201143442 六、發明說明： 【發明所屬之技術領域】 本發明係有關於一種三維之影像切割方法，特別是有 關於一種根據時間、空間及品質等三個參數將影像資料切 割成複數洪流(torrent)單元之三維之影像切割方法。 【先前技術】 隨著網際網路的普及，愈來愈多的使用者透過網路收 看影片或電視節目。然而，由於網路使用者的激增，傳統 主從(Client-Server)式的資料傳送方式已經不符合需求。有 鑑於此，習知方式係利用多台影片伺服器並以群播方式分 放伺服器和網路的負擔。然而，這種方式需耗費大量的成 本建置伺服器，若用戶·端不在其佈建範圍内將會受到諸多 限制。 為了解決上述問題，便將傳統之主從式伺服器改成分 散式的點對點(peer_t0_]peer，P2P)網路，透過將每一用戶也 當成一個資料來源端的方式來分散負載。因此，每部影片 將會被分割並分散到不同的用戶端，藉由同時向多個用戶 下載影片内容的方式來提高下載效率並縮短下载時間。然 1在目刖的點對點播放系統中，大多還是以檔案傳輸的 I式來長:供影音串流服務，而無法針對影音串流所要求的 及時性進行改善’因此常常造成晝面停格或破碎的現象。 ，為了解決畫面停格或破碎的現象，習知技術係根據晝 面群組(Group ofPicture，G〇p)來進行影片檔案的切割。此 種方式雖然改善了晝面破碎的現象，但由於無法改變影片 201143442 的位元率(bit她）’因此仍會有影片停格或跳躍的現象。有 鑑於此，另-傳統方式係採用可擴展視訊編碼咖趾 Video Coding，SVC)的機制來調整影片之品質層（职&1办 layer) ’藉此改變影》之位元率以適應網路戦。但這種方 式只能調整影片之品質層，無法根據使用者的裝置健愈 面大小，因此在調整的彈性上尚有不足之虞，亦可能發= 以顯示螢幕小的使用者裝置接收大晝面的影片的情況，將 因此而造成頻寬浪費的現象。 此外，另-種可調整料位元率的f知方式係轉瑪 (:SC〇ding)機制，其原理係將經由可擴展視訊編碼所編碼 的衫片重新轉碼成其他格式以娜則位轉 轉碼機制需要將影片重新編碼，因此較 —成= 象。此外’在點對點的網路架構中影片= =由夕個用戶端下載下來’若利用該轉碼格式將影片重 單一層級(品質層），則依然會存在有上述彈性不 基於上述問題，需要-種三維之影像切割方法。 【發明内容】 時間本種之影像切割方法，其主要係根 並根據使用者端之裝置等級 明數目，藉此達到檔案下载之彈性，斤=載之洪流早 本發明係提供一種三維之影;;: 時間、空間及品質等三個參數將影像資料切割成 201143442 = ==::裝置等級決定所需下载之洪流單 為、查、、 ，礤成本，為其發明又一目的。 由該技術手運狀技術手段及藉 得-維之Γ象切割方法，包含一資料練步驟，取 二：=組中該畫面群組包含-1晝面、-”面 赫細Γ母一該1晝面、該P晝面和該3畫面各包含 抽象層封包，每一該網路抽象層封包係包含-標 識別係包含一晝面相依性識別碼搁位、一晝面品質 Γ和—畫科間朗碼欄位；—資料比對步驟， 抽象i封1面、豕:=面畫面所包含之所有該網路 仙/、匕’、一比對其標頭所記載之該晝面相依性識別 的立該晝面品質識別碼攔位和該晝面時間識別碼欄位 =值；及-資料整合步驟’根據該資料比對步驟之比對結 將該I畫面、該Ρ畫面和該3晝面中，該晝面相依性 =、另·！馬攔位、5玄晝面品質識別碼攔位和該畫面時間識別碼 ，位具有相同值之所有該網路抽象層封包整合成單一洪流 單元。 次，此外，本發明揭露一種三維之影像切割方法，包含一 、貝料擷取步驟，係取得經由一編碼機制編碼後之一影像資 =，一資料切割步驟，係將該影像資料切割成複數影像區 ^ ’其中每一影像區段包含複數晝面群組，每一該晝面群 組包含一I晝面、一Ρ晝面和一Β晝面，每一該J晝面、 該Ρ晝面和該Β晝面各包含複數網路抽象層封包，每一該 網路抽象層封包係包含一標頭，該標頭係包含一晝面相依 201143442 i 性識別瑪欄位、-晝面品質識別碼欄師—晝面時間 碼攔位；-步驟，觸每1雜隨之所抖 晝面、該P晝面和該B晝面所包含之所有賴路抽象層 封包’逐-崎顯記载之該依性識別^ 位、該晝面品質識別碼攔位和該晝面時間識別碼攔位 值；及-資料整合步驟，根據該資料比對步驟之崎結果， 將每-該影像區段之所有該！畫面、該p晝面和該b晝面 中，該畫面相依性識別碼欄位、該晝面品質識別碼搁位和201143442 VI. Description of the Invention: [Technical Field] The present invention relates to a three-dimensional image cutting method, and more particularly to cutting a video data into a plurality of torrents according to three parameters: time, space and quality. The three-dimensional image cutting method of the unit. [Prior Art] With the popularity of the Internet, more and more users are watching movies or TV programs over the Internet. However, due to the proliferation of Internet users, the traditional Master-Server data transfer method has not met the demand. In view of this, the conventional method utilizes multiple video servers and distributes the burden of the server and the network in a multicast manner. However, this method requires a large amount of cost-building servers, and there are many restrictions if the user terminal is not in the scope of its deployment. In order to solve the above problem, the traditional master-slave server is changed to a distributed point-to-point (peer_t0_) peer, P2P) network, and the load is distributed by using each user as a data source. As a result, each movie will be split and spread across different clients, increasing download efficiency and reducing download time by simultaneously downloading video content to multiple users. However, in the point-to-point playback system that is witnessed, most of them are still based on the I-type of file transmission: for video streaming services, and cannot improve the timeliness required for video streaming, so it often causes a stoppage or Broken phenomenon. In order to solve the phenomenon of screen stoppage or fragmentation, the conventional technology cuts the video file according to the Group of Picture (G〇p). Although this method improves the phenomenon of broken face, it cannot change the bit rate of the film 201143442 (bit her), so there will still be a film stop or jump. In view of this, the other-traditional method uses the mechanism of the scalable video coding toe video Coding (SVC) to adjust the quality layer of the film (the role of the game), thereby changing the bit rate of the film to adapt to the network. Roller. However, this method can only adjust the quality layer of the film, and cannot be based on the size of the user's device. Therefore, there is still a shortage in the flexibility of the adjustment, and it is also possible to send a large user device with a small screen display. The situation of the film will cause a waste of bandwidth. In addition, another method of adjusting the bit rate of the material is a SC〇ding mechanism, and the principle is to re-transcode the film encoded by the expandable video coding into other formats. The transcoding mechanism requires re-encoding the movie, so it is more like = image. In addition, in the peer-to-peer network architecture, the video == downloaded by the client. If you use the transcoding format to re-weight the video to a single level (quality layer), there will still be the above-mentioned flexibility. A three-dimensional image cutting method. [Summary of the Invention] The image cutting method of the time type is mainly based on the number of the device level of the user end, thereby achieving the flexibility of the file downloading, and the invention provides a three-dimensional image; ;: Three parameters of time, space and quality cut the image data into 201143442 = ==:: The device level determines the cost of the flood of the download, check, and, and the cost of the download is another purpose. The technique of the hand-crafted technique and the method of cutting the image of the image-dimensional method include a data training step, taking two: = the group of the group in the group contains -1 face, - "face" 1昼, the P昼 and the 3 pictures each comprise an abstraction layer packet, and each of the network abstraction layer packets includes a face recognition identifier, a face quality, and a quality. The code field of the drawing room; - the data comparison step, the abstract i block 1 side, the 豕: = all the network cents contained in the face picture, 匕', one than the face number recorded in the header Dependent identification of the face quality identification code block and the face time identification code field = value; and - data integration step 'according to the ratio of the data comparison step to the I picture, the picture and In the 3 facets, the face-to-face dependency=, another·! horse block, 5 Xuanzang face quality identification code block and the picture time identification code, all the network abstraction layer packets with the same value are integrated into A single flooding unit. In addition, the present invention discloses a three-dimensional image cutting method, including a The image data is encoded by an encoding mechanism, and the data cutting step is to cut the image data into a plurality of image regions, where each image segment includes a plurality of image groups, each of which is The face group includes an I face, a face and a face, each of the face, the face and the face each comprise a plurality of network abstraction layers, each of the network abstractions The layer package includes a header, which includes a face-to-face 201143442 i-recognition mare position, a face-to-face quality identification code bar division, a face time code block stop, and a step-by-step The swaying surface, the P 昼 surface, and all of the 抽象 、 、 、 、 、 、 、 、 、 、 、 、 、 、 、 、 、 、 、 、 、 、 、 、 、 、 、 、 、 、 、 a time identification code blocking value; and a data integration step, according to the data comparison step of the step, the screen is dependent on all of the picture, the p-plane and the b-plane of the image segment Sex identification code field, the face quality identification code position and
該晝面時間制碼攔位具有相同值之所有該網路抽象 包整合成單一洪流單元。 【實施方式】 特徵及優點能更明顯 ’並配合所附圖式， 為讓本發明之上述及其他目的、 易懂’下文特舉本發明之較佳實施例 作詳細說明如下： 請參照第1圖，其係緣示本發明三維之影像切割方法 所執行之點對點網路架構的示意圖。在第！圖中，宜係顯 示多個使用者端i透過-點對點_2互相聯 用者端1具有一儲存媒體和-中央處理單元（Centeal Processing Unit，CPU)。_存媒體係連接至該㈣。該 儲存媒體係存放至少-影片資料，該cpu係執行本發明之 三維之影像切割方法。 請參照第2圖，其係綠示本發明之三維之影像切割方 法的流程圖。本發明之三維之影像⑽方法包含一資 取步驟S1、-資料切割步驟S2、一資料比對步驟s3、一 201143442 資料整合步驟S4和一資料傳送步驟S5。 在資料擷取步驟S1中，係取得經由可擴展視訊編碼 (SVC)機制編碼後之一影像資料，如第3圖所示。在資料 切割步驟S2中，係將該影像資料切割成複數影像區段 (segment)。其中，每一影像區段包含複數晝面群組，每一 晝面群組包含内部編碼畫面(Intra frame，以下稱I晝面）、 預測晝面(Predict frame，以下稱p晝面)和雙向預測晝面 (Bi-directional frame，以下稱B晝面）。由於該I晝面、p 畫面和B畫面係熟悉此領域者之通常知識，因此在此不多 做敘述。舉例來說’影像區段1〇包含晝面群組^至仏。 其中’以畫面群組G〗為例，其包含一I畫面、至少一 p圭 面和至少一 Β晝面。該I晝面和該Ρ晝面又可分成基底層 (base layer)和加強層（enhancement layer)，而該 Β 晝面僅為 加強層。每一 I晝面、P晝面和B晝面皆由複數網路抽象 層(Network Abstraction Layer，NAL)封包(NAL packet)組成 ’用以承載該晝面之影像資料。每一 NAL封包皆包含兩部 份：一標頭(header)和一資料(data)部份。第4圖係繪示NAL 封包之標頭結構。 在第4圖中，NAL封包之標頭結構系可包含一禁止零 位元(Forbidden zero bit，F)攔位、一網路抽象層參考指標 (NAL Ref Idc，NRI)攔位、一網路抽象層單元類別(NAL unit type ’ NUT)攔位、一保留位元(Reserved bit，R)欄位、一即 時解碼刷新旗標(Instantaneous Decoding Refreshing flag，I) 攔位、一優先順序識別碼(Priority ID，PID)攔位、一非内 層預測編碼旗標(No inter layer pred flag，N)欄位、一晝面 201143442 相依性識別碼(Dependency ID ’ DID)欄位、一晝面品質識 別碼(Quality ID，QID)攔位、一晝面時間識別碼(Temporal ID，TID)欄位、一使用參考基準圖片旗標(Use ref base pic flag，U)欄位、一可棄置旗標(Discardable flag ’ D)攔位、 一輸出旗標（output flag，O)欄位和一保留位元(Reserved bits，R2)欄位等。其中網路抽象層單元類別(NUT)攔位係 記載目前NAL封包的種類，其值14表示目前NAL封包為 前置(prefix)封包，其值1或5表示目前NAL封包為影音 資料封包。網路抽象層單元類別欄位值為1和5的不同在 於’其值為5的NAL封包只會出現於I晝面的基底層。 晝面相依性識別碼(DID)攔位係記載影像畫面之空間可擴 展性(Spatial Scalability)，亦即晝面的大小。晝面相依性識 別碼的值可從〇〇〇至111，〇〇〇為最小的畫面，而ηι為最 大的晝面。晝面品質識別碼(QID)攔位係記載影像晝面之品 質可擴展性(Quality Scalability)，亦即晝面品質（或解析度) 的好壞。晝面品質識別碼的值可從〇〇〇〇至111丨，〇〇〇〇為 最差的晝面品質，而1111為最佳的晝面品質。晝面時間識 別石馬(TID)欄位係記載影像畫面之時間可擴展性㈤叫㈣The kneading time code blocker has all of the network abstraction packets of the same value integrated into a single torrent unit. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS The features and advantages of the present invention will become more apparent from the following description of the preferred embodiments of the invention. The figure shows a schematic diagram of a point-to-point network architecture executed by the three-dimensional image cutting method of the present invention. In the first! In the figure, it is preferred to display a plurality of user terminals i-peer-to-peer_2. The inter-connector 1 has a storage medium and a central processing unit (CPU). _Save media is connected to (4). The storage medium stores at least - film material, and the cpu is a three-dimensional image cutting method of the present invention. Referring to Figure 2, there is shown a flow chart of the three-dimensional image cutting method of the present invention. The three-dimensional image (10) method of the present invention comprises an acquisition step S1, a data cutting step S2, a data matching step s3, a 201143442 data integration step S4, and a data transfer step S5. In the data acquisition step S1, one of the image data encoded by the scalable video coding (SVC) mechanism is obtained, as shown in FIG. In the data cutting step S2, the image data is cut into a plurality of image segments. Each image segment includes a plurality of face groups, each of which includes an intra-coded picture (Intra frame, hereinafter referred to as an I-plane), a predictive face (hereinafter referred to as a p-plane), and a two-way Bi-directional frame (hereinafter referred to as B-plane). Since the I face, the p picture, and the B picture are familiar with the usual knowledge of those skilled in the art, there is no description here. For example, the image segment 1 contains the page group ^ to 仏. For example, the picture group G is taken as an example, and includes an I picture, at least one p-plane, and at least one face. The I face and the face can be further divided into a base layer and an enhancement layer, and the face is only a reinforcement layer. Each of the I, P, and B faces is composed of a Network Abstraction Layer (NAL) packet to carry the image data of the page. Each NAL packet contains two parts: a header and a data part. Figure 4 shows the header structure of the NAL packet. In FIG. 4, the header structure of the NAL packet may include a Forbidden zero bit (F) block, a Network Abstraction Layer Reference Indicator (NAL Ref Idc, NRI) block, and a network. NAL unit type 'NUT' block, a reserved bit (R) field, an Instantaneous Decoding Refreshing flag (I) block, a priority order identifier ( Priority ID, PID), a non-intermediate predictive coding flag (N) field, a faceted 201143442 Dependency ID 'DID' field, a face quality identification code (Quality ID, QID) block, a Temporal ID (TID) field, a Use ref base pic flag (U) field, and a discardable flag (Discardable) Flag ' D) Block, an output flag (O) field and a reserved bit (R2) field. The network abstraction layer unit class (NUT) interception system records the type of the current NAL packet, and the value 14 indicates that the current NAL packet is a prefix (prefix) packet, and the value 1 or 5 indicates that the current NAL packet is a video data packet. The network abstraction layer unit category field values are different from 1 and 5 in that the NAL packet whose value is 5 will only appear on the base layer of the I plane. The facet correlation identifier (DID) interceptor records the spatial Scalability of the image frame, that is, the size of the facet. The value of the facet recognition code can be from 〇〇〇 to 111, which is the smallest picture, and ηι is the largest picture. The face quality identification code (QID) is used to describe the quality of the image, ie the quality of the face (or resolution). The facet quality identification code can range from 〇〇〇〇 to 111丨, which is the worst kneading quality, and 1111 is the best kneading quality. The time of recognizing the stone horse (TID) field is the time scalability of the recorded image (5) called (4)
Scalability)’其值代表影像於壓縮時之框率饵繼的倍 數，例如其值為001時代表影像壓縮時之框率倍數為2。 NAL封包標縣獅其侧位由於與本料無關，因此在 此不多加敘述。 以第3圖之晝面群組為例，其包含一 j晝面、至少 二P畫面和多個B晝面。其中’ Ϊ畫面3包含三—NAL封包 NAL(0,0,0) 3卜NAL(〇,〇,〇) 32和财⑽丄❻）％。對封包 201143442 ^AL(0,0，G) 31而。，其包含—標頭和__資料，其標頭所包 3之晝面相依性識別碼、晝面品質識別碼和晝面時間識別 碼的值皆為〇，而其NUT為14。對封包nal(〇,〇,〇)幻而 s，其包含-標頭和-資料，其標頭所包含之晝面相依性 識別碼、晝面品質識別石馬和畫面時間識別碼的值亦為〇， 然其NUT為1或5。對封包鼠⑽⑼33而言，亦包含 ‘頭和-貝料’其標頭所包含之晝面相依性識別碼的值 為〇、晝面品質識別碼的值為！而晝面時間識別碼的值為〇 ’而其NUT為20。如圖所示，B畫面4包含三隱封包 NAL(1，〇,0) 4 卜 NAL(1，〇,〇) 42 和取印山〇) 43 ; p 書面 $ 包含三 NAL 封包 NAL(〇,〇,〇) 51、nal(〇,〇,〇) Μ 和 NAL(〇，l，〇) 53 ; B 畫面 6 包含三 NAL 封包 NAL(_ Μ 、NAL(l，〇,〇) 62 和 NAL(l，l，0) 63 ; ρ 畫面 7 包含三職 封包 NAL(0,0,0) 71、NAL(〇,〇,〇) 72 和 73。每 一上述NAL封包皆同樣包含一標頭和一資料。 據此，本發明之資料比對步驟S3係針對i畫面3、ρ 晝面5、7和所有B畫面(含未魏之B畫面)所包含之所有 NAL封包，逐-比對其標頭所記載之該晝面相依性識別碼 、該晝面品質識別碼和該畫面時間識別碼的值，以便進行 資料整合步驟S4。詳言之，NAL封包31_33、4M3、We 、61-63、71_73和其他未標號者，係彼此相互比對其標頭 所記載之該晝面相依性識別碼、該畫面品質識別褐和該晝 面時間識別碼的值。 旦 本發明之資料整合步驟S4係根據資料比對步驟幻之 比對結果，將該I畫面、該p晝面和該6晝面中，標頭具 201143442 有相同該晝面相依性識別碼、該 。 時間識別碼之所有NAL封包整合碼㈣晝面 。舉例來說侧3圖，14^；^== =、Ρ畫面5…包一，。，。)5:和5=) ρ晝面7之NAL封包隱(GA0) 71和72 = 之畫面相依性識別碼、畫面品 1碩所包含 的值皆相同（同為0)，因此該等_封^面 之；1和:2,成- T_單元一 封包 NAL 二 33 ’ P旦面5之NAL封包狐⑽⑻^，以及p ， 之匡封包狐⑽）73，其標頭所包含之晝面=性7 =別碼1面品質識別碼和晝面時間識別碼的值皆相同 一面相依性識別碼同為〇、畫面品質識別碼同為J，晝 間識別碼同為0)，因此該等NAL封包33、53和^係替 合成一 Torrent單元(0，1，0)，亦稱為和加強層之T〇rrem單 兀(0，1，0)。。據此，NAL封包31、32、51和52係整合 Torrent單元(1，〇,〇) ’亦稱為和加強層之丁_加單元 。同理’ NAL封包43和63係整合成一 Torrent單元’ ’亦稱為和加強層之Torrent單元(1，1，〇彡。 為方便說明本發明上述Torrent單元(〇,〇，…、（〇，〗〇)、 (】，〇,〇)和(1，1，0)之糊過程，係僅以晝面群組&為例 上述四T〇rrent單元係為晝面群組Gi之所有j晝面、p晝面 和B晝面的NAL封包所構成。然而，在實際的實施例中 ，上述 Torrent 單元(〇,〇,〇)、（〇，〗，〇)、（1，〇,〇)和(〗，1〇)係為 有晝面群組G]至GK之所有I晝面、P晝面和B晝面的 201143442 封包所共同構成，而非單純僅由晝面群組G]之所有j晝面 、P晝面和B晝面的NAL封包所構成。其中，晝面群組 ，至GK之I畫面、P晝面和B晝面的NAL封包構成T〇rrent 單元(0,0,0)、（0，1，0)、（1，〇,〇)和〇，〗，〇)之切割過程係與畫面 群組G]相同’因此不再重複敛述。 至此 ’ Torrent 單元(〇,〇,〇)、（〇，1〇)、（1,〇,〇)和(u，〇)由 晝面群組01至GK組成之過程已詳述如上。然而，在實際 的實施例中，Torrent單元並非只有(〇,〇,〇)、（〇1，〇)、（1，〇,〇) 和（1，1，0)四者，而是由((^刈至仏叫十其中卜瓜和^^皆 為任意整數。Torrent單元之分布情形可參見第5圖所示。 以上係以第3圖之影像區段1〇為例說明Torrent單元之切割 凊形，其他影像區段L至IN之T〇rrent單元切割程序亦同 ，因此不再贄述。 據此，由複數影像區段至ιΝ所形成的影像資料可被 切割成許多Torrent單元(〇,〇,〇)至(1，m，n)，並為使用者端工 透過點對點網路2互相傳送。當其中一使用者端丨欲下載 影像資料時，可視使用者端1本身的配備水準決定所需下 裁之T0rrent單元的多寡。舉例來說，參見第如圖當使 用者端1為不具有高階顯示功能的手機裝置時，使用者端 1僅需從點對點網路2下載基底層之Torrent單元(〇,〇,〇)即 可滿足手機之顯示功能。此時，參見第6a圖之Torrent單 凡結構，所下载的影像資料因僅含單一 Torrent單元(〇,〇,〇) ’因此係無法調整所下載之影像資料的影像品質、晝面大 1和框率等等。此外，如第6b圖所示，當使用者端1為顯 不功能略佳的個人數位助理(Personal Digital Assistant)時 一 12 — 201143442 ，㈣· 1可從輯點 (1，〇,〇)、⑽）、（u 載 T_nt 皁元(〇,〇,〇)、 影像資料具有 ，（2山〇)。據此，所下載的 畫面品質1)，以及：古°〇t(基本的畫面品質0和較佳的 1和框率最㈣(框率G、框率中等的框率 干取内的框率2)，但盔 干 第6C圖所示，當使用者端二^ =尺寸）。如 資料除了具有如“圖 擴===_== 係代表較大的晝面;Γ:， 於較小的蚩 Μ ρ’所下載的影像資料可切換 間二 = 擴展性0)和較大的晝面空間(空 功能^強的I)之間。參見第6d圖，當使用者端1為顯示 ^ 、桌上型電腦時’射從輯闕路2下載更多 ’，）（，2,〇)、(1，2,〇)、(2,2,〇)、(〇,〇，”、(1，〇 ，山1)、（⑽、⑽、（咖、（咖 所下载的轉㈣她於帛0e目具妓佳的 當使_ 1為具有更強顯;功:的 解析又電視(High Definition Television)時，其可從點對點 網路2下載更完整數量的Torrent單元以迎合高解析度電視 強大的顯示功能，如第6e圖所示。據此，本發明允許使用 者端1根據其自身的顯示功能強弱，選擇所欲下載之 —】3 — 201143442 T_nt單元的多寡。如此—來，#使用者端1本 功能較差時，只需要下载少數的^咖單元即 ， 硬體設備的播放需求，而不需要下載完整數量的 身 單元而浪費錢的㈣。反之，當❹麵2本 功能較強時，可選擇下載更錢量的加咖單 二不 精緻的影像播放，提升影像播放的水準。 仃更 至=本發明已說明影像請(見第3圖)如何被切 Torrent單元，麟送至賴點網路2供使用者端1下截。 以下，本發明將說明使用者端丨下载Twem $元、 Torrent單元及播放重組之T〇rrent單元的程序。 、、且 第7圖係緣示本發明之使用者端1之操作方塊圖，复 =包含-設定程序S71、-通訊程序奶、一頻寬管理程 序S73、-資料存取程序S74、一洪流比對更新程序奶 、。-候選人管理程序S76、一時間檢知程序S77和一播放 程序S78。 在流程開始之初，係於使用者端1建立-用戶介面。 在設定程序S71中’係根據使用者端1的裝置類型來設定 像資料以及Torrent單元。亦即，當使用者端1 的裝置愈碰時，則下載愈多的T_nt單元，以便符合自 ，，設備。#欲下载之影像㈣以及T_nt 早兀決;t之後’ it訊料S72㈣始資料的傳送。在資料 傳送之初’通訊程序S72係先向頻寬管理程序S73取得連 線數f ’再根據連線數目建立多條連線下載丁_扮單元。 3管理程序S73係不斷侧點對點網路2的頻寬狀況並 计异出使用者端1的剩餘可用連線數目，使得通訊程序S72 201143442 能藉此建立使用者端1車蝻 ^ 像資料的T_m單元時，合s72接收到影 町θ將接收的Torrent單元交由資 =程序S74進行存取》其中資料存取程序S74二 ="=健存至相關的储存裝置，例如記憶體。此 、省t 更斤程序S75係偵測所下載的加咖單元是 :已：完整接收’並將了_單元的接收狀態 ==S74t候選人管理程序挪。候選人管理程序咖 將視Tb刪t早摘接收狀敎新其㈣人料，Scalability)' represents the multiple of the frame rate of the image at the time of compression. For example, the value of 001 represents a frame rate multiple of 2 when the image is compressed. The NAL package is not related to the material, so it is not mentioned here. Taking the face group of FIG. 3 as an example, it includes a j-plane, at least two P pictures, and a plurality of B-planes. Where ' Ϊ screen 3 contains three - NAL packets NAL (0, 0, 0) 3 NAL (〇, 〇, 〇) 32 and Cai (10) 丄❻)%. For the packet 201143442 ^AL(0,0,G) 31 instead. It contains the header and __ data, and the value of the facet identification code, the face quality identification code and the face time identification code of the header of the package 3 are both 〇 and the NUT is 14. For the package nal (〇, 〇, 〇) illusion and s, which contains - the header and - data, the header contains the face-to-face identification code, the face quality identification stone horse and the picture time identification code value For 〇, its NUT is 1 or 5. For the packaged mouse (10) (9) 33, the value of the face-to-face identification code included in the header of the head and the shell material is also the value of the 〇 and the face quality identification code! The facet time identification code has a value of 〇 ' and its NUT is 20. As shown in the figure, B picture 4 contains three hidden packets NAL (1, 〇, 0) 4 卜 (1, 〇, 〇) 42 and 取 〇 〇 43; p written $ contains three NAL packets NAL (〇, 〇,〇) 51, nal(〇,〇,〇) Μ and NAL(〇,l,〇) 53 ; B Picture 6 contains three NAL packets NAL(_ Μ , NAL(l,〇,〇) 62 and NAL( l,l,0) 63 ; ρ picture 7 contains the three-part packet NAL(0,0,0) 71, NAL(〇,〇,〇) 72 and 73. Each of the above NAL packets also contains a header and a According to the present invention, the data comparison step S3 of the present invention is directed to all NAL packets included in i picture 3, ρ 昼 5, 7, and all B pictures (including un-B pictures). The face correlation identification code, the face quality identification code and the value of the picture time identification code are recorded in the head for performing the data integration step S4. In detail, the NAL packets 31_33, 4M3, We, 61-63, 71_73 and other unlabeled persons are mutually compared with each other according to the facet identification code recorded by the header thereof, the picture quality recognition brown and the face time identification code. The data integration step S4 of the present invention is According to the data For the result of the step phantom comparison, in the I picture, the p 昼 face and the 6 昼 face, the header has the same facet identification code as the 201143442, and all the NAL packet integration codes of the time identification code (4) For example, side 3, 14^; ^== =, Ρ screen 5... package one, ., .) 5: and 5=) NAL 7 7 NAL packet hidden (GA0) 71 and 72 = The screen dependency identifier and the screen product 1 are all of the same value (the same as 0), so the _ 封 面 ;; 1 and: 2, into - T_ unit a package NAL two 33 'P Dan Face 5 of the NAL packet fox (10) (8) ^, and p, then the package fox (10) 73, its header contains the face = sex 7 = the code 1 face quality identification code and the face time identification code are the same value The dependency identifiers are the same, the picture quality identification code is J, and the inter-day identification code is 0), so the NAL packets 33, 53 and ^ are combined to form a Torrent unit (0, 1, 0), also known as T和rrem is a single 兀 (0, 1, 0) for the reinforcement layer. . Accordingly, the NAL packets 31, 32, 51, and 52 are integrated with the Torrent unit (1, 〇, 〇) ‘also referred to as the tens-add unit of the enhancement layer. Similarly, the 'NAL packets 43 and 63 are integrated into a Torrent unit' and are also referred to as the Torrent unit of the enhancement layer (1, 1, 〇彡. For convenience of description of the above Torrent unit of the present invention (〇,〇,...,(〇, 〇), (], 〇, 〇) and (1,1,0) paste process, only the face group & for example, the above four T〇rrent unit is all of the face group Gi The NAL packet of the facet, the facet, and the B face is constructed. However, in the actual embodiment, the above Torrent unit (〇, 〇, 〇), (〇, 〗, 〇), (1, 〇, 〇 ) and (〗, 1〇) are composed of all the I昼, P, and B faces of the face group G] to GK, not just the face group G] All the NAL packets of the J-plane, the P-plane and the B-plane are formed, wherein the NAL packet of the I-picture, the P-plane and the B-plane of the GK constitute a T〇rrent unit (0, The cutting process of 0,0), (0,1,0), (1,〇,〇) and 〇, 〗, 〇) is the same as the group of pictures G] 'so no longer repeats. So far' Torrent unit (〇,〇,〇), (〇,1〇), (1,〇,〇) The process of combining (u, 〇) by the face group 01 to GK has been described above in detail. However, in the practical embodiment, the Torrent unit is not only (〇, 〇, 〇), (〇1, 〇), (1, 〇, 〇) and (1,1,0) four, but by ((^刈到仏叫十中卜瓜和^^ are all arbitrary integers. For the distribution of Torrent units, see Figure 5 As shown in the above figure, the image segment 1〇 of Fig. 3 is taken as an example to illustrate the cutting shape of the Torrent unit, and the T〇rrent unit cutting procedures of other image segments L to IN are also the same, and therefore will not be described again. The image data formed by the plurality of image segments to ι can be cut into a plurality of Torrent units (〇, 〇, 〇) to (1, m, n), and transmitted to each other through the peer-to-peer network 2 for the user to work. When one of the user terminals wants to download the image data, the size of the T0rrent unit to be cut can be determined according to the level of the user terminal 1 itself. For example, see the figure when the user terminal 1 does not have a high order. When the function of the mobile phone device is displayed, the user terminal 1 only needs to download the base layer of the Torrent unit from the peer-to-peer network 2 (〇, 〇, 〇) The display function of the mobile phone can be satisfied. At this time, referring to the Torrent single structure in Fig. 6a, the downloaded image data only contains a single torrent unit (〇, 〇, 〇) 'so it is impossible to adjust the downloaded image data. Image quality, face size 1 and frame rate, etc. In addition, as shown in Figure 6b, when the user terminal 1 is a personal digital assistant (Personal Digital Assistant) that is not functionally better, a 12-201143442, (4) 1 can be from the point (1, 〇, 〇), (10)), (u contains T_nt soap yuan (〇, 〇, 〇), image data has, (2 hawthorn). According to this, the downloaded picture quality 1), and: ancient ° 〇 t (basic picture quality 0 and better 1 and frame rate most (four) (frame rate G, frame rate medium frame rate within the frame rate 2), but the helmet is shown in Figure 6C, when the user ends with ^^ size). For example, if the data has a larger image such as “Graphic expansion ===_==, the image data downloaded by the smaller 蚩Μ ρ′ can be switched between two = scalability 0) and larger. Between the facet space (empty function ^ strong I). See Figure 6d, when the user terminal 1 is for display ^, desktop computer 'shot from the bridge 2 download more',) (, 2 ,〇), (1,2,〇), (2,2,〇), (〇,〇,", (1,〇,山1),((10),(10),(咖,( (4) When she is better at 帛0e, she can download a more complete number of Torrent units from peer-to-peer network 2 to meet the high level of _1. The powerful display function of the resolution TV is as shown in Fig. 6e. Accordingly, the present invention allows the user terminal 1 to select the number of T_nt units to be downloaded according to its own display function strength.] 3 - 201143442 T_nt unit. Come, #用户端1 This function is poor, only need to download a few ^ coffee unit, that is, the playback requirements of the hardware device, without having to download a complete number of body units Costly (4). Conversely, when the function of the face 2 is strong, you can choose to download a more expensive amount of the coffee and the two are not refined image playback, and improve the level of image playback. 仃更到=The invention has explained the image please (See Figure 3) How to cut the Torrent unit, and send it to the network 2 for the user terminal 1. Below, the present invention will explain that the user downloads the Twem $ element, the Torrent unit, and plays the reorganized T. The program of the 〇rrent unit, and the seventh diagram shows the operation block diagram of the user terminal 1 of the present invention, the complex=include-setting program S71, the communication program milk, the bandwidth management program S73, and the data storage The program S74, a flooding update program milk, a candidate management program S76, a time detecting program S77, and a playing program S78. At the beginning of the process, the user interface 1 is established with the user interface. In the setting program S71, the image data and the Torrent unit are set according to the device type of the user terminal 1. That is, when the device of the user terminal 1 is more touched, the more T_nt units are downloaded, so as to comply with the device. #要影像的影像(四)和T_ Nt early decision; after t's message S72 (four) start data transmission. At the beginning of the data transmission 'communication program S72 first to the bandwidth management program S73 to obtain the number of connections f ' and then establish a number of connections according to the number of connections Download Ding_Dressing Unit. 3 Management program S73 is constantly on the side-to-point network 2 bandwidth condition and counts the number of remaining available connections of user terminal 1, so that communication program S72 201143442 can establish user terminal 1 car蝻^ When the T_m unit of the image data is received, the s72 is received by the shadow town θ, and the received Torrent unit is transferred to the program = program S74 for access. The data access program S74 2 ="= is stored in the relevant storage device. For example, memory. This saves the program. The S75 system detects that the downloaded add-on unit is: already: complete reception ‘and the reception status of the _ unit == S74t candidate management program. The candidate management program will treat the Tb as early as the receipt of the new (4) person.
^載之了_辑摘料，並蚊㈣要下狀τ_； 早兀時間檢知程序S77包含-計時器，當影像資料之某 一影像區段的播放時間到達時，時間檢知程序奶將通知 資料存取程序S74將該影像區段的如_單元進行重組， 並將重組後的Torrent單元送至播放程序S78進行播放γ在 播放程序S78中，會對該影像區段進行解碼 後的影㈣放料。 在上述私序中，當使用者端1完成一個影像資料之影 像區段之T〇rrent單元的下載時’使用者端i會判斷其緩ς 空間是否奴夠的空間暫存新的影像區段。若有足夠的空 間可再下載該新的影像區段，則判斷是否還有時間可下g 該新的影像區段’或者是該影像區段所有1〇订6加單元已於 之剷下載。假设該影像區段剩餘可下載之時間為 . * J bailing *^Loaded _ series of extracts, and mosquitoes (four) to be under the shape τ_; early detection time program S77 contains - timer, when the playback time of a certain image segment of the image data arrives, the time detection program milk will The notification data accessing program S74 reorganizes the _ unit of the image segment, and sends the reorganized torrent unit to the playing program S78 for playing γ. In the playing program S78, the image segment is decoded. (4) Discharge. In the above private sequence, when the user terminal 1 completes the download of the T〇rrent unit of the image segment of the image data, the user i will judge whether the buffer space is sufficient or not to temporarily store the new image segment. . If there is enough space to download the new image segment, it is determined whether there is still time to download the new image segment or all of the image segments of the image segment have been downloaded. Assume that the remaining downloadable time of the image segment is . * J bailing *
Waning = tsegmenl x An » An = SNi-SNJ 其中挪，為欲下載之影像區段的序列號碼，聊"為正被 解碼之影像區段的序列號碼。 7 —15 — 201143442 一如果沒有剩餘的時間或是該影像區段所有Torrent單 7G已於之前下載’料準備下載下—鋪的影像區段的 單元。如果尚有時間且該影像區段的所s To職t 單兀尚未下载完全，則計算剩餘的可用連線數目，如下： 連線數目=可用頻寬/連線之佔用頻寬 …田尚有連、線數目時就開始下載尚未下載的T_nt單 二並判斷其疋否完整接收’若未完整接收卿續判斷是否 ^有時間!_触’若已完整接㈣賴TGiTem單元桿記 為已下载。’並，__是否需要繼續下載其他影像區段的 _n'm此’使用者端！即可成功地接收所需的 Torrent早兀，並進行後續的播放事宜。 的是，雖然在以上的實施财，複數影像區 而又得^ t切割步驟S2中切割影像資料而得，進 像區段所包含的複數晝面群組。然而，資料 擷取步驟S1亦可直接取得一或多 料切割步驟S2。 4夕個晝面群組，因而省略資 本發明係提供一種三維之影像 1 夺間、空間及品質等三個參數將影像資料切割成 皁兀，並根擄使用者端之裝置 歿數洪/瓜 元的數目，而達到檔案下载彈性的功效/下载之洪流單 時間其主要係根 單元，並根據使用影像資料切割成複數洪流 佩优用考糕之裴置等級決 几的數目，而達到節省時間成本的功效。载之洪流單 雖然本發明已利用上述較佳實施例揭示，然其並非用 201143442 以限定本發明，任何熟習此技藝者在不脫離本發明之精神 和範圍之内，相對上述實補進行各種更動峰改仍屬本 者:準此本發明之倮護範圍當視後附 —17 ^ 201143442 f圖式簡單說明】 第】圖·本發明一較佳實 方法所執行之點# 仏疋一維之衫像切割 心點對點網路架構的示意圖。 第2圖··本發明該較佳實施 方法的流程圖。 κ 一、、隹之衫像切割 第3圖.本發明該較佳實 成複數影像區段，以及影像區段之架構圖。像貧科切割 包的=圖圖。:本發明該較佳實施例所述之一網路抽象層封 法所辑之爾像切割方 第6曝·本㈣#_使用者端為—手機裝置 之T0rrent單元的示意圖。 「執 第6b目··本發明當該使用者端為一個人數位助 所下载之Torrent單元的示意圖。 第6C圖：本發明當該使用者端為—筆記型電腦時 载之Torrent單元的示意圖。 、第6d圖：本發明當-使用者端為—桌上型電腦時 下载之Torrent單元的示意圖。 、第6e圖··本發明當-使用者端為_高解析度電視時所 下载之Torrent單元的示意圖。 第7圖·本發明之使用者端之操作方塊圖。 201143442 【主要元件符號說明】 〔本發明〕Waning = tsegmenl x An » An = SNi-SNJ where is the sequence number of the image segment to be downloaded, and the chat is the serial number of the image segment being decoded. 7 —15 — 201143442 If there is no time remaining or all the Torrent single 7Gs in the image section have been downloaded before the unit is ready to download the image section of the shop. If there is still time and the s To job t of the image segment has not been downloaded completely, calculate the remaining number of available connections as follows: Number of connections = available bandwidth / bandwidth occupied by the line... Tian Shangyou When the number of lines is connected, the T_nt single 2 that has not been downloaded will be downloaded and judged whether it will be completely received. If it is not completed, the judge will continue to judge whether there is time! _Touch' If it is completely connected (4) Lai TGI unit pole is recorded as downloaded . ‘And, __Do you need to continue downloading _n’m this user side of other image segments! You can successfully receive the required Torrent and carry out subsequent playback. In the above, in the above implementation, the plurality of image areas are obtained by cutting the image data in the cutting step S2, and the plurality of face groups included in the image segment. However, the data capture step S1 can also directly take one or more of the cutting steps S2. 4 昼 a face group, thus omitting the capital invention department to provide a three-dimensional image 1 capture, space and quality and other three parameters to cut the image data into saponins, and roots / user The number of yuan, and the effect of the file download flexibility / downloading the flood flow time is mainly the root unit, and according to the use of image data to cut into the number of multiple torrents, the number of the number of the test cake, to save time The cost of the effect. The present invention has been disclosed in the above-described preferred embodiments, and it is not intended to limit the present invention to those skilled in the art, and various modifications may be made to the above-described embodiments without departing from the spirit and scope of the invention. The peak change is still the same: the scope of protection of the present invention is attached as follows - 17 ^ 201143442 f Simple description of the diagram] The figure is executed by a better method of the present invention # 仏疋一维之A schematic diagram of a shirt-like peer-to-peer network architecture. Figure 2 is a flow chart of the preferred embodiment of the present invention. κ 一, 隹 隹 像 第 第 图 图 图 图 图 图 图 图 图 图 图 图 图 图 图 图 图 图 图 图 图 图 图 图 图 图 图 图 图 图 图 图 图 图 图 图 图 图 图 图 图 图 图 图 图 图 图 图 图 图Like the poor cut package = map. The method of the network abstraction layer sealing method according to the preferred embodiment of the present invention is a schematic diagram of the T0rrent unit of the mobile phone device. The present invention is a schematic diagram of the Torrent unit downloaded by the user terminal as a number of people. Figure 6C: Schematic diagram of the Torrent unit carried by the user terminal when the user terminal is a notebook computer. Figure 6d: Schematic diagram of the Torrent unit downloaded when the user terminal is a desktop computer. The 6th figure is the Torrent downloaded when the user terminal is a high resolution television. Figure 7 is a block diagram of the operation of the user end of the present invention. 201143442 [Description of main components] [Invention]
1 使用者端 2 點對點網路 3 I晝面 31 網路抽象層封包 32 網路抽象層封包 33 網路抽象層封包 4 B晝面 41 網路抽象層封包 42 網路抽象層封包 43 網路抽象層封包 5 P晝面 51 網路抽象層封包 52 網路抽象層封包 53 網路抽象層封包 6 B畫面 61 網路抽象層封包 62 網路抽象層封包 63 網路抽象層封包 7 P畫面 71 網路抽象層封包 72 網路抽象層封包 73 網路抽象層封包1 Client 2 Point-to-Point Network 3 I Surface 31 Network Abstraction Layer Packet 32 Network Abstraction Layer Packet 33 Network Abstraction Layer Packet 4 B Side 41 Network Abstract Layer Packet 42 Network Abstract Layer Packet 43 Network Abstraction Layer packet 5 P昼 51 Network abstraction layer packet 52 Network abstraction layer packet 53 Network abstraction layer packet 6 B picture 61 Network abstraction layer packet 62 Network abstraction layer packet 63 Network abstraction layer packet 7 P picture 71 network Road abstraction layer packet 72 network abstraction layer packet 73 network abstraction layer packet