEL DECODING OF INTRA-ENCODED VIDEO

目标专利名称:285多线程纹理解码CN104041050B

模型名称:DeepSeek-R1

### 特征比对表格

技术特征描述及公开性判断结果对比文件原文引用公开性论述
**技术特征A:** 其包含:将VP8帧的第一宏块指派给第一硬件线程并将所述VP8帧的第二宏块指派给第二硬件线程<br>**《未公开》**无相关内容。对比文件公开了并行解码H.264/MPEG4视频的宏块,并涉及多个工作实体(线程或处理器)同时处理不同的编码宏块信息部分(EMIPs)。例如,第[0046]段提到“The decoding of each of these EMIPs is carried out by a different working entity... Each DSP processor executes a separate thread of instructions”。第[0052]段描述了将EMIPs推送到共享工作队列并由不同线程弹出处理(图15,MB2和MB5)。然而,对比文件通篇针对的是H.264或MPEG4格式的视频解码,并未提及“VP8帧”或其宏块。VP8与H.264/MPEG4是不同的视频编码标准。因此,对比文件未公开将**VP8帧**的宏块指派给硬件线程的技术特征。
**技术特征B:** 以及经由以下步骤来通过所述第一硬件线程解码所述第一宏块并同时通过所述第二硬件线程解码所述第二宏块:重构所述第一宏块及所述第二宏块<br>**《隐含公开》**第[0048]段:“In the presently described example, only one EMi P was identified ... This EMIP for MBO is placed on the shared queue. The .first thread (denoted Tl in Figure .1) picks it up for decoding on the first virtual DSP processor and decodes the EMIP for MBO...”。<br>第[0052]段:“...the EMIP for MB2 and the EMIP for MB 5 therefore are both determined to be substantially decoded in step 103 at the same lime... the EMIPs for MB3 and MB6 are thereafter decoded simultaneously by different work entities (in this case, different threads).”<br>第[0034]段:“...difference macroblock. D is. added to a pixel value from macroblock P to generate a pixel value of a reconstructed (but uniiltered) macrobioek (UF).”对比文件明确描述了多个编码宏块信息部分(EMIPs)由不同的工作实体(如线程T1、T2)同时进行解码(第[0052]段)。解码过程包括逆变换(第[0031]段)和重构(第[0034]段)等步骤,以生成重建的(但未滤波的)宏块(UF)。因此,对比文件公开了通过第一线程解码第一宏块(如MBO)并同时通过第二线程解码第二宏块(如MB2和MB5),且该解码过程包含“重构”步骤。虽然对比文件未使用“硬件线程”这一术语,但其描述的“thread”或“work entity”在上下文中(如运行在DSP处理器上的指令线程)即对应于执行解码任务的硬件线程。因此,本领域技术人员能够从对比文件公开的内容中合理推断出该技术特征。
**技术特征C:** 将所述第一宏块和所述第二宏块的经重构像素存储于行缓冲器及列缓冲器中的至少一者内以执行帧内预测<br>**《直接公开》**第[0057]段:“Buffer 28 includes several vertically extending column sections 29-32 and several horizontally extending row sections 33-37.”<br>第[0057]段:“when an EMIP for a macroblock in a column is substantially complete, its bottom strip of prefiltered values is - written into the portion of buffer 2S corresponding to that column. These stored prefiltered. values . are then available for use in the decoding of the macroblock EMIP immediately below in the column,”<br>第[0058]段:“when an EMIP for a macroblock in a row is substantially complete, its right edge strip of prefiltered values is written, into the portion of buffer 28 corresponding to that row. These stored prefiltered values are then available for use in the decoding of the macroblock EMIP immediately to the right in the row,”<br>第[0056]段:“To provide access to these prefiltered values for the subsequent decoding of EMIPs for other macroblocks, the bottom edge strip of prefiltered values of the EMIP being decoded is stored into a buffer prior to deblocking filtering. Similarly, ... the right edge strip of prefiltered values of the EMIP being decoded is stored into the buffer.”对比文件详细描述了一个缓冲区28(buffer 28),该缓冲区包含垂直延伸的列部分(column sections 29-32)和水平延伸的行部分(row sections 33-37)(第[0057]段)。当对一个宏块(EMIP)的解码基本完成(即完成重构但未进行去块滤波)时,其底部边缘的预滤波值条带被存储到对应列的缓冲区部分,其右侧边缘的预滤波值条带被存储到对应行的缓冲区部分(第[0056]-[0058]段)。这些存储的预滤波值(即经重构但未滤波的像素值)正是用于后续宏块的**帧内预测**(intra-decoding),如第[0056]段所述。这与目标专利中为了执行帧内预测而将经重构像素存储于行/列缓冲器中的作用完全相同。因此,本领域技术人员能够毫无疑义地从对比文件得出该技术方案。
**技术特征D:** 以及在所述第一硬件线程或所述第二硬件线程重构另外的宏块之前,紧随所述第一宏块及所述第二宏块的重构来执行对所述第一宏块及所述第二宏块的回路滤波。<br>**《未公开》**第[0048]段:“When it. is determined that the decoding of an EMi P has been substantia].]}' completed (step 103), then process flow returns to step 10 ϊ... From step 103, decoding also continues such that the reconstructed macroblock (UF) is deblock filtered to generate a reconstructed and filtered macroblock. The resulting reconstructed and filtered macroblock is output (step 104)...”<br>第[0056]段:“...the bottom edge strip of prefiltered values of the EMIP being decoded is stored into a buffer prior to deblocking filtering.”对比文件描述了在解码(包括重构)基本完成后(step 103),继续进行去块滤波(deblock filtering)(第[0048]段)。然而,该过程描述并未限定必须在同一线程“重构另外的宏块之前”就“紧随重构”执行滤波。事实上,第[0056]段指出,为了给后续解码提供预滤波值,在去块滤波**之前**就将预滤波值条带存入缓冲区。这表明重构(产生UF)和去块滤波(产生F)是两个步骤,且存储预滤波值发生在滤波之前,但并未明确要求滤波必须“紧随”重构完成且必须在同一线程处理下一个宏块之前立即进行。对比文件的流程(图5)中,步骤103(解码完成)后返回步骤101(识别新任务),滤波(步骤104)似乎是解码完成后的后续步骤,但未强调其与“重构另外的宏块”在时序上的紧耦合关系。因此,对比文件未公开“在重构另外的宏块之前,紧随重构执行回路滤波”这一特定时序限定特征。
**技术特征E:** 其中所述第一宏块和所述第二宏块来自不同行。<br>**《直接公开》**第[0052]段及图15:“...the EMIPs for MB 2 and MB 5 are identified... Thread Ti pops the EMP for MB2 from the queue, and thread T2 pops the EMIP for MB 5 from the queue. This is the state of processing illustrated in Figure 15.”<br>(结合图7、图15可知,MB2和MB5位于帧的不同行)对比文件第[0052]段及图15明确示出了MB2和MB5的EMIPs被同时解码。根据对比文件图7的宏块帧结构,MB2和MB5显然位于不同的行(例如,MB0-MB3为第一行,MB4-MB7为第二行)。因此,对比文件直接公开了由不同线程同时解码的两个宏块(MB2和MB5)来自不同行。
**技术特征F:** 其进一步包含将未经滤波像素存储于所述行缓冲器及所述列缓冲器中的至少一者中。<br>**《直接公开》**第[0056]段:“To provide access to these prefiltered values for the subsequent decoding of EMIPs for other macroblocks, the bottom edge strip of prefiltered values of the EMIP being decoded is stored into a buffer prior to deblocking filtering. Similarly, ... the right edge strip of prefiltered values of the EMIP being decoded is stored into the buffer.”<br>第[0057]-[0058]段描述了这些值存储于缓冲区28的行和列部分中。对比文件多次明确提及,在去块滤波**之前**,将宏块边缘的“prefiltered values”(预滤波值)存储到缓冲区中(第[0056]段)。“Prefiltered values”即指经重构但尚未进行去块滤波的像素值,也就是“未经滤波像素”。这些值被存储在缓冲区28的列部分和行部分中(第[0057]-[0058]段),这与目标专利的“行缓冲器及列缓冲器”相对应。因此,该技术特征被对比文件直接公开。
**技术特征G:** 所述解码进一步包含:在每一硬件线程中重构一个宏块<br>**《隐含公开》**第[0046]段:“Each DSP processor executes a separate thread of instructions and is considered to be a different work entity.”<br>第[0048]段:“...decodes the EMIP for MBO in accordance with the methodology of Figure 4,”<br>第[0034]段:“...generate a pixel value of a reconstructed (but uniiltered) macrobioek (UF).”<br>图4及描述展示了解码流程,其中每个EMIP的解码包括逆变换(18)和重构(21)以生成UF。对比文件描述了多个工作实体(如不同的DSP线程)并行解码不同的EMIPs(第[0046]、[0052]段)。每个EMIP对应一个宏块的信息(第[0024]段)。解码方法(图4)包括重构步骤以生成重建的(未滤波)宏块(UF)(第[0034]段)。因此,每个工作实体(线程)处理一个EMIP,即解码一个宏块,该过程必然包含重构该宏块的步骤。虽然对比文件没有明确写出“在每一硬件线程中重构一个宏块”这句话,但其描述的工作模式(每个线程解码一个EMIP/宏块,解码包含重构)使得本领域技术人员能够直接且合理地推断出该技术特征。
**技术特征H:** 且接着对所述经重构宏块进行滤波。<br>**《隐含公开》**第[0048]段:“From step 103, decoding also continues such that the reconstructed macroblock (UF) is deblock filtered to generate a reconstructed and filtered macroblock.”对比文件明确说明了在解码基本完成(生成了重建的未滤波宏块UF)后,继续进行(continues such that)去块滤波(deblock filtered)以生成重建且滤波后的宏块。这公开了“对经重构宏块进行滤波”的步骤,并且“且接着”所表达的先后顺序关系也蕴含在“continues such that”的描述中。因此,本领域技术人员能够从对比文件得出该技术特征。
**技术特征I:** 其中所述解码进一步包含:将从先前宏块到所述第一宏块和所述第二宏块的经重构像素存储于所述行缓冲器和所述列缓冲器中<br>**《直接公开》**第[0057]段:“when an EMIP for a macroblock in a column is substantially complete, its bottom strip of prefiltered values is - written into the portion of buffer 2S corresponding to that column. These stored prefiltered. values . are then available for use in the decoding of the macroblock EMIP immediately below in the column,”<br>第[0058]段:“when an EMIP for a macroblock in a row is substantially complete, its right edge strip of prefiltered values is written, into the portion of buffer 28 corresponding to that row. These stored prefiltered values are then available for use in the decoding of the macroblock EMIP immediately to the right in the row,”对比文件明确描述了当一个宏块(如MB7)解码基本完成后,将其底部边缘的预滤波(即经重构)像素条带存储到缓冲区28的对应列部分(第[0057]段,图25),以及将其右侧边缘的预滤波像素条带存储到缓冲区28的对应行部分(第[0058]段,图26)。这些存储的像素值正是用于解码**后续**宏块(如下方或右方的宏块,如MB12或MB8)。这完全公开了“将从先前宏块到所述第一宏块和所述第二宏块的经重构像素存储于所述行缓冲器和所述列缓冲器中”这一技术特征,其中“先前宏块”对应已完成解码的宏块(如MB7),“第一宏块和第二宏块”对应后续待解码的宏块(如MB12、MB8)。
**技术特征J:** 以及使用所述先前宏块的经重构邻近者信息执行帧内预测。<br>**《直接公开》**第[0056]段:“To provide access to these prefiltered values for the subsequent decoding of EMIPs for other macroblocks...”。<br>第[0057]段:“These stored prefiltered. values . are then available for use in the decoding of the macroblock EMIP immediately below in the column,”<br>第[0058]段:“These stored prefiltered values are then available for use in the decoding of the macroblock EMIP immediately to the right in the row,”<br>第[0034]段:“...if the current EMIP being decoded was an intra-encoded EMIP, then blocks 22 and 23 cause a portion of a previously decoded macroblock of the current frame to be supplied... to the summer... In this intra-decoding, note that the portion of the prediction macroblock (P) is a portion of a prefϊltered macroblock...”对比文件第[0034]段明确说明了在进行帧内解码(intra-decoding)时,预测宏块(P)来自同一帧中先前已解码宏块的**预滤波**(prefiltered,即经重构但未滤波)部分。第[0056]-[0058]段进一步解释了将预滤波值(即经重构邻近者信息)存储到缓冲区28的行/列部分,正是为了用于后续宏块的帧内解码。因此,对比文件直接公开了使用先前宏块的经重构(预滤波)邻近者信息来执行帧内预测。
**技术特征K:** 其中解码包含对所述VP8帧的所述第一宏块和所述第二宏块中的每一者进行纹理解码。<br>**《未公开》**无相关内容。对比文件公开的是对H.264或MPEG4格式视频的并行解码。虽然“纹理解码”在视频解码领域可能广义地指代涉及纹理(像素)数据的解码过程,但目标专利明确限定其应用于“VP8帧”的解码。对比文件全文未提及VP8标准。VP8与H.264/MPEG4是不同的视频编码格式,其解码流程、语法和工具存在差异。因此,对比文件未公开对“VP8帧”的宏块进行解码(纹理解码)的技术特征。
**技术特征L:** 其进一步包含将所述多线程处理器集成到移动电话、机顶盒、音乐播放器、视频播放器、娱乐单元、导航装置、计算机、手持式个人通信系统PCS单元、便携式数据单元及固定位置数据单元中的至少一者中。<br>**《直接公开》**第[0003]段:“Cellular telephones provide their users more functionality...”。<br>第[0023]段:“Mobile communication device 1 in this example is a cellular telephone...”。<br>第[0062]段:“The decoding method need not be embodied on a mobile communication device, hut rather may he embodied on another device, fa one embodiment, the decoding method is performed by a video reader (for example, Windows Media Flayer) application thai executes on a personal computer. The decoding method can be performed by video decoding hardware and/or software embodied in televisions, video disc players, and other video decoding and viewing devices.”对比文件明确将其实施例描述为应用于移动通信设备(例如蜂窝电话,即移动电话)(第[0003]、[0023]段)。此外,第[0062]段进一步说明了解码方法也可以体现在其他设备上,例如执行视频播放器应用程序的**个人计算机**、**电视**、**视频光盘播放器**和其他视频解码/观看设备。这些设备涵盖了目标专利权利要求中列举的“计算机”、“视频播放器”、“娱乐单元”等。因此,对比文件直接公开了将执行并行解码的多线程处理器(或解码方法)集成到移动电话、计算机、视频播放器等设备中。

<<<C>>><<<E>>><<<F>>><<<I>>><<<J>>><<<L>>><<<b>>><<<g>>><<<h>>>

权利要求与技术特征
Powered by Django

网站备案号:渝ICP备2023012882号


重庆市非显而易见网络科技有限责任公司 A Anti NPE NPE