TW200428880A - Method and apparatus for moving picture compression rate control using bit allocation with initial quantization step size estimation at picture level - Google Patents

Method and apparatus for moving picture compression rate control using bit allocation with initial quantization step size estimation at picture level Download PDF

Info

Publication number
TW200428880A
TW200428880A TW92115798A TW92115798A TW200428880A TW 200428880 A TW200428880 A TW 200428880A TW 92115798 A TW92115798 A TW 92115798A TW 92115798 A TW92115798 A TW 92115798A TW 200428880 A TW200428880 A TW 200428880A
Authority
TW
Taiwan
Prior art keywords
day
complexity
current
activity
current day
Prior art date
Application number
TW92115798A
Other languages
Chinese (zh)
Other versions
TWI221739B (en
Inventor
Yung-Ching Chang
Chia-Chieh Chen
Teng-Kai Wang
Original Assignee
Silicon Integrated Sys Corp
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Silicon Integrated Sys Corp filed Critical Silicon Integrated Sys Corp
Priority to TW92115798A priority Critical patent/TWI221739B/en
Application granted granted Critical
Publication of TWI221739B publication Critical patent/TWI221739B/en
Publication of TW200428880A publication Critical patent/TW200428880A/en

Links

Landscapes

  • Compression Or Coding Systems Of Tv Signals (AREA)

Abstract

A method and apparatus for rate control in moving picture compression. Bit allocation with initial quantization step size estimation is used at picture level. With the relationship between pre-analyzed activity of current picture to be encoded and actual complexity of previously encoded picture of the same type, a target bit budget can be allocated to the current picture in accordance with the present invention. Once the target bit budget has been determined, an initial value for an average quantization step size is estimated on the basis of the target bit budget, along with the relationship between pre-analyzed activity of current picture and actual complexity of previously coded picture. Such an initial value of the average quantization step size is useful to achieve higher picture quality at a given bit allocation.

Description

200428880 五、發明說明(1) 【發明=屬之技術領域】 於視關於資料壓縮之領域’特別係指-種適用 方法及i置統以估計的初始量化位階執行位元率控制之 【先前技術】 位表示ί=t ’電子通訊世界正經歷—次數位革命,以數 傳收、、處理r主要優點在於能夠幾近無誤地儲存、再生、 ,NTSC的〜ΐ運用的資料位元流(MtStream)。舉例而言 480條可視訊影像每秒有29· 97張晝面,每張畫面約 ^ T目田線’每條掃瞄線約需480點以紅、綠、誃三 素(Plxel),但若每種色彩成分以8位元編 、 的位兀率(bitrate)每秒約168百萬位元(Mbits/ S \ *種視訊格式其未經壓縮處理的位元率是非常高而 不、! T二因此不適於多數的應用。 ^ 與電腦、、電信網路、消費性產品整合的數位音訊和視 : 更加刺激資訊革命的前進,而這革命的核心則是視、 音訊的數位壓縮技術。許多的壓縮標準,包含以壓縮技術 通之精知為基礎的演算法,如:I τ u — τ (前身係c C I T T )建 議書 H.261 和 H.2 6 3,以及 IS0/IEC 的 MPEG-1、MPEG-2和 MPEG-4標準。MPEG的演算法是由動態影像專家群組 (Moving Picture Experts Group,MPEG)所發展出來,該 動態影像專家群組係國際標準組織(internat ional Standards Organ i zat ion,ISO)及國際電子技術委員會200428880 V. Description of the invention (1) [Invention = technical field of the invention] The field of data compression regards' specially refers to an applicable method and system to perform bit rate control based on the estimated initial quantization level [prior art 】 Bit representation ί = t 'The world of electronic communications is undergoing a revolution of digital digits. The main advantage of digital transmission and processing is that it can store and reproduce almost error-free data streams. ). For example, 480 video images have 29.97 diurnal surfaces per second, each frame is about ^ T Mada Line 'each scanning line needs about 480 points to be red, green, and trioxin (Plxel), but If each color component is coded at 8 bits, the bitrate is about 168 million bits per second (Mbits / S \ * video formats, the uncompressed bit rate is very high, T2 is therefore not suitable for most applications. ^ Digital audio and video integrated with computers, telecommunications networks, and consumer products: It further stimulates the information revolution, and the core of this revolution is video and audio digital compression technology. Many compression standards include algorithms based on the general knowledge of compression techniques, such as: I τ u — τ (formerly c CITT) Recommendations H.261 and H.2 6 3, and IS0 / IEC MPEG-1, MPEG-2, and MPEG-4 standards. MPEG algorithms were developed by the Moving Picture Experts Group (MPEG), which is an international standard organization (international standards) Organ i zat ion (ISO) and International Electrotechnical Commission

200428880 五、發明說明(2) (International Electrotechnical Commission , IEC)的 聯合技術委員會,致力於發展視、音訊的壓縮及多路傳輸 的表現方式’這些標準規定了壓縮位元流的語法(S n t a Χ ) 以及解碼的方法,但對於編碼器所使用的演算法而言,卻 為編碼的新穎性與多樣性保留了相當多的自由度。 根據Μ P E G ’ 一連串的視訊畫面(p i c t u r e )分成一序列 的晝面群組(group of picture,G〇p),其中每組GOPwj — 晝面開始,後面跟著p-晝面和B—晝面的安排,第1圖以顯 示的順序說明一組典型的G0P。I -畫面之編碼毋須參考先 别或將來的晝面,P-畫面則參考連串的視訊晝面中在時間 上最接近的I -晝面或P-晝面來進行編碼,而B—晝面散佈於 I -晝面和Ρ-晝面之間。並且利用先前的、將來的或兩者兼 具的緊鄰I -晝面和Ρ-晝面來編碼。雖然好幾張Β -晝面可以 緊接著連續出現,但絕不能以Β -晝面預估其他的晝面。 母張晝面具有三種成分:亮度值(luminance,以γ表 示)’紅色差值(red color difference,以Cr表示),以 及藍色差值(blue color difference,以Cb表示)。對於 MPEG的4 ·· 2 : 0的格式而言,每一種Cr和Cb成分的取樣點在 水平與垂直方向只有Y成分的一半。如第2圖所示,一張 MPEG的晝面其基本構成方塊係大區塊(macrobiock,以MB 表示)。以4 : 2 : 0的視訊為例,每個MB包含一個Y成分16x1 6 取樣點的陣列,以及兩個C r和C b成分8 X 8取樣點的區塊, 其中Y成分1 6 X1 6取樣點的陣列實際上由4個8x8取樣點的區 塊所組成。200428880 V. Joint Technical Committee of the Invention Description (2) (International Electrotechnical Commission, IEC) is committed to the development of video and audio compression and multiplexing expressions. These standards specify the syntax of compressed bit streams (Snta X ) And decoding methods, but for the algorithms used by the encoder, it retains a considerable amount of freedom for the novelty and diversity of the encoding. According to M PEG 'a series of video pictures (picture) divided into a sequence of group of pictures (Gop), where each group GOPwj — day starts, followed by p-day and B — day Arrangement, Figure 1 illustrates a typical set of GOPs in the order shown. The I-picture coding does not need to refer to the day-to-day or future day-to-day. The P-picture refers to the I-day or P-day-plane that is closest in time to the series of video day-days. B-day The planes are interspersed between the I-day plane and the P-day plane. And use the previous, future, or both of the immediate I-day plane and P-day plane to encode. Although several B-days can appear next to each other in succession, the other day can not be estimated by the B-days. The mother day surface has three components: a brightness value (represented by γ), a red color difference (represented by Cr), and a blue color difference (represented by Cb). For the MPEG 4 · 2: 0 format, the sampling points for each Cr and Cb component are only half of the Y component in the horizontal and vertical directions. As shown in Figure 2, the basic composition of a MPEG diurnal surface is a large block (macrobiock, expressed in MB). Taking the video of 4: 2: 0 as an example, each MB contains an array of 16x1 6 sampling points of the Y component, and two blocks of 8 X 8 sampling points of the C r and C b components, of which the Y component 1 6 X 1 6 The array of sampling points is actually composed of 4 blocks of 8x8 sampling points.

200428880 五、發明說明(3) 、、扁碼态的作用在於決定何種畫面編碼型態以及何種預 測模式為最佳。對每張卜晝面,MB中的每個8χ8方塊均經 過離散餘弦轉換(dlscrete c〇slne transf〇rn],DCT)而形 成8x8轉換係數陣列,轉換係數接著以一量化矩陣進行量 =,然後用Z字形(zig-zag)掃瞄DCT係數的量化結果而得 到一連串的DCT係數,並且此DCT係數序 (=ri_e length)進行編碼。卜書面必^\ = : = 1型㈣型0進行編碼,1型0的編碼係以、 前圭面之’而對於每俯麵,則需得到繼以先 v:c^r)^^ # - ^ *(m〇t1〇n 及其在先^面巾的1晝面巾即將編碼的Μβ 間的預測誤差則以DCT、量曰字:?『預估Μβ與目前MB 碼。 Z子形拎瞄以及VLC進行編 處理B-畫面時,必須決 個MB :1模式、F模式、B模式以列^種MB模式來編碼每 身而不借助移動補償來編 、 旲式。1模式係以MB本200428880 V. Description of the invention (3) The function of the flat code state is to determine which picture encoding type and which prediction mode are the best. For each daylight surface, each 8x8 block in MB is subjected to a discrete cosine transform (dlscrete c0slne transfrn) (DCT) to form an array of 8x8 conversion coefficients. The conversion coefficients are then processed by a quantization matrix =, then The zigzag (zig-zag) is used to scan the quantized results of the DCT coefficients to obtain a series of DCT coefficients, and the DCT coefficients are sequenced (= ri_e length) for encoding. You must write ^ \ =: = Type 1 type 0 is used for encoding. Type 1 type 0 is based on the previous surface, and for each plane, you need to get the following v: c ^ r) ^^ # -^ * (m〇t1〇n and its previous ^ 1-day facial towel will be encoded by the prediction error between the β β, DCT, the amount of words:? "Estimated Μ β and the current MB code. Z sub-shaped target And when VLC edits B-pictures, it is necessary to determine MB: 1 mode, F mode, and B mode to encode each body in a series of MB modes without the use of motion compensation to edit and format. 1 mode is based on MB.

的向前預測編碼,以(:⑽-般);F模式為單向 麵-般L反之,= 到移動補償的預估(如P 來的畫面得到移動補償箱:°白、向後預測編碼,係以後 向的預測編碼,運肖別的抓模式,其係雔 做内插 對F、β和FB模式而言, β杈式之移動補償預估。 形掃目苗以及VLC進行編碼、。心差可以使用DCT、量化、2字 0702-7400TWf(Nl) ; 90P113 ; Lin.ptd 第8頁 ^UU428880 五、發明說明(4) 視訊編碼器很重要的_ 控制之主要目的係很有二:即為位元率的㈣。位元率 個MR所田从a - 有曰w地分配編碼每張晝面及其中每 ,T f數,使編碼過的視訊於解碼器進行解碼時 張;編碼視m的視覺品胃。編碼器必須為整 真7然:::位階:便在給定的位元率下控制可見的失 二數V妙乂〉取的$化位階編碼-張畫面所用的實際位 ::逆:L貫際編碼之後才齡,現實中並不存在 -張晝面所用的實p位^希广達到的量化位階下’決定 應性(或是可變。咖的關鍵特徵則係採用適 不同區域以不同的程产編2,這種技術允許每張晝面中的Forward prediction encoding of ():-general; F mode is one-way surface-general L and vice versa, = the estimate of motion compensation (such as the picture from P gets a motion compensation box: ° white, backward prediction encoding, It is a backward-looking predictive coding, which is different from the grasping mode. It does interpolation for F, β, and FB modes. Β-branch-type motion compensation estimation. Scanning seedlings and VLC encoding. Heart Difference can use DCT, quantization, 2 words 0702-7400TWf (Nl); 90P113; Lin.ptd page 8 ^ UU428880 V. Description of the invention (4) Video encoder is important _ The main purpose of control is two: The bit rate is the bit rate. The bit rate of each MR field is from a to w to assign a code to each day and each time, T f number, so that the coded video is decoded by the decoder; the coded video The visual taste of the m. The encoder must be true 7 ::: level: then control the visible missing two at a given bit rate Rank :: Inverse: It is only after the L-intercoding, it does not exist in reality-the real p-bit used by Zhang Tianmeng 'Decision atopic (or variable. Coffee key features are adapted system using different regions with different production processes knitting 2, this technique allows each day plane

方法卻相當複雜,通貝。不過,傳統的位元率控制 處理,除此之外好幾次才能完成視訊編碼的 以運用的簡單機制,用二,欠^ 一種適應性量化方式可 更為平均。 、°又疋初始量化位階而讓晝面品質 有鐘於此,亟需一插立 單次、即時的視訊編碼哭Γ 率控制技術,可用於 ^ ^ #1 ir ^ ά ^ ”、、叩,再者,亦期待能提供一種位元 工制方法及衣置,於晝面層級利用估計的初始量化位階 來進行位元分配壓縮動態影像。 【發明内容】 根據本發明,每一張晝面其目標位元的分配,係基於 先W晝面的編碼結果和對即將被編碼的目前晝面其預先分The method is quite complicated, Tumbe. However, the traditional bit rate control processing, in addition to several times to complete the video encoding using a simple mechanism to use, using two, a less adaptive quantization method can be more average. , ° and the initial quantization level, so that the quality of the day and night are here, and a single, real-time video encoding is urgently needed. The rate control technology can be used for ^ ^ # 1 ir ^ ά ^ ”, 叩, In addition, it is also expected to provide a bit manufacturing method and clothing that use the estimated initial quantization level at the day level to perform bit allocation to compress the dynamic image. [Summary of the Invention] According to the present invention, each day The allocation of target bits is based on the encoding result of the previous day surface and its pre-scoring of the current day surface to be encoded.

200428880 五、發明說明(5) 析的活動量,一旦目標位元分配好之後,則平均量化位階 的初始值也隨之決定。藉由目前晝面之預先分析的活動量 以及先前編碼晝面之實際複雜度間的關係,目前畫面之複 雜度將能夠被估計而得,這種預估的目前晝面複雜度對視 訊編碼器是非常有用的,可以更精確地分配目標位元額度 給每張晝面。再者,利用上述的平均量化位階初始值,視 訊編碼器在既定的位元配額條件下,可輸出較佳的晝面品 質。 本發明係針對於一種利用初始量化位階預估之動態影 像壓縮位元率控制方法。首先為一即將被編碼的目前晝面 來計算其全活動量,其中該目前晝面係在一視訊序列的一 組連續晝面之中;接著基於目前晝面之全活動量以及這組 連續晝面中同型態的先前編碼畫面其活動量對複雜度之比 ,估計目前晝面之複雜度,並且以目前晝面預估的複雜度 更新本組連續晝面之瞬間複雜度;目前畫面之目標位元額 度的分配,係依預估的複雜度以及瞬間複雜度而定;用來 編碼目前晝面之平均量化位階初始值,則可根據目前晝面 之全活動量、目標位元額度以及同型態的先前編碼晝面其 活動量對複雜度之比來決定;因此,基於平均量化位階初 始值,對視訊序列中之目前畫面進行編碼。在目前晝面編 碼之後,根據目前晝面之全活動量、目前畫面之實際消耗 位元數和目前晝面之平均量化位階,計算目前晝面其活動 量對複雜度之比,就這點而言,目前晝面的活動量對複雜 度之比與目前晝面之全活動量成正比,而與目前晝面之實200428880 V. Description of the invention (5) Once the target bit is allocated, the initial value of the average quantization level will be determined accordingly. Based on the relationship between the pre-analyzed activity of the current diurnal surface and the actual complexity of the previously encoded diurnal surface, the complexity of the current picture can be estimated. This estimated current diurnal complexity is an important factor for video encoders. It is very useful to assign the target bit quota more accurately to each day. Furthermore, by using the above initial value of the average quantization level, the video encoder can output better daylight quality under the condition of a predetermined bit quota. The invention is directed to a dynamic image compression bit rate control method using an initial quantization level estimation. First calculate its full activity for the current day surface to be coded, where the current day surface is in a set of continuous day surfaces in a video sequence; then based on the current full day activity surface and this set of continuous day The ratio of the amount of activity to the complexity of the previously coded pictures of the same type in the face, estimates the current day-to-day complexity, and updates the instantaneous complexity of this group of continuous day-to-day pictures with the current day-to-day estimated complexity; The allocation of the target bit quota is based on the estimated complexity and instantaneous complexity. The initial value of the average quantization level used to encode the current daytime plane can be based on the current total activity of the daytime plane, the target bitrate, and The previously encoded day-to-day surface of the same type has its activity to complexity ratio determined; therefore, the current picture in the video sequence is encoded based on the initial value of the average quantization level. After the current diurnal coding, the ratio of the current diurnal activity to the complexity is calculated based on the current full diurnal activity, the actual number of bits consumed in the current picture, and the average quantization level of the current diurnal surface. In other words, the ratio of current day-to-day activity to complexity is directly proportional to the current day-to-day total activity, and

0702-7400TWf(Nl) ; 90P113 ; Lm.ptd 第10頁 200428880 五、發明說明(6) 際消耗位元 另一方 之動態影像 一複雜度估 一參數更新 晝面中即將 塊匹配時, 全活動量以 動量對複雜 面之複雜度 本組連縯晝 間複雜度, 晝面之平均 晝面之全活 面其活動量 據目前晝面 前晝面之平 之比,其中 之全活動量 平均量化位 【實施方式 數以及平 面,本發 壓縮位元 計器、一 為。移動 被編碼的 计异此目 及這組連 度之比, 。以目前 面之瞬間 分配目標 量化位階 動量、目 對複雜度 之全活動 均量化位 ,目前晝 成正比, 階成反比 均量化位階 明還揭露一 率控制裝置 配額分派器 估算單元接 目别晝面, 前晝面之全 續畫面中同 複雜度估言十 畫面預估的 複雜度,且 位元額度給 初始值,則 標位元額度 之比為基礎 量、目前畫 階,計算目 面的活動量 而與目前畫 成反比。 種利用初始量化 ,包括一移動估 、一量化位階估 收視訊序列搜的 用來在執行移動 活動量。基於目 型態之先前編碼 器可因此用來估 複雜度,配額分 根據預估的複雜 目前畫面。用來 由量化位階估計 以及同型態的先 來決定。而參數 面之實際消耗位 前畫面其活動量 對複雜度之比與 面之貫際消耗位 位階預估 异早7L 、 計器以及 一組連續 估算的區 前晝面之 畫面其活 计目前畫 派器更新 度以及瞬 編碼目前 器以目前 前編碼晝 更新器根 元數和目 對複雜度 目前晝面 元數以及 為使本發明之上述目的、特徵和優點能更明顯易4 下文特舉一較佳實施例,並配合所附圖式,作礞y 懂’ 卬评細說明如0702-7400TWf (Nl); 90P113; Lm.ptd Page 10 200428880 V. Description of the invention (6) Dynamic image of the other side of the consumed bit-complexity estimation-parameter update When the block is matched in the daytime, the full activity Complexity of Complex Surfaces by Momentum This group performed the daytime complexity. The average daily activity of the full-time surface of the daily surface is based on the ratio of the current daily surface to the average daily surface. The number of modes and planes, the compression bit counter of the hair, one for. The mobile code is calculated differently for this purpose and the ratio of this group of connections,. Instantly assign the target quantization level momentum, and the full activity average quantization level of complexity to the current day, which is directly proportional to the current day, and inversely proportional to the average quantization level. The rate control device quota dispatcher estimation unit is also exposed to the day and time. In the full-continuity picture of the day-to-day picture, the estimated complexity of the ten pictures is estimated, and the bit quota is given the initial value. The ratio of the standard quota is the basic amount and the current drawing level to calculate the current activity. The amount is inversely proportional to the current picture. This method uses initial quantization, including a mobile estimation and a quantized level estimation. The video sequence search is used to perform mobile activity. The previous encoder based on the model can be used to estimate the complexity, and the quota points are based on the estimated complexity of the current picture. It is used to determine by the quantization level estimation and the isotype first. The actual consumption level of the front surface of the parameter surface is compared with the ratio of the amount of activity to the complexity of the continuous consumption level of the surface. The estimation is different as early as 7L. The degree of renewal and instant encoding is based on the current pre-encoded day-to-day updater root element number and complexity. The current diurnal surface element number and the above-mentioned objects, features, and advantages of the present invention can be made more obvious and easier. 4 A better implementation is given below. Examples, and in accordance with the drawings, make 礞 y understand '

0702-7400TWf(Nl) ; 90P113 ; Lm.ptd 第11頁 200428880 五、發明說明(7) 下·· 一開始,預測用的I -畫面和p—晝面必須先在―“編碼 器内解碼,利用計算原始晝面及解碼的晝面間的均方根 (root mean square)誤差,可以得到解碼晝面其品質的客 觀量度。以均方根決差作為解碼晝面之失真度,則解碼晝 面的複雜度(c 〇 m p 1 e X) - C定義如下: C = r 乂 d 其中,r係編碼該晝面所需的位元率,d係解碼的晝面之失 真度。理論上’平均的量化位階(q u a n t i z a t i 〇 n s t e p s l z e )和失真度之間為一比例關係,故,晝面的複雜度可0702-7400TWf (Nl); 90P113; Lm.ptd Page 11 200428880 V. Description of the invention (7) Next ... At the beginning, the I-picture and p-day plane for prediction must be decoded in the "" encoder, By calculating the root mean square error between the original diurnal plane and the decoded diurnal plane, an objective measure of the quality of the decoded diurnal plane can be obtained. Using the root mean square difference as the distortion of the decoded diurnal plane, the decoded diurnal plane is decoded. The complexity of the surface (c 0mp 1 e X)-C is defined as follows: C = r 乂 d where r is the bit rate required to encode the day plane, and d is the distortion of the decoded day plane. In theory ' There is a proportional relationship between the average quantizati onstepslze and the degree of distortion, so the complexity of the day surface can be

疋義成· ^ rxq C «-- .2 其中,q係該晝面的平均量化位階。在晝面編碼之後,將 可得知位兀率(、貫際消耗位元數)以及平均量化位階,而該 晝面的複雜度亦因此獲得。 晝面之複雜度端視t面的本f及其編碼的形式而定 先耵技術利用取近編碼晝面之複雜崖來表示目前晝面之 :隹度並:,I P及B _晝面分別保持其複雜度以減輕 同編碼形式的衫喜,偈若視訊序列的晝面 變化,則此方式可達到柏#兀拉& μ m 門令馬十順 一 咬判相當不錯的效果。然而,由於書Wu Yicheng ^ rxq C «-.2 where q is the average quantization level of the day surface. After the day-plane coding, the bit rate (and the number of bits consumed across the interval) and the average quantization level will be known, and the complexity of the day-plane will be obtained as a result. The complexity of the diurnal plane depends on the f of the t-plane and its coding form. The a priori technology uses a complex cliff near the coded diurnal plane to represent the current diurnal plane: 并 Degree and: IP and B _ diurnal plane respectively Maintaining its complexity to alleviate the same coding style. If the day-to-day changes of the video sequence, this method can achieve a very good effect of Bai Yiwu's gate order Ma Shishun. However, due to the book

編碼的目標位兀配額與其實際消耗位元數之間不相符; 如果視訊序列的内容大幅變動,上述方式 成一致的視覺品質。 卜乙口巾木The encoding target bit quota does not match the actual number of bits it consumes; if the content of the video sequence changes significantly, the above methods achieve consistent visual quality. Bu Yi

0702-7400TWf(Nl) ; 90P113 ; Lin.ptd 第12頁 200428880 五、發明說明(8) 根據本發明,預先分析目前晝面的活動量(a c t i v i t y ) 以及先前編碼晝面的實際複雜度之間的關係,目前畫面之 複雜度將能藉此預估。活動量係一種在晝面編碼後其位元 率和品質的量度嘗試,假設一整張晝面的全活動量A正比 於複雜度C,則 k A~kxC^ — xrxq — k]xr x a 2 ^ 其中,k係比例常數。若k’為活動量對複雜度之比 (activity-t〇- complexity ratio ,ACR),則全活動量A 還 可以下列式子表示: A = A CR xr x q 根據本發明,可以從相同型態的先前編碼晝面來預估A CR 。觀念上,先在編碼一張晝面前計算其全活動量A,然 後,即將被編碼的晝面之複雜度以如下方式估算: //(0 广⑴=Λ ^est - ACR{1-1) 其中,」(/)係晝面·Ρ(2〕的全活動量,-:4C/?(:!_1:I係同型態的先前編 碼晝面其活動量對複雜度之比。以下,在符號或變數中的 上標(i)表示該符號或變數係與即將被編碼的目前晝面有 關;同理,在符號或變數中的上標則與先前編碼晝面 相關。 估計的複雜度可用來為合適型態的晝面更新其複 雜度,所以三種型態的晝面其複雜度q、Cp和CB將分別儲存0702-7400TWf (Nl); 90P113; Lin.ptd Page 12 200428880 V. Description of the invention (8) According to the present invention, the current activity between the current day surface and the actual complexity of the previously coded day surface are analyzed in advance. Relationship, the complexity of the current picture will be able to estimate. Activity is a measure of its bit rate and quality after daytime coding. Assuming that the total activity A of a whole daytime surface is proportional to complexity C, then k A ~ kxC ^ — xrxq — k] xr xa 2 ^ Where k is the proportionality constant. If k 'is the activity-to-complexity ratio (ACR), the full activity A can also be expressed by the following formula: A = A CR xr xq According to the present invention, the same type Previously coded day to day to estimate A CR. Conceptually, calculate the full activity A before coding a day, and then the complexity of the day surface to be coded is estimated as follows: // (0 广 ⑴ = Λ ^ est-ACR {1-1) Among them, "(/) is the total activity of the day surface · P (2),-: 4C /? (:! _ 1: I is the ratio of the activity amount to complexity of the previously coded day surface of the same type. Below, The superscript (i) in a symbol or variable indicates that the symbol or variable is related to the current day surface to be encoded; similarly, the superscript in a symbol or variable is related to the previously coded day surface. Estimated complexity It can be used to update the complexity of the appropriate type of diurnal surface, so the complexity of the three types of diurnal surfaces q, Cp, and CB will be stored separately

0702-740.0TWf(Nl) ; 90P113 ; Lm.ptd 第13頁 200428880 五、發明說明(9) 以利後續之處理。一組連續畫面之瞬間複雜度 (instantaneous complexity)亦以如下方式更新: INST_C = Nj x + Np x Cp + N3 x C5 其中,Ni、NP和NB分別是該組連續晝面中I -、P -及B -晝面之 數目,而且此處所指的一組連續畫面至少包含一組GOP。 一旦瞬間複雜度·^ST_C 已經決定,則目前晝面户〕之目標位 元額度為 ΤΒ(ϋ ρ·⑴ ——-—xR^ INST C ’ 其中,Reii係一組連續晝面的有效位元用量。由上式可知, 目標位元額度係與複雜度⑵成正比,而與瞬間複雜度 JJVST_C成反比。 所有的MPEG-2資料位元流均必須遵循MPEG-2標準的視 訊緩衝器檢驗者(V i d e ο B u f f e r V e r i f i e r,V B V )規則,分 配的目標額度必須受限以使V B V緩衝器不會滿溢(o v e r f 1〇w )或匱乏(u n d e r f 1 o w)。原本只有在編碼後才能得到目前晝 面之平均量化位階,則能於目標位元額度決定之後 予以估計,根據本發明,目前晝面Ρα〕其預估的平均量化位 階如下: =___ XB^xACR^ 其中,这2係當做平均量化位階之初始值。藉助這個平均0702-740.0TWf (Nl); 90P113; Lm.ptd page 13 200428880 V. Description of the invention (9) To facilitate subsequent processing. The instantaneous complexity of a group of continuous pictures is also updated in the following way: INST_C = Nj x + Np x Cp + N3 x C5 where Ni, NP, and NB are I-, P- And B-the number of diurnal planes, and the set of consecutive pictures referred to here contains at least one set of GOPs. Once the instantaneous complexity ^ ST_C has been determined, the target bit quota of the current day-time household] is TB (Β ρ · ⑴ ——- xR ^ INST C ', where Reii is a set of effective day-time effective bits The amount of usage. According to the above formula, the target bit quota is directly proportional to the complexity and inversely proportional to the instantaneous complexity JJVST_C. All MPEG-2 data bit streams must follow the MPEG-2 standard video buffer checker (V ide ο Buffer Verifier, VBV) rule, the allocated target quota must be limited so that the VBV buffer will not overflow (overf 10w) or be scarce (underf 1 ow). It was originally only available after encoding. The current average quantization level of the daytime plane can be estimated after the target bit quota is determined. According to the present invention, the current average quantization level of the current daytime plane Pα] is as follows: = ___ XB ^ xACR ^ where 2 are taken as The initial value of the average quantization level.

III 0702-7400TWf(Nl) ; 90P113 ; Lin.ptd 第14頁 200428880 五、發明說明(10) 量化位階之初始值忍,視訊編碼哭 挪·)下,輸出較佳的晝面品質。^匕在查既定的位元額度 度及量化位階初始值決定好後,便可利^面之目標位兀額 影像編碼器,像是MPEG-2標準描述的 =多不同型態的III 0702-7400TWf (Nl); 90P113; Lin.ptd Page 14 200428880 V. Description of the invention (10) Under the initial value of the quantization level, the video code is crying.), Output better daylight quality. ^ After checking the determined bit quota and the initial value of the quantization level, you can use the target location. Image encoder, such as described by the MPEG-2 standard = many different types of

Model 5),根據目標位元額度來;^型5 (Test 目前晝面在完成編碼之後,其i;:的;縮。 ^ 貝1不的複雜度將可得 ,而目刖旦面的全活動量和實際複雜度 灰^〕可從以下計算得到 間的關係,Model 5), according to the target bit quota; ^ Type 5 (Test is currently completed after the coding, i ;: of; contraction. ^ The complexity of 1 will be available, and The relationship between the amount of activity and the actual complexity can be obtained from the following calculations,

AQ(i) XAQ (i) X

ACR ;:U與全活動成正比’而與目前畫面的實際 消耗位兀數UB(o以及平均量化位階犯(0 、不 八十w & 白匕成反比。此⑺可 手來預估下一張同型態的晝面之複雜度。^妒〕 ACT〔2-1〕做線性結合來避免受到那歧富 17 太恭昍从口口 w 丨一田3亦隹矾晝面的影響。 本1月的早一人(31叫161&33)視訊編碼可藉由 較佳實施例並配合第4圖之操作流程圖來解釋。日如第3 = 不,動態影像視訊編碼器30 〇包括一移動估算單元3 :二 複雜度估計器3 2 0、一配額分派器3 3 〇、— :-量化位階估計麵以及一影像編碼』更 二收視訊序列裡的一組連續晝面中即將被編碼二 目則旦面π ,用來在執行移動估算的區塊匹配時, 目前晝面’)之全活動量妒(步驟S410)。基於全活動量: 以及這組連續晝面中同型態之先前編碼畫面其,複ACR;: U is directly proportional to the full activity 'and is inversely proportional to the actual number of consumed frames UB (o and the average quantizer (0, not eighty w & white dagger). This can be used to estimate the next picture by hand The complexity of the diurnal surface of the same type. ^ Jealousy] ACT [2-1] makes a linear combination to avoid being affected by the richness 17 Too respectful from the mouth w 丨 Yitian 3 is also affected by the diurnal surface. Ben 1 The video coding of the early person of the month (31 called 161 & 33) can be explained by the preferred embodiment and the operation flow chart of Fig. 4. If the day is 3 = No, the moving image video encoder 30 includes a motion estimation. Unit 3: two complexity estimators 3 2 0, a quota dispatcher 3 3 0,-:-quantization level estimation surface and an image coding "more than two sets of continuous day planes in the video sequence will soon be encoded. Then, the surface π is used to perform the current activity of the daytime surface when performing block matching of the motion estimation (step S410). Based on the full activity volume: and the previous encoding of the same type in this set of continuous daylight surfaces Screen its, complex

200428880 五、發明說明(11) 雜度估計器3 2 0可用來估計目前晝面之複雜度巧(步驟S42〇 )°配額分派器3 30以預估的複雜度⑵更新本組連續晝面其 瞬間複雜度/滞—c ,ϋ絲擄c%及〇分配目標位 元額度給目前畫面#(步驟S4 30 )。用來編碼目前晝面 _之平均量化位階初始值忠,“量化位階估計器35〇以 全活動量,〗、目標位元額度以及活動量 為基礎來決定(步驟S440)。在本實施例中像編 碼器3 6 0基於初始值忠以適應性量化方式編碼目前晝面 押,並且在完成後,回報實際消耗位元數迎(〇以及平Γ均量 化位階却(0給參數更新器34 0 (步驟S45 0 )。以全活動量 ⑽、實際的消耗位元數㈤《和實際的平均量化位階乂泛0為 基礎’參數更新器340為目前晝面ρ«計算其活動量對複雜 度之比乂cf)(步驟S 4 6 0 )。 3在等,上,第3圖之較佳實施例可考量以硬體以及/或 是軟體來實現。根據本發明,移動估算單元3丨〇和 的其他元件可以管線(Pipeline)模式運作,在複 二 器3 2 0開始計#目前晝面的複雜度之前,移動估嘗、隹,人: 必須先完成即將被編碼的目前晝面其全活 二早—凡3 10 移動向量之估算,並且,當移動估算單元3ι〇里之叶算以及 作準備時,複雜度估計器3 2 0和其他元件仍正^下張晝面 有與目前晝面相關的運作。接下來將詳細地於完成—所 例中的每個元件。 t車父佳實施200428880 V. Description of the invention (11) The complexity estimator 3 2 0 can be used to estimate the complexity of the current day (step S42). The quota dispatcher 3 30 updates the continuous day and time of this group with the estimated complexity. Instantaneous complexity / hysteresis, c, ϋc% and 0 allocate target bit quota to the current picture # (step S4 30). It is used to encode the initial value of the average quantization level of the current diurnal plane. "The quantization level estimator 35 is determined on the basis of the full activity amount, the target bit quota, and the activity amount (step S440). In this embodiment, Like the encoder 3 6 0 based on the initial value, adaptively quantify the current day betting, and after completion, return the actual number of bits consumed (0 and the average quantization level but 0 (0 to the parameter updater 34 0). (Step S45 0). Based on the full activity amount, the actual number of consumed bits, and the actual average quantization level, pan 0, the parameter updater 340 calculates the amount of activity versus complexity for the current day (乂 cf) (step S 460). 3 Wait, above, the preferred embodiment of FIG. 3 can be implemented by hardware and / or software. According to the present invention, the mobile estimation unit 3 and 0 Other components can be operated in pipeline mode. Before the second device 3 2 0 starts to calculate the current complexity of the day, move estimates, sigh, people: You must first complete the full activity of the current day that will be coded. Erzao—Estimate of the 3 10 moving vector, and When the mobile estimation unit calculates and prepares for 3 mile, the complexity estimator 3 2 0 and other components are still working under the current day surface. The next step will be completed in detail— Each element in the example.

200428880200428880

一、移動估算單元310的主要目的之一為決定用何種預測 模式來編碼一張晝面裡的每個Μ β,如果必要的話,亦進行 向鈾和向後的移動預測,還可從區塊匹配運算中提取書面 /舌動里的資訊。首先,計算每個Μ Β之自身活動量(丨n t η — activUy),將一個〇中4個8以亮度值區塊之像素強度以 Ym,n’m-〇,···,7,η = 0,···,7來表示,並且每個8x8區塊其 平均值為?,則每個MB之内活動量IntraAct ·· 31. One of the main purposes of the motion estimation unit 310 is to decide what prediction mode to use to encode each M β in a daytime plane. If necessary, it also performs forward uranium and backward motion prediction. Extracting information from written / tongue movements during matching operations. First, calculate the amount of activity of each MB (丨 nt η — activUy), and set the pixel intensity of 4 of 8 in a block with a luminance value as Ym, n'm-0, ..., 7, η = 0, ..., 7, and the average value of each 8x8 block? , Then the amount of activity within each MB IntraAct ·· 3

MroAct =k 免=〇 其中 V ??ϊ=0?2=43 倘若需要較低的計算複雜度,可以相對的絕對差值代 替: 3MroAct = k == 〇 Where V ?? ϊ = 0? 2 = 43 If a lower computational complexity is required, the relative absolute difference can be substituted: 3

IntrcL^ct = 其中 ^-ςςΚ,--ψ 7?Ί=0?2=0 由於I-晝面中的MB僅有一種模式·· I模式,因此IntraAct 即為I -畫面的每個Μ B之活動量。 如果晝面為Ρ-或Β-畫面,則需進行移動估算。最常使IntrcL ^ ct = where ^ -ςςK, --ψ 7? Ί = 0? 2 = 0 Because MB in I-day has only one mode ... I mode, so IntraAct is every MB of I-picture Of activity. If the daytime picture is a P- or B-picture, then motion estimation is required. Most often

200428880 五、發明說明(13) 用來發現最佳移動向量之技術係區塊匹配。對非自身編碼 (non in tra coding)而言(如P-和B-晝面),利用將失真標 準如變異數或失真絕對值和減至最低,來選擇向前、向 後、雙向的預測或不需移動補償。一旦P—或B—晝面中每個 MB之MB模式決定,在每個移動補償的差值Μβ中4個8χ8區塊 之變異數亦可求得,移動補償的差值ΜΒ係待處理〇和預估 MB間像素對像素之差;失真絕對值和通常具有較佳的計算 效率’因此可用來取代k異數。將4個8 X 8區塊之變異數或 失真絕對值和相加以求得非内編碼晝面中每個M B之相互活 動量(inter-activity) ’ In terAct,然後把非内編碼晝面 中每個MB之IntraAct及其InterAct拿來作比較,以判斷 InterAct是否較小’若是’則以interAct作為該Μβ之活動 量,並以相互模式(inter-mode)編碼該MB ;否則以 IntraAct作為該MB之活動量,並以自身模式(intra —㈣心) 編碼該MB。最後’對目前的:[―、p—或^晝面,將所有㈣之 活動量相加而得全活動量,,移動估算單元31〇再把全活 動量,#送給複雜度估計器3 2 0、參數更新器34〇以及量化 位階估計器3 5 0。 接著’複雜度估計器3 2 0為某型態的目前晝面〆〕估i 其複雜度/且依照三種晝面型態,引進加權係數至預估: 複雜度β 。因為絕不能以Β-晝面預估其他的晝面,故可 減少Β-晝面之加權係數以分配較少的位元給^一晝面而保留 較多的位元給I-和Ρ-畫面;一般而言,編碼卜畫面會產生200428880 V. Description of the Invention (13) The technique used to find the best motion vector is block matching. For non in tra coding (such as P- and B-day), choose to use forward, backward, or bidirectional prediction or No motion compensation is required. Once the MB mode of each MB in the P- or B-day is determined, the variation number of 4 8 × 8 blocks in the difference Mβ of each motion compensation can also be obtained. The difference MB of the motion compensation is pending. The pixel-to-pixel difference between the estimated and MB; the absolute value of distortion and usually have better computational efficiency 'so it can be used to replace the k-different. Add the sum of the absolute values of the variations or distortions of the four 8 X 8 blocks to obtain the inter-activity of each MB in the non-encoded diurnal surface. The IntraAct and InterAct of each MB are compared to determine whether the InterAct is smaller. If it is, then the interAct is used as the Mβ activity amount, and the MB is coded in inter-mode; otherwise, the IntraAct is used as the The amount of MB activity and encodes the MB in intra mode. Finally, for the current: [―, p—, or ^ diurnal surface, add all the activity amounts of ㈣ to get the full activity amount, and the motion estimation unit 310 then sends the full activity amount, # to the complexity estimator 2 0, parameter updater 34 0 and quantization level estimator 3 5 0. Next, the 'complexity estimator 3 2 0 is the current diurnal plane of a certain type.] Estimating the complexity of i / and according to the three diurnal types, introduce a weighting coefficient to the estimation: complexity β. Since it is impossible to estimate the other day planes with the B-day plane, the weighting coefficient of the B-day plane can be reduced to allocate fewer bits to the ^ -day plane while retaining more bits to I- and P- Picture; in general, encoding a picture will produce

0702-7400TWf(Nl) ; 90P113 ; Lin.ptd 第18頁 200428880 五、發明說明(14) 最多的位元,因此P -晝面之加權係數又小於I -畫面之加權 係數。複雜度估計器3 2 0根據目前晝面的型態更新複雜度 q、CP或CB三者其中之 方式估算: ί/(ΐ-畫面) e/m/(P-晝面) else z/(B-畫面) 乂1 0) ACR\ .4' 目前晝面之複雜度以如下 -1; (00702-7400TWf (Nl); 90P113; Lin.ptd Page 18 200428880 V. Description of the invention (14) The most bits, so the weighting coefficient of P-day is smaller than the weighting coefficient of I-picture. The complexity estimator 3 2 0 updates the complexity q, CP, or CB based on the current type of the daytime surface to estimate one of three ways: ί / (ΐ-picture) e / m / (P-daytime) else z / ( B-picture) 乂 1 0) ACR \ .4 'At present, the complexity of the daytime surface is as follows -1; (0

ACRACR

0-D 乂 | (0 ACR(rl) 其中,Hf—1; 4〇邱〜及係一組連續晝面中,I、p 和B型態的先前編碼晝面各自的活動量對複雜度之比。這 些nf 〃 、乂和up存放在參數更新器3 4 〇之中, 而複雜度估計器3 2 0會為適當型態的先前晝面讀取對應的 活動量對複雜度之比。L、KP及KB分別係I -、P -和B -晝面之 加權係數,其範圍一般是在0到1 . 0之間。至於I -晝面,較 佳實施例可以採用1.0 。 當收到之時,配額分派器3 3 0為該組連續晝面更新0-D 乂 | (0 ACR (rl) where Hf-1; 4〇 Qiu ~ and a series of successive day-planes, each of the previously coded day-planes of the I, p, and B types have their respective activity to complexity These nf 〃, 乂, and up are stored in the parameter updater 3 4 0, and the complexity estimator 3 2 0 will read the corresponding amount of activity to the complexity of the previous diurnal surface of the appropriate type. L , KP, and KB are the weighting coefficients of I-, P-, and B-days, respectively, and their ranges are generally between 0 and 1.0. As for I-days, the preferred embodiment can use 1.0. When received At this time, the quota dispatcher 3 3 0 continuously updates the group daily.

0702-7400TWf(Nl) ; 90P113 ; Lm.ptd 第19頁 200428880 五、發明說明(15) 其瞬間複雜度迅切二〇且分配目標位元額度給目前晝面 Pco ,瞬間複雜度迅「ST—C之更新如下:0702-7400TWf (Nl); 90P113; Lm.ptd Page 19 200428880 V. Description of the invention (15) Its instantaneous complexity is quickly cut to 20 and the target bit quota is allocated to the current daytime Pco. The instantaneous complexity is "ST- The update of C is as follows:

S^T_C = NT xCj + Np X Cp + Nb X CB 其中,%、NP和Νβ分別是本組連續晝面中I-、P-及B-晝面之 數目。再者,目前晝面之目標位元額度:Γ5(2)為 TB(lJ =——^—X —xi? II'IST^C f 其中,n係本組連續晝面之晝面數,f係每秒晝面數,即: 圖幀率(f r am e r a t e ).,R則係每組連續晝面之期望平均位 元率。然而,實際的消耗位元數並不會和期望的位元額度 剛好相等,因此發展一種回授策略來使畫面實際消耗的位 元數接近於目標位元額度,在較佳實施例中,編碼至目前 為止的超用位元量,可由正在編碼的目前晝面攤還一部 份:S ^ T_C = NT xCj + Np X Cp + Nb X CB where%, NP, and Nβ are the numbers of I-, P-, and B- diurnal planes in the continuous diurnal plane of this group, respectively. In addition, the current target bit quota of the day: Γ5 (2) is TB (lJ = —— ^ — X —xi? II'IST ^ C f, where n is the number of consecutive day and day in this group, f Is the number of diurnal planes per second, that is: the frame rate (fr am erate). R is the expected average bit rate of each group of continuous diurnal planes. However, the actual number of consumed bits does not match the expected bit rate. The quotas are just equal, so develop a feedback strategy to make the number of bits actually consumed by the picture close to the target quota. In a preferred embodiment, the amount of super-used bits encoded to date can be calculated from the current day being encoded. Part of the noodles:

TB{0 = TB(i)-ARxEB 其中,EB係參數更新器3 4 0傳來的超用位元量,AR則係既 定之攤還率,其範圍一般是在0 . 0 5到0 . 2之間。配額分派 器3 3 0須調整目標位元額度.以符合VBV規範,所以還定 下額度的上限(ϋ —bound)及下限(L_b〇und)。就固定位元率 (constant bitrate’CBR)的操作而言,分配給一張晝面 的目標位元額度須使VBV缓衝器不會滿溢或匱乏,因此目 標位元額度烈⑺限制在上、下限範圍内:TB {0 = TB (i) -ARxEB Among them, the EB is the overused bit amount transmitted by the parameter updater 3 4 0, and AR is the predetermined amortization rate, and its range is generally from 0.5 to 0. Between 2. The quota dispatcher 3 3 0 must adjust the target bit quota to comply with the VBV specification, so it also sets the upper limit (ϋ —bound) and lower limit (L_bund) of the quota. For constant bitrate'CBR operations, the target bit quota allocated to a daytime plane must be such that the VBV buffer will not overflow or be scarce, so the target bit quota is strongly limited to Within the lower limit:

0702-7400TWf(Nl) ; 90P113 ; Lm.ptd 第 2◦頁 2004288800702-7400TWf (Nl); 90P113; Lm.ptd page 2 ◦ 200428880

五、發明說明(16) if (Γ3ω > u —bound) then 了召⑴=U —bound if (< L —bound) then tb(i) =L_bound 若是可變位元率(v ar i ab 1 e b i t r a t e,VBR)的操作,則口、 要防止VBV匱乏即可,故: 、^ if (rs〔2) > u —bound) then 了S0.)二U —bound 然後,目標位元額度rs⑴會被傳送到參數更新器3 4 0、量化 位階估計器3 5 0以及影像編碼器3 6 〇。 一旦目標位元額度to⑴決定好之後,量化位階估計器 350即可為目前晝面p(o預估其編碼所需的平均量化位階初 始值⑶。用來量化一特定型態的目前晝面p(2〕之初始值 以如下方式估算: = 乂⑺ — T^xACR^·^ e/w i/(P-畫面)V. Description of the invention (16) if (Γ3ω > u —bound) then ⑴ = U —bound if (< L —bound) then tb (i) = L_bound if it is a variable bit rate (v ar i ab 1 ebitrate, VBR) operation, you can prevent VBV scarcity, so: ^ if (rs [2) > u —bound) then S0.) Two U —bound Then, the target bit quota rs⑴ Will be transmitted to the parameter updater 3 4 0, the quantization level estimator 3 5 0, and the image encoder 3 6 0. Once the target bit quota to⑴ is determined, the quantization level estimator 350 can estimate the initial initial value of the average quantization level ⑶ required for its encoding for the current day plane p (o. It is used to quantify a specific type of current day plane p (2) The initial value is estimated as follows: = 乂 ⑺ — T ^ xACR ^ · ^ e / wi / (P-picture)

T^xACR^^ p //(B-畫面)T ^ xACR ^^ p // (B-picture)

0702-7400TWf(Nl);90P113;Lin.ptd 第21頁 200428880 五、發明說明(17) ⑦一 , 一 T^^xACR(i~^0702-7400TWf (Nl); 90P113; Lin.ptd page 21 200428880 V. Description of the invention (17) First, one T ^^ xACR (i ~ ^

B 接著將平均量化位階初始值您傳給影像編碼器3 6 〇。藉由 對晝面中每一空間區域其量化位階的適應性改變,本實施 例的視訊編碼器3〇〇採用虛擬緩衝器(virtual buf f er)以 提供位元率控制的回授機制,由於虛擬緩衝器及適應性量 化對熟悉此技藝者而言,乃為習知技術,因此不再做贅述 。虛擬緩衝斋的利用狀況(〇 C c U P a n c y )控制了量化位階, 從而控制了位元率,然而,虛擬緩衝器必須指定一個初始 的利用狀況方能運作,並且虛擬緩衝器的利用狀況也反應 了用來編碼母一空間區域的量化位階,所以虛擬緩衝器的 利用狀況不僅控制了量化位階的初始值,亦在給定位元率 的條件下,控制了視覺品質。 在初始值您的幫助之下,影像編碼器3 6 0可以讓視訊 序列中的晝面品質更為一致。透過訊號線312,影像編碼 為3 6 0從移動估算單元31〇接收畫面資料以及每個帅之移動 =量和MB模式。為了在給定位元額度瓜⑴的情形下,儘可 月匕勺減/可見的失真,影像編碼器3 6 0依據平均量化位階 來決定編碼目前晝面的每侧利之量化位階, &二貝料為基礎,影像編碼器3 6 0開始對目前晝面〆〕進 :::且輸出編碼後的資料CD。在目前晝面押編碼之後, 广、,扁碼斋3 6 0將量化位階做平均並且計算實際消耗位元 數成),再回報目前晝面的L妒〕以及平均量化位階犯(〇給來B then passes the initial value of the average quantization level to the image encoder 36. By adaptively changing the quantization level of each spatial region in the daylight, the video encoder 300 of this embodiment uses a virtual buffer (virtual buffer) to provide a bit rate controlled feedback mechanism. Virtual buffers and adaptive quantification are familiar to those skilled in the art, so they will not be repeated here. The utilization status of the virtual buffer (OCC UP ancy) controls the quantization level, thereby controlling the bit rate. However, the virtual buffer must specify an initial utilization status before it can operate, and the utilization status of the virtual buffer also reflects In order to encode the quantization level of the mother-spatial region, the utilization of the virtual buffer not only controls the initial value of the quantization level, but also controls the visual quality under the condition of giving a localization rate. With the help of the initial values, the image encoder 360 can make the quality of the day and night in the video sequence more consistent. Through the signal line 312, the image is coded 3 to 60 to receive the picture data from the motion estimation unit 31, as well as the amount of movement and the MB mode of each handsome. In order to reduce the amount of visible distortion in the case of locating the amount of positioning elements, the image encoder 3 60 determines the quantization level of each side of the current day and day based on the average quantization level. &Amp; Based on the data, the image encoder 360 starts to enter the current day and time] :: and outputs the encoded data CD. After the current day-day betting coding, Guangzhou, Bianmao Zhai averages the quantization levels and calculates the actual number of bits consumed), and then returns the current day-time jealousy] and the average quantization level offender (0 to come

200428880 五、發明說明(18) 數更新器340。 晝面P(z)之目標位元額度ra(2〕以及實際消耗位元數你(〇 的差距,會由參數更新器34 0做累計以便在晝面〆〕編石馬 後得到超用位元量EB : ^ < EB - EB X (1 - ) + UB(i) - TB(i) 其中,AR係既定之攤還率。因此,活動量和實際複雜声 間的關係,乂,可從以下計算得到 又 ACR^ =--^pr ,4gu) X UB(2) 此可用來預估同型態的下張晝面之複雜度。為使 乂Ci?(2)不會受到富含雜訊晝面的影響,較佳實施例利用了 和乂CR:(2-1)的線性組合,例如: ί/(Ι_畫面) ACR(ji} = ACR(/a) x (l - CW)~l· ACR(0 x CIV else i/(P-畫面) ACR(;} = ACR^a) x (l - CW)+ ACR1'0 x CW else i/(B-畫面)200428880 V. Description of the invention (18) Number updater 340. The target bit quota ra (2) of the day surface P (z) and the actual number of bits you consume (the gap of 〇 will be accumulated by the parameter updater 34 0 so that the day horse surface) will get the super-used bit Elementary EB: ^ < EB-EB X (1-) + UB (i)-TB (i) where AR is the predetermined rate of return. Therefore, the relationship between the amount of activity and the actual complex sound is awkward. From the following calculation, ACR ^ =-^ pr, 4gu) X UB (2) can be used to estimate the complexity of the next diurnal surface of the same type. In order that 乂 Ci? (2) will not be enriched The effect of noise day and day, the preferred embodiment uses a linear combination of 乂 CR: (2-1), for example: ί / (Ι_Picture) ACR (ji) = ACR (/ a) x (l-CW ) ~ l · ACR (0 x CIV else i / (P-picture) ACR (;) = ACR ^ a) x (l-CW) + ACR1'0 x CW else i / (B-picture)

ACR(^} = ACR(J~l) x (l - CIV)+ ACR(i) x CW 其中,CW係線性組合既定之加權係數。更新過的ACRf;、 及〃會被傳送到複雜度估計器3 2 〇以及量化位階 200428880 五、發明說明(19) 計器3 5 0,以便為下一張適當型態的畫面分別估算其複雜 度和量化位階之初始值,此外,超用位元量EB則送至配額 分派器3 3 0作為位元分配之用。 雖然本發明已以一具體實施例揭露如上,然其僅為了 易於說明本發明之技術内容,而並非將本發明狹義地限定 於該實施例,任何熟習此技藝者,在不脫離本發明之精神 和範圍内,當可作些許之更動與潤飾,因此本發明之保護 範圍當視後附之申請專利範圍所界定者為準。ACR (^) = ACR (J ~ l) x (l-CIV) + ACR (i) x CW where CW is a predetermined weighting coefficient for the linear combination. The updated ACRf; and 〃 are transmitted to the complexity estimate Device 3 2 0 and quantization level 200428880 V. Description of the invention (19) The device 3 50 is used to estimate the initial value of the complexity and the quantization level for the next appropriate type of screen. In addition, the bit amount EB is overused. Then it is sent to the quota dispatcher 3 3 0 for bit allocation. Although the present invention has been disclosed as above with a specific embodiment, it is only for easy explanation of the technical content of the present invention, and is not limited to the present invention in a narrow sense. In the embodiment, anyone skilled in the art can make some modifications and retouching without departing from the spirit and scope of the present invention. Therefore, the protection scope of the present invention shall be determined by the scope of the attached patent application.

0702-7400TWf(Nl) ; 90P113 ; Lm.ptd 第24頁 200428880 圖式簡單說明 第1圖係以顯示的順序呈現典型的晝面群組(GO P); 第2圖係Μ P E G之大區塊, 第3圖係本發明較佳實施例的位元率控制裝置之方塊 不意圖,以及 第4圖係本發明之操作流程圖。 符號說明 3 0 0〜動態影像視訊編碼器; 3 1 0〜移動估算單元; 3 1 2〜訊號線; 3 2 0〜複雜度估計器; 3 3 0〜配額分派器; 3 4 0〜參數更新器; 3 5 0〜量化位階估計器; 3 6 0〜影像編碼器; Ρ⑴〜晝面; 乂(2〕〜全活動量; 巧〜預估複雜度; 〜活動量對複雜度之比; ΕΒ〜超用位元量; 〜目標位元額度; 〜平均量化位階; 〜實際消耗位元數;0702-7400TWf (Nl); 90P113; Lm.ptd Page 24 200428880 Brief description of the diagram Figure 1 shows a typical daytime group (GO P) in the order shown; Figure 2 is a large block of M PEG FIG. 3 is a block diagram of a bit rate control device according to a preferred embodiment of the present invention, and FIG. 4 is an operation flowchart of the present invention. Symbol description 3 0 0 ~ Motion video encoder; 3 1 0 ~ Motion estimation unit; 3 1 2 ~ Signal line; 3 2 0 ~ Complexity estimator; 3 3 0 ~ Quota dispatcher; 3 4 0 ~ Parameter update 3 0 0 ~ quantization level estimator; 3 6 0 ~ image encoder; ⑴ ~ day surface; 乂 (2) ~ full activity; Qiao ~ estimated complexity; ~ ratio of activity to complexity; ΕΒ ~ Over-used bit amount; ~ Target bit amount; ~ Average quantization level; ~ Actual number of bits consumed;

0702-7400TWf(Nl) ; 90P113 ; Lm.ptd 第25頁 2004288800702-7400TWf (Nl); 90P113; Lm.ptd page 25 200428880

0702-7400TWf(Nl) ; 90P113 ; Lm.ptd 第26頁0702-7400TWf (Nl); 90P113; Lm.ptd page 26

Claims (1)

200428880 六、申請專利範圍 1. 一種利用初始量化位階預估之動態影像壓縮位元率 控制方法,至少包含下列步驟: 為一即將被編碼的目前晝面而計算一全活動量,該目 前晝面在一視訊序列的一組連續晝面之中; 基於該目前晝面之該全活動量以及該組連續晝面中之 一同型態的先前編碼晝面其活動量對複雜度之比,估計該 目前晝面之一複雜度; 以該目前晝面預估的該複雜度更新該組連續晝面之一 瞬間複雜度; 依據該估計複雜度以及該瞬間複雜度,分配一目標位 元額度給該目前晝面; 基於該目前晝面之該全活動量、該目前晝面之該目標 位元額度以及該同型態的先前編碼晝面其活動量對複雜度 之比,決定用來編碼該目前晝面之一平均量化位階初始 值; 根據該平均量化位階初始值,對該視訊序列中之該目 前晝面進行編碼;以及 在編碼該目前晝面之後,根據該目前晝面之該全活動 量、該目前晝面之實際消耗位元數和該目前晝面之平均量 化位階,計算該目前晝面其活動量對複雜度之比; 其中,該目前晝面的該活動量對複雜度之比與該目前 晝面之該全活動量成正比,而與該目前畫面之實際消耗位 元數以及該目前晝面之平均量化位階成反比。 2. 如申請專利範圍第1項所述之動態影像壓縮位元率200428880 6. Scope of Patent Application 1. A method for controlling bit rate of dynamic image compression using initial quantization level estimation, including at least the following steps: Calculating a full activity amount for a current day surface to be encoded, the current day surface In a set of continuous diurnal planes of a video sequence; based on the ratio of the amount of activity to the complexity of the previously coded diurnal planes of the same type of continuous diurnal planes in the current diurnal plane, estimate the One of the current day-to-day complexity; update the instantaneous complexity of the set of consecutive day-to-day surfaces with the estimated complexity of the current day-to-day; assign a target bit quota to the based on the estimated complexity and the instantaneous complexity Current day; based on the total activity of the current day, the target bit quota of the current day, and the ratio of the amount of activity to the complexity of the previously coded day of the same type, a decision is made to encode the current day An initial value of the average quantization level of one of the diurnal planes; encoding the current diurnal plane in the video sequence according to the initial value of the average quantization level; and encoding the current diurnal plane After that, the ratio of the current day's activity to the complexity is calculated based on the full activity of the current day, the actual number of bits consumed by the current day, and the average quantization level of the current day. The ratio of the activity amount to the complexity of the current day surface is directly proportional to the total activity amount of the current day surface, and is inversely proportional to the actual number of bits consumed by the current picture and the average quantization level of the current day surface. 2. Bit rate of video compression as described in item 1 of the scope of patent application 0702-7400TWf(Nl) ; 90P113 ; Lm.ptd 第27頁 200428880 六、申請專利範圍 控制方法,更至少包含下列步驟: 執行上述目前晝面其活動量對複雜度之比和上述同型 態的先前編碼晝面其活動量對複雜度之比兩者間的一線性 組合運算。 3. 如申請專利範圍第1項所述之動態影像壓縮位元率 控制方法,其中計算上述目前晝面之上述全活動量係在上 述目前晝面進行移動估算時。 4. 如申請專利範圍第1項所述之動態影像壓縮位元率 控制方法,其中上述估計複雜度之步驟係依據一方程式進 行計算: 似 ACR 其中,Cest係上述目前晝面預估的上述複雜度,K係一既定 的加權值,其範圍在0到1之間,A係上述一組連續晝面中 的上述目前晝面之上述全活動量,ACR係上述同型態的先 前編碼晝面其活動量對複雜度之比。 5. 如申請專利範圍第1項所述之動態影像壓縮位元率 控制方法,其中上述分配的目標位元額度係與上述目前晝 面之上述複雜度成正比,而與上述一組連續晝面之上述瞬 間複雜度成反比。 6. 如申請專利範圍第1項所述之動態影像壓縮位元率 控制方法,其中上述平均量化位階初始值由下列函數決 定:0702-7400TWf (Nl); 90P113; Lm.ptd Page 27 200428880 6. The method for controlling the scope of the patent application, including at least the following steps: Perform the above-mentioned current day-to-day activity ratio to complexity ratio and the previous type of the previous type A linear combination of the ratio of the activity to the complexity of the coded day surface. 3. The bit rate control method for moving image compression as described in item 1 of the scope of patent application, wherein the above-mentioned full activity amount for calculating the current daytime surface is when the current daytime surface is subjected to motion estimation. 4. The bit rate control method for moving image compression as described in item 1 of the scope of patent application, wherein the above step of estimating the complexity is calculated according to a formula: like ACR, where Cest is the above-mentioned complexity of the current day-time estimation Degree, K is a predetermined weighted value, ranging from 0 to 1, A is the above-mentioned full activity of the current day in the above-mentioned group of continuous day, and ACR is the previously coded day of the same type The ratio of its activity to complexity. 5. The bit rate control method for moving image compression as described in item 1 of the scope of patent application, wherein the allocated target bit quota is directly proportional to the above-mentioned complexity of the current day surface, and is related to the above-mentioned group of continuous day surfaces The above-mentioned instantaneous complexity is inversely proportional. 6. The bit rate control method for moving image compression according to item 1 of the scope of patent application, wherein the initial value of the average quantization level is determined by the following function: 0702-7400TWf(Nl) ; 90P113 ; Lm.ptd 第28頁 200428880 六、申請專利範圍 ㈣ TBxACR 其中,MQest代表上述平均量化位階初始值,A係上述一組連 續晝面中的上述目前晝面之上述全活動量,TB係上述目前 晝面之上述目前晝面之上述目標位元額度,ACR係上述同 型態的先前編碼晝面其活動量對複雜度之比。 7. 如申請專利範圍第1項所述之動態影像壓縮位元率 控制方法,其中上述一組連續晝面至少包含一晝面群組 (group of picture,G0P),且該晝面群組符合MPEG視訊 標準。 8. —種利用初始量化位階預估之動態影像壓縮位元率 控制裝置,至少包含: 一移動估算單元,接收一視訊序列裡的一組連續晝面 中即將被編碼之一目前晝面,用來在執行移動估算的區塊 匹配時,計算該目前晝面之一全活動量; 一複雜度估計器,基於該目前晝面之該全活動量以及 該組連續晝面中一同型態的先前編碼晝面其活動量對複雜 度之比,用來估計該目前晝面之一複雜度; 一配額分派器,以該目前晝面預估的該複雜度更新該 組連續晝面之一瞬間複雜度,用來根據該估計複雜度以及 該瞬間複雜度,分配一目標位元額度給該目前晝面; 一量化位階估計器,以該目前晝面之該全活動量、該 目前晝面之該目標位元額度以及該同型態的先前編碼畫面 其活動量對複雜度之比為基礎,決定用來編碼該目前晝面0702-7400TWf (Nl); 90P113; Lm.ptd Page 28 200428880 6. Patent application scope x TBxACR, where MQest represents the initial value of the above average quantization level, and A is the above-mentioned current day-plane in the above-mentioned group of continuous day-planes For full activity, TB is the above-mentioned target bit quota for the current day and above, and ACR is the ratio of its activity to complexity for the previously coded day of the same type. 7. The bit rate control method for dynamic image compression according to item 1 of the scope of the patent application, wherein the above-mentioned group of continuous day-planes includes at least one group of pictures (G0P), and the day-plane groups conform to MPEG video standard. 8. — A dynamic image compression bit rate control device using initial quantization level estimation, including at least: a motion estimation unit that receives a current day surface that is about to be encoded in a group of continuous day surfaces in a video sequence, and uses Comes to calculate the full activity of one of the current diurnal planes when performing block matching of the motion estimation; a complexity estimator based on the full active amount of the current diurnal planes and the previous type of the same group of consecutive diurnal planes Encodes the ratio of the amount of activity to the complexity of the diurnal surface to estimate the complexity of the current diurnal surface; a quota dispatcher that updates the instantaneous complexity of one set of consecutive diurnal surfaces with the complexity estimated by the current diurnal surface Degree, used to allocate a target bit quota to the current day plane based on the estimated complexity and the instantaneous complexity; a quantized rank estimator, based on the full activity of the current day plane, the current day plane Based on the target bit quota and the activity-to-complexity ratio of the previously encoded picture of the same type, the decision is made to encode the current day 0702-7400TWf(Nl) ; 90P113 ; Lm.ptd 第29頁 200428880 六、申請專利範圍 之一平均量化位階初始值;以及 一參數更新器,以該目前晝面之該全活動量、該目前 晝面之實際消耗位元數和該目前晝面之平均量化位階為基 礎,計算該目前晝面其活動量對複雜度之比; 其中,該目前晝面的該活動量對複雜度之比與該目前 晝面之該全活動量成正比,而與該目前晝面消耗之位元數 以及該目前晝面之平均量化位階成反比。 9.如申請專利範圍第8項所述之動態影像壓縮位元率 控制裝置,其中上述參數更新器還執行上述目前晝面其活 動量對複雜度之比和上述同型態的先前編碼晝面其活動量 對複雜度之比兩者間的一線性組合運算,以減輕雜訊影 響。 1 0 .如申請專利範圍第8項所述之動態影像壓縮位元率 控制裝置,更至少包含一影像編碼器,根據上述平均量化 位階初始值來對上述目前晝面進行編碼,且將上述目前晝 面之實際消耗位元數以及上述目前晝面編碼所用之平均量 化位階回報給上述參數更新器。 1 1 .如申請專利範圍第8項所述之動態影像壓縮位元率 控制裝置,其中上述複雜度估計器所估計得到的上述目前 晝面之上述複雜度,係依據一方程式: 似 ACR 其中,Cest係上述目前晝面預估的上述複雜度,K係一既定0702-7400TWf (Nl); 90P113; Lm.ptd page 29 200428880 VI. One of the scope of patent application average initial quantization level initial value; and a parameter updater, based on the current activity of the current day surface, the current day surface Based on the actual number of consumed bits and the average quantization level of the current day, calculate the ratio of the current day's activity to the complexity; where the ratio of the current day's activity to the complexity and the current day The total amount of activity on the diurnal surface is directly proportional to the number of bits consumed by the current diurnal surface and inversely proportional to the average quantization level of the current diurnal surface. 9. The dynamic image compression bit rate control device according to item 8 of the scope of the patent application, wherein the parameter updater also executes the ratio of the current day's daytime activity to the complexity and the previously coded daytime surface of the same type. A linear combination of the ratio of activity to complexity to reduce noise. 10. The dynamic image compression bit rate control device described in item 8 of the scope of the patent application, further comprising at least one image encoder, which encodes the current day surface according to the initial value of the average quantization level, and The actual number of bits consumed by the diurnal plane and the average quantization level used by the current diurnal coding are reported to the parameter updater. 11. The bit rate control device for moving image compression as described in item 8 of the scope of patent application, wherein the above-mentioned complexity of the current daytime surface estimated by the above-mentioned complexity estimator is based on a formula: like ACR, Cest is the above-mentioned complexity of the current day-to-day estimates, and K is an established 0702-7400TWf(Nl) ; 90P113 ; Lm.ptd 第30頁 200428880 六、申請專利範圍 的加權值,其範圍在0到1之間,A係上述一組連續晝面中 的上述目前晝面之上述全活動量,A C R係上述同型態的先 前編碼晝面其活動量對複雜度之比。 1 2 .如申請專利範圍第8項所述之動態影像壓縮位元率 控制裝置,其中上述分配的目標位元額度係與上述目前晝 面之上述複雜度成正比,而與上述一組連續晝面之上述瞬 間複雜度成反比。 1 3 .如申請專利範圍第8項所述之動態影像壓縮位元率 控制裝置,其中上述量化位階估計器以下列函數決定上述 平均量化位階初始值: 其中,MQest代表上述平均量化位階初始值,A係上述一組連 續晝面中的上述目前晝面之上述全活動量,TB係上述目前 晝面之上述目前晝面之上述目標位元額度,ACR係上述同 型態的先前編碼晝面其活動量對複雜度之比。 1 4.如申請專利範圍第8項所述之動態影像壓縮位元率 控制裝置,其中上述一組連續晝面至少包含一晝面群組 (group of picture,G0P),且該晝面群組符合MPEG視訊 標準。0702-7400TWf (Nl); 90P113; Lm.ptd Page 30 200428880 VI. The weighted value of the patent application range, which ranges from 0 to 1, A is the above-mentioned current day-plane in the above-mentioned group of continuous day-planes Full activity, ACR is the ratio of the activity to complexity of the previously coded diurnal surface of the same type. 12. The bit-rate control device for moving image compression as described in item 8 of the scope of the patent application, wherein the allocated target bit quota is proportional to the above-mentioned complexity of the current day surface and is proportional to the above-mentioned group of consecutive day The above-mentioned instantaneous complexity is inversely proportional. 13. The dynamic image compression bit rate control device according to item 8 of the scope of the patent application, wherein the quantization level estimator determines the initial value of the average quantization level with the following function: wherein MQest represents the initial value of the average quantization level, A is the above-mentioned full activity of the above-mentioned current day-plane in the above-mentioned group of continuous day-planes, TB is the above-mentioned target bit quota of the above-mentioned current day-plane, and ACR is the previously-encoded day-plane of the same type Activity to complexity ratio. 14. The dynamic image compression bit rate control device according to item 8 of the scope of the patent application, wherein the above-mentioned group of continuous daylight planes includes at least a daylight plane group (GOP), and the daylight plane group Compliant with MPEG video standards. 0702-7400TWf(Nl) ; 90P113 ; Lin.ptd 第31頁0702-7400TWf (Nl); 90P113; Lin.ptd page 31
TW92115798A 2003-06-11 2003-06-11 Method and apparatus for moving picture compression rate control using bit allocation with initial quantization step size estimation at picture level TWI221739B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
TW92115798A TWI221739B (en) 2003-06-11 2003-06-11 Method and apparatus for moving picture compression rate control using bit allocation with initial quantization step size estimation at picture level

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
TW92115798A TWI221739B (en) 2003-06-11 2003-06-11 Method and apparatus for moving picture compression rate control using bit allocation with initial quantization step size estimation at picture level

Publications (2)

Publication Number Publication Date
TWI221739B TWI221739B (en) 2004-10-01
TW200428880A true TW200428880A (en) 2004-12-16

Family

ID=34389034

Family Applications (1)

Application Number Title Priority Date Filing Date
TW92115798A TWI221739B (en) 2003-06-11 2003-06-11 Method and apparatus for moving picture compression rate control using bit allocation with initial quantization step size estimation at picture level

Country Status (1)

Country Link
TW (1) TWI221739B (en)

Also Published As

Publication number Publication date
TWI221739B (en) 2004-10-01

Similar Documents

Publication Publication Date Title
CN104885455B (en) A kind of computer implemented method and device for Video coding
CN104854866B (en) The content-adaptive of next-generation video, characteristic compensation prediction
US10013746B2 (en) High dynamic range video tone mapping
TW404132B (en) A real-time single pass variable bit rate control strategy and encoder
CN103402099B (en) The method and system that the picture-level rate of Video coding controls
CN101390401B (en) Enhanced image/video quality through artifact evaluation
US9258568B2 (en) Quantization method and apparatus in encoding/decoding
TWI304297B (en) Video encoding methods and systems with frame-layer rate control
WO2019127136A1 (en) Bit rate control method and encoding device
US8406297B2 (en) System and method for bit-allocation in video coding
CN101584215B (en) Integrated spatial-temporal prediction
CN105900420A (en) Selection of motion vector precision
JP5173409B2 (en) Encoding device and moving image recording system provided with encoding device
KR20010021879A (en) Apparatus and method for macroblock based rate control in a coding system
KR100922510B1 (en) Image coding and decoding method, corresponding devices and applications
CN107431814A (en) The change of ρ domains speed control
KR20070028404A (en) Method of storing pictures in a memory using compression coding and cost function including power consumption
CN107155112A (en) A kind of compressed sensing method for processing video frequency for assuming prediction more
JP3872849B2 (en) Video encoding device
TWI247546B (en) A video encoding method which carries out the encoding of P frame or B frame by utilizing I frame
CN102187678A (en) Encoding method and encoding device for compression encoding of moving images
WO2022252222A1 (en) Encoding method and encoding device
US8111753B2 (en) Video encoding method and video encoder for improving performance
TW200428880A (en) Method and apparatus for moving picture compression rate control using bit allocation with initial quantization step size estimation at picture level
CN102238382B (en) Bit rate control method and device for image compression

Legal Events

Date Code Title Description
MM4A Annulment or lapse of patent due to non-payment of fees