TWI221739B - 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
TWI221739B
TWI221739B TW92115798A TW92115798A TWI221739B TW I221739 B TWI221739 B TW I221739B TW 92115798 A TW92115798 A TW 92115798A TW 92115798 A TW92115798 A TW 92115798A TW I221739 B TWI221739 B TW I221739B
Authority
TW
Taiwan
Prior art keywords
day
current
complexity
activity
picture
Prior art date
Application number
TW92115798A
Other languages
Chinese (zh)
Other versions
TW200428880A (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

狄、發明說明: [發明所屬之技術領域】 明係有關於資 如、 研%,衍別係指一藉请 ;、硯訊編碼系統以估計的初始量化 方法及裳置。 ^執仃位π率控制之 【先前技術】 眾所周知,電子通訊世界正經歷一次數位革命,以數 位表示資訊的主要優點在於能夠幾近無誤地儲存、再生、傳 收、處理及運用的資料位元流(bltstr咖)。舉例而言,峨 的=色視訊影像每秒有29. 97張晝面,每張晝面約彻條可 見掃瞄線,每條掃瞄線約需48〇點以紅、綠、藍三色呈現的 像素(Pixel),但若每種色彩成分以8位元編碼,則所產生 的位元率(bit rate)每秒約168百萬位元(Mbits/s),故各 種視訊格式其未經壓縮處理的位元率是非常高而不經濟,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 訊,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,包含以壓縮技術共通 之精髓為基礎的演算法,如:ιτυ-τ(前身係CCITT)建議書 H· 261 和 H. 263,以及 ISO/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(M〇ving 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(Internati〇nai standards Organization,ISO)及國際電子技術委員會(Ini:ernai:i〇nal 1221739D. Description of the invention: [Technical field to which the invention belongs] The Ming refers to information, research, and refer to a borrowing; the Xun code system uses the estimated initial quantification method and clothes. ^ [Previous technology] for controlling π rate control It is well known that the world of electronic communication is undergoing a digital revolution. The main advantage of representing information digitally is that it can store, reproduce, transmit, process and use data bits almost without error. Stream (bltstr coffee). For example, E = color video images have 29.97 day-to-day images per second. Each day-to-day image has approximately four visible scanning lines. Each scanning line requires approximately 48 o'clock in red, green, and blue. Pixels to be rendered, but if each color component is encoded in 8 bits, the resulting bit rate is about 168 million bits per second (Mbits / s), so various video formats are not The bit rate after compression processing is very high and uneconomical, so it is not suitable for most applications. "The integration of digital audio and video with computers, telecommunications networks, and consumer products has further stimulated the information revolution. The core of this revolution is digital compression technology for video and audio. Many compression standards include the common use of compression technology. Essence-based algorithms, such as: ιτυ-τ (formerly CCITT) Recommendations H.261 and H.263, and ISO / IEC MPEG-1, MPEG-2, and MPEG-4 standards. MPEG algorithms It was developed by the Moving Picture Experts Group (MPEG), which is an International Standards Organization (ISO) and the International Electronic Technology Commission (Ini: ernai: i〇nal 1221739

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殊的表現方式,這二俨 準規疋了壓縮位元流的語法(syntax)以及解碼的方法1作對 於編碼為所使用的演算法而言,卻為編碼的新穎性與多性 保留了相當多的自由度。 ^ 根據MPEG,一連串的視訊畫面(picture)分成_序列的 晝面群組(group of picture,G0P),其中每組G〇p以卜書 · 面開始,後面跟著P—晝面和B—晝面的安排,第工圖以顯示 的順序說明一組典型的G0P。丨―晝面之編碼毋須參考先前或 · 將來的晝面,P-畫面則參考連串的視訊畫面中在時間上最接 近的I-晝面或P-晝面來進行編碼,而B-晝面散佈於卜晝面 和P-畫面之間。並且利用先前的、將來的或兩者兼具的緊 · 鄰I-晝面和P-畫面來編碼。雖然好幾張B-畫面可以緊接著 · 連續出現’但絕不能以B-晝面預估其他的畫面。 母張畫面具有三種成分:亮度值(luminance,以Y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 φ 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(macr〇|3l〇ck,以MB表 示)。以4:2:0的視訊為例,每個MB包含一個Y成分16x16 取樣點的陣列,以及兩個Cr和Cb成分8x8取樣點的區塊, 其中Y成分16x16取樣點的陣列實際上由4個8x8取樣點的 區塊所組成。 8 1221739 、“、、扁碼益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8x8方塊均經 過離政餘弦轉換(discrete CC)sine加㈣。],㈣而形成 :8轉換係數陣列,轉換係數接著以一量化矩陣進行量化, 然後用z字形(zig_zag)掃瞄DCT係數的量化結果而得到一 、串的DCT係數,並且此DCT係數序列以可變長度碼 (variable length c〇de,VLC)進行編碼。p_晝面必須決定 將每個MB以!型MB或p型Μβ進行編碼,^型Μβ的編碼係 以j逑的方式為之,而對於每個?型mb,則需得到該抓以 先前晝面所做之預測,此預測係藉由一種移動向量(motlon vector^得’移動向量象徵著目前畫面中即將編碼的肋 及其在先前畫面中的預估Μβ之間的變動,預估Μβ與目前 MB間的預測誤差則以DCT、量化、ζ字形掃瞎以及孔 編碼。 處理Β-晝面時’必須決定以下列何種Μβ模式來編媽每 個MB I換式、F拉式、β模式以及FB模式。I模式係以肋 本身而不借助移動補償來編碼(如j㉟Μβ _般),· f模式為 單向的向前預測編碼’係以先前畫面得到移動補償的^估 (如P 3L MB般),反之’ β模式為單向的向後預測編瑪, 係以後來的晝面得到㈣補償的預估。特別的是FB模式, 其係雙向的預測編碼,運用向前的和向後的移動補償預估兩 者來做内插(她rpcaatiQn)而得到ρβ模式之移動補償預 估對F B和FB模式而言,預測誤差可以使用町、量化、 Z字形掃瞎以及VLC進行編碼。 、 視訊編碼器很重要的一點即為位元率的控制。位元率 、二制之主要目的係很有智慧地分配編碼每張畫面及其中每 们MB所用的位元數,使編碼過的視訊於解碼器進行解碼 月b彳胤了犯的&咼編竭視訊的視覺品質。編碼器必須為整 張旦面%取里化位階以便在給定的位元率下控制可見的失 真,然而,以選取的量化位階編碼一張晝面所用的實際位元 婁:必然疋在貫際編碼之後才會得知,現實中並不存在一個 史轉函數,能夠在給定希望達到的量化位階下,決定一張畫 面所用的實際位认。MPEG的關鍵特徵則係 = 是可細化方式,這種技術允許每張晝面中的不㈣; 以不同的程度編碼,而使每張畫面内及不同的畫面間達到平 均-致的視覺品質。不過,傳統的位元率控制方法卻相當複 雜逋吊需進订好幾次才能完成視訊編瑪的處理,除此之 外,先前的技術欠缺一種適岸枓吾 迥應性里化方式可以運用的簡單機 制,用於設定初始量化位階而讓畫面品質更為平均。 有鑑於此,…種新穎的位元率控制技術,可用於 早二人、即㈣視訊編碼器,再者,亦期待能提供 控制方法及裝置,於書面層級彳彳$ … —U彻估相㈣量化位階來進 盯位70分配壓縮動態影像。 【發明内容】 據本發明,每一張畫面其目標位元的分配,係美於 先則旦面的編碼結果和對即將被編碼的目前畫面^分 析的活動量,—旦目標位元分配好之後,則平立階的 1221739 初^值也之決定。藉由目前畫面之預先分析的活動量以及 & “,馬旦面〜貫際複雜度間的關係,目前晝面之複雜度將 Γ夠相知,這種預估的目前晝面複雜度對視訊編碼器 是非# β用的,可以更精確地分配目標位元額度給每張晝 面三再者1用上述的平均量化位階初始值,視訊編碼器在 既定的攸歧額條件下,可輸出較佳的晝面品質。 p 11明係針對於一種利用初始量化位階預估之動態影 、、宿位元率控制方法。首先為_即將被編碼的目前晝面來 :异其全活動量,其中該目前晝面係在一視訊序列的一組連 繽畫面〈中’接著基於目前畫面之全活動量以及這組連續晝 :中同型態的先前編碼晝面其活動量對複雜度之比,估計目 ^旦回之4雜度’並且以目前晝面預估的複雜度更新本組連 、、’、旦面之瞬間複雜度;目前畫面之目標位元額度的分配,係 依預=的複雜度以及瞬間複雜度而用來編碼目前晝面之 平均置化位階㈣值,則可根據目前晝面之全活動量、目標 位元額度以及同型悲的先前編碼晝面其活動量對複雜度之 比來決疋’因此’基於平均量化位階初始值,對視訊序列中 之目前晝面進行編碼。在目前晝面編碼之後,根據目前晝面 =全活動量、目前畫面之實際消耗位元數和目前晝面之平均 里化位δ十异目刖晝面其活動量對複雜度之比,就這點而 言,目前畫面的活動量對複雜度之比與目前畫面之全活動量 成正比而與目4畫面之貫際消耗位元數以及平均量化位階 成反比。 另方面本發明還揭露一種利用初始量化位階預估 11 ^221739 之動態影像昼縮位元率控 數更新器。移動估算單元接收視訊 食 即將被編碼的目前晝面,用來在執行移連績晝面中 時,計算此目前畫面之全活動量。基於目前書"'的^塊匹配 以及這組連讀晝面中同型1息面之全活動量 雜度之比,複靜料哭刊 馬旦面其活動量對複 面广_度,配額分派器更新本組連= 二Γ:,且根據預估的複雜度以及瞬間複雜度:: a己目才示位兀額度拾目i查;^ 刀 則旦面。用來編碼目前晝面之平始旦几 立階初始值,則由量化位階估計器以目前畫面之 1 ;標位元額度以及同型態料前編财面其活動量=複\、 Ϊ之比為基礎來決定。而參數更新器根據目前書面之=; :、二前晝面之實際消耗位元數和目前畫面之平均:化= 階’计异目前晝面其活動量對複雜度之比,盆中,目二查 =量對複雜度之比與目前晝面之全活動量成正:::: 則旦面之貫際消耗位元數以及平均量化位階成反比。” 【實施方式】 為使本發明之上述目的、特鮮m更明 下文特舉-較佳實施例,並配合所附圖式,作詳細說明,. 一開始,預測用的I-晝面和p_晝面必須先在 红 碼器内解碼,湘計算原始晝面及解碼的晝面間的^ (⑽t _n square)誤差’可以得到解碼畫面其品質的客: 12 1221739 里度。以均方根誤差作為經笔查 卜马解碼畫面之失真度,則解碼書面的 複雜度(c〇mpiex)-c:定義如下: 一 其H編碼該晝面所需的位元率,d係解碼的畫面之失 /、又丄理娜上平均的量化位階(quan1:izati〇n step size) 和失真度之間為;比例關係,故,晝面的複雜度可定義成: C ^ —~ί 2 _ ,中1係該畫面的平均量化位階。在晝面編碼之後,將可 得知位元率(實際消耗位元數)以及平均量化位階,而該畫面 的複雜度亦因此獲得。The Electrotechnical Commission (IEC) 's joint technical committee is committed to the development of video and audio compression and multi-channel representation. These two standards standardize the syntax of the compressed bit stream and the decoding method. As far as the algorithm used is concerned, coding has kept a considerable amount of freedom for the novelty and diversity of coding. ^ According to MPEG, a series of video pictures are divided into _sequence groups of pictures (G0P), where each group G0p starts with a book face, followed by P-day face and B-day In the above arrangement, the first drawing illustrates a typical set of GOPs in the order shown.丨 ―The encoding of the diurnal plane does not need to refer to the previous or future diurnal plane. The P-picture refers to the I-day or P-dimension plane that is closest in time to the serial video picture, and the B-day The surface is scattered between the Bu day surface and the P-picture. And use the previous, future, or both tightly adjacent I-day planes and P-pictures to encode. Although several B-pictures can be followed by consecutive appearances', other pictures cannot be estimated from the B-day. The master picture has three components: brightness (represented by Y), red color difference (represented by Cr), and blue color difference (represented by Cb). For the 4: 2: 0 format of MPEG φ, the sampling points of 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 block of an MPEG diurnal surface is a large block (macr0 | 3l0ck, expressed in MB). Taking 4: 2: 0 video as an example, each MB contains an array of 16x16 sampling points of the Y component and two blocks of 8x8 sampling points of the Cr and Cb components, where the array of 16x16 sampling points of the Y component is actually composed of 4 A block of 8x8 sampling points. The role of 8 1221739, ", and Bian Yiyi is to determine which picture encoding type and which prediction mode is the best. For each image, each 8x8 block in MB undergoes a discreet cosine transformation (discrete CC) sine plus ㈣.] To form an array of 8 conversion coefficients, and the conversion coefficients are then quantized by a quantization matrix, and then zigzag (zig_zag) is used to scan the quantization results of the DCT coefficients to obtain a series of DCT coefficients. And this DCT coefficient sequence is encoded with a variable length code (VLC). The p_day must decide to encode each MB with! -Type MB or p-type Μβ, and ^ -type Μβ is encoded with The method of j 逑 is this, and for each? -type mb, you need to get the prediction made by the previous day. This prediction is based on a motion vector (motlon vector ^ 得 'motion vector symbolizes the current picture The ribs to be encoded and the changes between their estimated Mβ in the previous picture, and the prediction error between the estimated Mβ and the current MB are coded with DCT, quantization, zeta-shaped literacy, and hole coding. Processing B-Day Surface Time ' Must decide which of the following Mβ modes to use We edit each MB I transform, F pull, β mode, and FB mode. I mode uses the rib itself to encode without motion compensation (like j㉟Μβ _), and f mode is one-way forward prediction coding 'Based on the previous picture's motion compensation estimation (like P 3L MB), otherwise' β mode is a one-way backward prediction editor, which is based on the prediction of the future compensation of the daytime. Especially the FB mode It is a two-way prediction coding. It uses both forward and backward motion compensation estimation to perform interpolation (her rpcaatiQn) to obtain motion compensation estimation of ρβ mode. For FB and FB modes, the prediction error can be used Map, quantization, zigzag literacy, and VLC encoding. The most important point of video encoder is the control of bit rate. The main purpose of bit rate and binary system is to intelligently allocate and encode each picture and its image. The number of bits used by each MB enables the encoded video to be decoded by the decoder. The video quality of the video has been exhausted. The encoder must take the rank of the entire image in order to reduce the visual quality. Control is possible at a given bit rate However, the actual bits used to encode a diurnal surface with the selected quantization level: it must be known only after transcoding, that there is no history transfer function in reality, which can be achieved in a given hope. The quantization level determines the actual recognition of a picture. The key feature of MPEG is that it is a thinnable method. This technology allows for the invariance of each day; encoding to different degrees, so that each The average and consistent visual quality is achieved in each frame and between different frames. However, the traditional bit rate control method is quite complicated. It needs to be ordered several times to complete the processing of video editing. In addition, the previous The technology lacks a simple mechanism that can be used to adapt the Ligularity of Ligularity to set the initial quantization level and make the picture quality more even. In view of this, ... a novel bit rate control technology can be used in the early two, that is, video encoders, and also look forward to providing control methods and devices at the written level. ㈣ Quantization level comes into bit 70 to distribute compressed video. [Summary of the Invention] According to the present invention, the allocation of the target bits of each picture is the result of the encoding of the first plane and the amount of active analysis of the current picture to be encoded ^, once the target bits are allocated After that, the initial value of 1221739 of the Heli order is also determined. Based on the pre-analyzed activity of the current picture and the relationship between "Ma Danian ~ inter-complexity, the current day-to-day complexity will be sufficiently known, and this estimated current day-to-day complexity is important for video The encoder is non- # β, which can more accurately allocate the target bit quota to each day and time. 1 Use the above-mentioned average quantization level initial value. Under the predetermined ambiguity conditions, the video encoder can output more The best daytime quality. P 11 is aimed at a dynamic shadowing and place rate control method that uses the initial quantization level estimation. The first is the current daytime to be coded: its full activity, where The current diurnal plane is a set of continuous images in a video sequence, “Zhong,” which is then based on the total activity of the current picture and this set of continuous diurnal: median isomorphic types of previously coded diurnal planes with their activity to complexity ratio. Estimate the projected complexity, and update the instantaneous complexity of this group with the current estimated complexity of the day and time; the allocation of the target bit quota of the current picture is based on the pre- = Complexity and instantaneous complexity To encode the current average level of the daytime plane, the value can be determined based on the current daytime plane's total activity, the target bit quota, and the previously coded daytime plane's activity to complexity ratio of the same type of sadness. Based on the initial value of the average quantization level, the current diurnal plane in the video sequence is encoded. After the current diurnal plane coding, the current diurnal plane = full activity, the actual number of bits consumed by the current picture, and the average diurnalization of the current diurnal plane. Bit δ ten different eyes: The ratio of the amount of activity to the complexity of the day. In this regard, the ratio of the amount of activity to the complexity of the current picture is proportional to the total amount of activity of the current picture and is consistent with the picture of the fourth picture. The number of consumed bits is inversely proportional to the average quantization level. On the other hand, the present invention also discloses an updater for daytime shrinkage bit rate control of a moving image that uses an initial quantization level to estimate 11 ^ 221739. The motion estimation unit receives video data that is about to be encoded. The current diurnal surface is used to calculate the total activity of the current screen when performing the continuous performance diurnal surface. Based on the current book " 'block matching and the same type of continuous reading diurnal surface 1 The ratio of the total activity volume and the complexity level. The activity volume of the complex static material is compared to the breadth of the complex surface. The quota dispatcher updates the group connection = two Γ :, and according to the estimated complexity and instantaneous complexity. ::: I have only shown the position of the eye to see the amount of i check; ^ The knife is the surface. It is used to encode the initial value of the first order of the current day, and then the quantization level estimator uses the current picture to 1; The bit quota and the amount of activity on the front of the financial platform of the same type are determined based on the ratio of the complex value = complex \, 复. The parameter updater is based on the currently written =;:, the actual number of bits consumed and the current The average of the picture: = = the level of the difference between the current day's activity and the complexity ratio, in the basin, the second check = the ratio of the amount to the complexity is positive with the current total activity of the day: ::: Zedan The number of consecutive consumed bits and the average quantization level are inversely proportional. "[Embodiment] In order to make the above-mentioned purpose of the present invention, the special m more clear, the following specific examples-preferred embodiments, and with the accompanying drawings, make In detail, at the beginning, the I-day and p_days used for prediction must be decoded in the red coder first. ^ (⑽t _n square) between the surface of the original daytime face day and decoding error 'which can be decoded picture quality passenger: 121221739 in degrees. Taking the root-mean-square error as the distortion of the Chama decoded picture, the written complexity (c0mpiex) -c: is defined as follows:-The bit rate required for H-encoding the daytime plane, d is The loss between the decoded picture and the average quantization level (quan1: izati〇n step size) and distortion is proportional to each other, so the complexity of the day surface can be defined as: C ^ — ~ ί 2 _, Middle 1 is the average quantization level of the picture. After the daytime coding, the bit rate (the number of bits actually consumed) and the average quantization level will be known, and the complexity of the picture will be obtained as a result.

^晝面之複雜度端視晝面的本質及其編碼的形式而定。 先刖技術利用最近編碼畫面之複雜度來表示目前晝面之複 雜度’並且為卜、p —及B_晝面分別保持其複雜度以減輕不 同編碼形式的影響,偏若視訊序列的畫面其内容為平順的變 化’則此方式可達到相當不錯的效果H由於畫面編碼 的目標位元轉與其實際消耗位元數m目符合,如果視 訊序列的内容大幅變動’上述方式並不適合用來達成—致的 根據本發明,減分析目前晝面的活動量(aetiv⑼ 以及先前編碼晝面的實際複雜度之間的關係,目前晝面之複 雜度將能藉此預估。活動量得、_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:比於複雜 度C,則 ’ A^kxC^^-xj^xq = kxxr x q 13 1221739 其中’ k係比例常數。若k,為活動量對複雜度之比 (activity-to-complexity rati〇,ACR),則全活動量 A 還 可以下列式子表示: 〇ί]=Ί^ 根據本發明,可以從相同型態的先前編碼畫面來預估ACR。 觀念上,先在編碼一張晝面前計算其全活動量A,然後,即 將被編碼的晝面尸w之複雜度以如下方式估算: 一、 J0) ^ 其m系晝面,的全活動量,似㈣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⑴表示該符號或變數係與即將被編碼的目前晝 同理’在符號或變數=上標(h)則與先前編瑪晝面相關。 估計的複雜度邮可用來為合適型態的晝面 、包 :度’所以三種型態的晝面其複雜度C/、。和::二 存以利後續之處理。一組連續書 .tant繼uscomplexitw以如;;方式^間後雜度^ The complexity of the day surface depends on the nature of the day surface and its encoding form. The prior art uses the complexity of the recently coded picture to represent the current complexity of the diurnal plane, and maintains its complexity separately to reduce the impact of different encoding forms. The content is smooth change ', then this method can achieve quite good results. H Since the target bit encoding of the picture is consistent with the actual number of bits it consumes, if the content of the video sequence changes significantly, the above method is not suitable for achieving — According to the present invention, the analysis of the current amount of diurnal activity (aetiv⑼) and the actual complexity of the previously encoded diurnal surface can be reduced, and the current complexity of the diurnal surface can be estimated from this. Attempt to measure the bit rate and quality after daytime encoding, assuming the full amount of activity A of a daytime surface: compared to complexity C, then 'A ^ kxC ^^-xj ^ xq = kxxr xq 13 1221739 where' k is a proportional constant. If k is the activity-to-complexity rati0 (ACR), then the total activity A can also be expressed by the following formula: 〇ί] = Ί ^ According to the present invention, Can be from the same type State the previous coded picture to estimate the ACR. Conceptually, first calculate the full activity amount A before coding a day, and then, the complexity of the daytime corpse w to be coded is estimated as follows: 1. J0) ^ The m is the total activity of the diurnal surface, which is the ratio of the activity to the complexity of the previously coded diurnal surface of the same type of actinide. In the following, the superscript ⑴ in a symbol or variable indicates that the symbol or variable is the same as the current day to be coded. 'The symbol or variable = superscript (h) is related to the previous edited day surface. The estimated complexity post can be used for the appropriate type of diurnal surface, including the degree: so the complexity of the three types of diurnal surface C / ,. And ::: Save for subsequent processing. A set of consecutive books .tant following uscomplexitw in such a way;

T-C,xC]+Np 乂 Cp+N0C 其中’I…和〜分別是該組連續畫面“ 之數目’而且此處所指的-組連續晝面至少包人_广晝面 -旦瞬間複雜度術—C已經決定,貝j 各;、、且咖。 元額度汉⑺為 |面之目標位 —_y J? INST 一 C eff 其中,~係一組連續晝面的有效位 目標位元額度仞(()係與複雜度成· °由上式可知, 而與瞬間複雜度 14 1221739 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(Video Buffer Verifier,vBV)規則,分配的目 標額度必須受限以使VBV緩衝器不會滿溢(〇vern〇w)或匱 乏(underflow)。原本只有在編碼後才能得到目前晝面戶⑺之 平均里化位階,則能於目標位元額度⑺⑴決定之後予以估 計,根據本發明,目前畫面/其預估的平均量化位階如下·· Q(0 = 一est TB^XACR{^ _ 其中,一如係當做平均量化位階之初始值。藉助這個平均量 化位階之初始值以丨,視訊編碼器能在既定的位元額度汉⑺ 下,輪出較佳的畫面品質。當目前晝面之目標位元額度及量 _ 化位階初始值決定好後,便可利用許多不同型態的影像編碼 杰,像是MPEG-2標準描述的測試原型5(Test M〇de;[ 5), 根據目標位元額度來完成晝面的壓縮。 目前晝面在完成編碼之後,其實際的複雜度將可得 知,而目前晝面的全活動量和實際複雜度之間的關係,及⑺ 可從以下計算得到 泰 ACR{i) 一 ^⑴ ~AQ{i) X UB{i) "中,與全活動量/)成正比,而與目前晝面的實際消 耗位元數汲《以及平均量化位階^以)成反比。此可拿 來預估下一張同型態的畫面之複雜度。jC/jW可以和』⑶㈣) 做線性結合來避免受到那些富含雜訊畫面的影響。 本發明的單次(single-pass)視訊編碼可藉由第3圖之 車乂佳貫施例並配合第4圖之操作流程圖來解釋。如第3圖所 15 1221739 示,動態影像視訊編碼器300包括一移動估算單元3ί〇、一 複,度估計器320、-配額分派器33()、_參數更新器⑽、 里化位階估計器350以及一影像編碼器36〇。移動估算單 j 3i0 =收視訊序列裡的一組連續晝面中即將被編碼:目 :畫面F,用來在執行移動估算的區塊匹配時,計算此目 前畫面_之全活動量,)(步驟S41 〇)。基於全活動量^)以 及這組連績晝面中同型態之先前編碼畫面其义❽叫) ^ ^ Λ ’、 ’複雜度TC, xC] + Np 乂 Cp + N0C where 'I ... and ~ are the number of continuous pictures in this group' and the -group of continuous daylight surfaces at least include people_ 广 日 面-旦C has decided, beijing each ;, and and coffee. Yuan quota Han ⑺ is the target position of the | — y J? INST a C eff where ~ is a set of effective digit target bit quotas of continuous day and time 仞 ((( ) Is related to complexity. ° can be known from the above formula, and is inversely proportional to the instantaneous complexity 14 1221739. All MPEG-2 data bit streams must comply with the MpEG_2 standard video buffer verifier (VBV) As a rule, the allocated target quota must be limited so that the VBV buffer does not overflow (0vern0w) or underflow. Originally, only the current average dailyization level of households can be obtained after encoding. It is estimated after the target bit quota ⑺⑴ is determined. According to the present invention, the current picture / its estimated average quantization level is as follows. Q (0 = one est TB ^ XACR {^ _ where, as if it were the average quantization level Initial value. With the initial value of this average quantization level,丨, the video encoder can turn out a better picture quality under the predetermined bit quota, and when the current day's target bit quota and amount_determination initial value is determined, many different types can be used State video coding, such as Test Prototype 5 (Test Mode; [5), described in the MPEG-2 standard, completes the compression of the daytime plane according to the target bit quota. At present, after the daytime plane is encoded, its actual The complexity will be known, and the relationship between the current day's total activity and the actual complexity, and ⑺ can be obtained from the following calculations: ACR {i) 一 ^ 一 ~ AQ {i) X UB {i) ", Is directly proportional to the full amount of activity /), and inversely proportional to the actual number of bits consumed on the current day, and the average quantization level ^)). This can be used to estimate the next picture of the same type Complexity. JC / jW can be linearly combined with 『⑶㈣) to avoid being affected by those noise-rich pictures. The single-pass video coding of the present invention can be implemented by the car in accordance with Figure 3. The example is explained in conjunction with the operation flow chart in Fig. 4. As shown in Fig. 15 1221739, the moving image The video encoder 300 includes a motion estimation unit 3o, a complex degree estimator 320, a quota dispatcher 33 (), a _parameter updater ⑽, a localization level estimator 350, and an image encoder 36. Single j 3i0 = will be encoded in a set of continuous day-to-day images in the video sequence: head: frame F, which is used to calculate the total activity of the current frame _ when performing block matching for motion estimation, (step S41 〇). Based on the full amount of activity ^) and the previously coded pictures of the same type in this group of consecutive days, the meaning is howl) ^ ^ Λ ',' complexity

叶益320可用來估計目前晝面之複雜度巧(步驟⑽)。 配額分派t二3〇以預估的複雜度更新本組連續書面立瞬 =—C,並爾^ 額度汉給目前晝面P⑺(步驟S430)。用來編碼目前主面俨 之平均量化位階初始值⑽,則由量化位階估計器咖以八 =量二目t標位元額度妒以及活動量物^ 為基礎來決定(步驟S440)。在本實施例中,影像編 =36G基於初始值㈤以適應性量化方式編碼目前畫面Ye Yi 320 can be used to estimate the complexity of the current day (step ⑽). The quota allocation t 230 is updated with the estimated complexity of the group's consecutive written instants = -C, and the quota is given to the current day P⑺ (step S430). The initial average value of the average quantization level 编码 used to encode the current principal plane ⑽ is determined by the quantization level estimator 八 based on the eight-quantity-two-t-bit quota and the amount of activity ^ (step S440). In this embodiment, the image coding = 36G is based on the initial value, and the current picture is encoded in an adaptive quantization manner.

务ίί完成後,回報實際消耗位元數⑽⑺以及平均量 二㈣处給參數更新器34〇(步驟S45〇)。以全活動量 只際的消耗位元數⑽(1)和實際的平均量化位階啦(0 = 器340為目前晝面 度之比况火(步驟S460)。 η在等放上帛3圖之較佳貫施例可寺量以硬體以及 =體來實現。根據本發明,移動估算單力31〇和第3圖中 以管線(Pipellne)模式… 開始計鼻目前畫面的複雜度之前,移動估算單元310必 16 1221739 須先完成即將被編碼的 動向量之估管,、,R 』旦命/、王活動里之計算以及移 備時,二二^ 前晝面相關的運作:接下=他7"件仍正忙於完成所有與目 個元件。 下來舲砰細地描述較佳實施例中的每 移動估算單开qin AA + & 模式來編碼味金 要目的之一為决定用何種預測 、:“碼一張晝面裡的每個mb,如 向前和向後的蒋翻苑、日丨w 受扪活,亦進仃 動量的資m i可從區塊匹配運算中提取畫面活 (lntra十百計算每個MB之自身活動量 此1Vity),將一個Μβ中4個8x8亮度值+ # 素強度以Ym,n,㈣ 7, _n 儿度值&塊之像 8x8 di# # ji ’二··’ n— , . _ .,7 來表示,並且每個 ”平均值為7;則每個㈣之内活動量intraAct: ^traAct^Yjjk 其中 k=0 降Uf (^7=0 77=〇 倘若需要較低的計算,雜度,可以相對^絕對差值代替:After the completion of the service, the actual number of consumed bits and the average amount are reported to the parameter updater 34 (step S45). The total number of consumed bits 全 (1) and the actual average quantization level (0 = device 340 is the current ratio of the daytime degree) (step S460). The preferred embodiment can be implemented in hardware and body. According to the present invention, the single force of motion estimation 31 and the pipeline (Pipellne) mode in Figure 3 ... Before starting to calculate the complexity of the current picture, move The estimation unit 310 must 16 1221739 to complete the estimation of the motion vectors to be coded, and the calculation of the R's life cycle, the activity in the king's activity, and the relocation, the operations related to the day-to-day plane: Continue = He 7 is still busy completing all the components. Let ’s describe in detail the per-movement estimation in the preferred embodiment. Open qin AA + & mode. One of the main purposes is to decide what to use. Prediction: "Code each mb in a daytime surface, such as forward and backward Jiang Fanyuan, day 丨 w subject to activity, and also the momentum of the asset mi can extract the picture activity from the block matching operation ( lntra calculates 100 MB of its own activity (1Vity), and will add 4 8x8 brightness values in one Mβ + # 素 强Degrees are represented by Ym, n, ㈣ 7, _n degrees & block image 8x8 di # # ji '二 ··' n—,. _., 7, and each "average is 7; then each IntraAct: ^ traAct ^ Yjjk where k = 0 and decrease Uf (^ 7 = 0 77 = 〇) If a lower calculation is required, the degree of heterogeneity can be replaced by the relative absolute value of ^:

IntraAct=Y^A 其中 k=0 w=Ow=:〇 模式’因此IntraAct 由於1一晝面中的mb僅有一種模式 即為卜晝面的每個MB之活動量。 17 1221739 (麵intra coding)而言(如p—和β_畫面),利 =變異數或失真絕對值和減至最低,來選擇向前^二 向的預測或不需移動補償。-旦…-晝面中每個Μβ: 二=動補償的差值Μβ中4物區塊之 -— W η , 真對值和通常具有較佳的計算 因此可用來取代變異數。將4個㈣區塊之變異 失心邑對值和相加以求得非内編碼畫面中每個仙之相互活IntraAct = Y ^ A where k = 0 w = Ow =: 〇 Mode ’Therefore IntraAct has only one mode for the mb in the diurnal surface, which is the activity amount of each MB in the diurnal surface. 17 1221739 (intra coding) (such as p- and β_ picture), the benefit = the number of mutations or the absolute value of distortion and minimized to choose forward ^ two-way prediction or no need for motion compensation. -Once ...-Each Mβ in the daytime surface: 2 = the difference between the four components in the dynamic compensation Mβ-W η, the true pair value usually has a better calculation, so it can be used instead of the number of mutations. Calculate the sum of the values of the four distorted blocks and add them to each other in the non-inner coded picture.

动:(:rr,lty),inierAct,然後把非内編碼書面中 母個 MB 之 IntraAr十 e 甘 T ‘ A ▲ ~ τ I—_較若?,= 叙旦、,、 右疋,則以inte^ct作為該ΜΒ之活 里亚以相互模式(inter_mode)編碼冑㈣;否則以 =traAct作為該MB之活動量,並以自身模式(intra-mode) 編瑪該MB。最後,對目前的卜、卜或卜畫面,將所有肋 之活=相加而得全活動量凡移動估算單元310再把全Move: (: rr, lty), inierAct, and then write the non-internal code in the MB of IntraAr 10 e Gan T ‘A ▲ ~ τ I — _? , = Syria, ,, and right, use inte ^ ct as the MB ’s living ria to encode in inter_mode; otherwise, use = traAct as the MB ’s activity and use its own mode (intra- mode) edit the MB. Finally, for the current Bu, Bu, or Bu picture, add all the cost of the activity = to get the full activity.

f動W料給複雜度估計器㈣、參數更新器340以及 1化位階估計器350。 接著,複雜度估計器320為某型態的目前畫面户(〇估瞀 产,且依照三種畫面型態,引進加權係數至預估的: ,、ς如。因為絕不能以B—畫面預估其他的畫面,故可減少 β—畫面之加權係數以分配較少的位元給Β-畫面而保留較多 的位兀給I-和ρ_畫面;一般而言,編碼卜畫面會產生最多 的位元’因此Ρ_晝面之加權係數又小朴晝面之加權係 數。複雜度估計器根據目前畫面的型態更新複雜度Q、 18 c 畫面户(0 '或G三者其中之一,目 姑算: 之·複雜度以如下方式 if (I-畫面) ACR(rl) e 1 se i f (p*~畫面) ci:!=Cp=KpXisr else if (B-畫面) ci:!=Cs=KBXisr y(i-l) C^^Cl=Klx-C-The data is fed to the complexity estimator ㈣, the parameter updater 340, and the rank estimator 350. Next, the complexity estimator 320 is the current picture user of a certain type (0 estimates the production, and according to the three picture types, introduces weighting coefficients to the estimated:,, 如. Because B-pictures must not be used to estimate For other pictures, the β-picture weighting coefficient can be reduced to allocate fewer bits to the B-picture and retain more bits to the I- and ρ_ pictures; generally speaking, encoding pictures will produce the most Bit 'so the weighting coefficient of P_day plane and the weighting coefficient of Xiaopu day plane. The complexity estimator updates the complexity Q, 18 c picture households (0' or G) according to the current picture type. Project calculation: The complexity is as follows if (I-picture) ACR (rl) e 1 se if (p * ~ picture) ci:! = Cp = KpXisr else if (B-picture) ci:! = Cs = KBXisr y (il) C ^^ Cl = Klx-C-

:中,蹲'考及蹲%'组連續畫面中小P ^和 B存放在翏數更新器340之中,而複 '旦、又估汁益320會為適當型態的先前晝面讀取對應的活動 :對複雜度之比。W嗜別係卜、p妙畫面之加 2數’其範圍一般是在0至,n 〇之間。至於卜畫面,較 1土 A方也例可以採用KI 丄.〇 〇 复_ Pt f到;配額分派器330為該組連續畫面更新 =間硬雜度術Γ—C且分配目標 前晝面 ,-間複雜度薦Γ一C之更新广又 矣今、N1、NP‘N \ Cb 之數目。再者_!Γ是本組連續晝面中im-畫面 η x —xi? ,刖重面P()之目標位元額度邡⑺為: In the squat 'examination and squat%' group of consecutive pictures, P ^ and B are stored in the number updater 340, and Fu 'dan, it is estimated that the juice benefit 320 will be the corresponding type of previous day-time reading correspondence. Activity: ratio of complexity. The number of W addi- tions, p, and the addition of 2 wonderful numbers' is generally in the range of 0 to, n 〇. As for the picture, the side A can also use KI 丄 .〇〇 复 _ Pt f to; the quota dispatcher 330 updates the continuous picture for this group = inter-hardness technique Γ-C and assigns the target day surface, The number of inter-complexity updates Γ-C, N1, N1, NP'N \ Cb. Furthermore, _! Γ is the im-picture η x —xi? In this group of continuous daytime planes, and the target bit quota 刖 of the heavy plane P () is

INST C 19 其中’ η係本組連續晝面之晝面數,f係每秒畫面數,即: 圖巾貞率(frame rate),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-種回授策略來使晝面實際消耗的位元數接 近於目^元減,在較佳實施例中,編碼至目前為止的超 用位元ϊ,可由正在編碼的目前晝面攤還一部份:INST C 19 where η is the number of continuous diurnal planes in this group, f is the number of frames per second, that is: frame rate, R is the expected average bit rate of continuous diurnal planes in each group. However, the actual number of bits consumed will not be exactly equal to the expected number of bits, so a feedback strategy is developed to make the number of bits actually consumed by the day close to the target minus. In the preferred embodiment, The super-used bits 编码 encoded so far can be partly amortized from the current daytime being encoded:

TB ⑴=TB ⑴-ARxEB 其中’ EB係蒼數更新器340傳來的超用位元量,AR則係既 定之攤還率,其範圍-般是在G.Q5到G.2之間。配额分派 | 器330須調整目標位元額度,以符合懦規範,所以還定 下額度的上限OLboimd)及下限(L—b_d)。就固定位元率 (constant Mt rate,CBR)的操作而言’分配給一張晝面的 目標位元額度須使VBV緩衝器不會滿溢或匱乏,因此目標位 元額度7SW限制在上、下限範圍内: if (7SW > u_bound) then Γ5⑺=u—bound if (7SW < L—bound) then ⑺=L—bound 若是可變位元率(variable bit rate,VBR)的操作,則只要 _ 防止VBV匱乏即可,故: if (7BW > u—bound) then ⑺=u—bound 然後,目標位元額度]會被傳送到參數更新器34〇、量化 位階估計器350以及影像編碼器36〇。 一旦目標位元額度沒⑴決定好之後,量化位階估計器 3 5 0即,可為目前畫面艸〉預估其編碼所需的平均量化位階初 始值这丨)。用來量化一特定型態的目前晝面之初始值 20 1221739 if (I-晝面)TB ⑴ = TB ⑴-ARxEB where ’EB is the amount of overused bits transmitted by the Cumulus Updater 340, and AR is a predetermined amortization rate, and its range is generally between G.Q5 and G.2. The quota allocation unit 330 must adjust the target bit quota to comply with the 懦 specification, so it also sets the upper limit OLboimd) and lower limit (L-b_d). As for the operation of constant bit rate (constant Mt rate, CBR), the target bit quota allocated to a daytime surface must be such that the VBV buffer will not overflow or be scarce, so the target bit quota 7SW is limited to above, Within the lower limit: if (7SW > u_bound) then Γ5⑺ = u—bound if (7SW < L_bound) then ⑺ = L_bound If it is a variable bit rate (VBR) operation, then only _ It is enough to prevent VBV shortage, so: if (7BW > u-bound) then ⑺ = u-bound Then, the target bit quota] will be sent to the parameter updater 34, the quantization level estimator 350, and the image encoder 36〇. Once the target bit quota has not been determined, the quantization level estimator 3 50 can estimate the initial initial value of the average quantization level needed for its encoding for the current picture (i). Initial value used to quantify the current day surface of a particular pattern 20 1221739 if (I-day surface)

Aii) TB{i) xACR°~l) 1 if (P-晝面) A⑴ TB{i) xACR(i~l) P if (B-畫面) A⑴ ί TB{i)xACR{i~l)Aii) TB {i) xACR ° ~ l) 1 if (P-day) A⑴ TB {i) xACR (i ~ l) P if (B-picture) A⑴ ί TB {i) xACR {i ~ l)

B 接者將平均量化位階初始值傳給影像編碼器36〇。藉由 為畫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300採用虛擬缓衝器(virtual buffer)以提供 位το率控制的回授機制,由於虛擬緩衝器及適應性量化對熟 悉亡技藝者而言,乃為f知技術,因此不再做贅述。虛擬= 衝益的利用狀況(〇CCUpanCy )控制了量化位階,從而押制了 然而,虛擬缓衝器必須指定一個初始的利用:況方 G區域亚二虛擬緩衝器的利用狀況也反應了用來編碼每-^ m 4里化位階’所以虛擬緩衝器的利用狀況不僅控制 覺品質。 兀在W疋位兀率的條件下,控制了視 在初始值必的幫助之下,影像編瑪 序列中的畫面品質更為—致。透、……ΐ9 Μ· MO從移動仕管-- 〇 、匕訊號線WE,影像編碼器 向量和m模;早=接收畫面資料以及每個MB之移動 能的減少可見定位元額度㈣的情形下,儘可 八如仏編碼器360依據平均量化位階初 21 1221739 始值〜邮來決定編碼目前晝面的每個肋所用之量化位階,以 這些資料為基礎,影像編碼器360開始對目前晝面艸]進行 編碼且輸出編碼後的資料CD。在目前晝面炉〉編碼之後,= 像編碼器刻將量化位階做平均並且計算實㈣耗位元^ ^,再回報目前畫面的劭w以及平均量化位階」給表數 更新器340。 / 晝面炉)之目標位元額度75(”以及實際消耗位元數汲(〇The B-connector transmits 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 picture, the video encoder 300 of this embodiment uses a virtual buffer to provide a feedback mechanism of bit το rate control. Adaptive quantification is a knowledge technique for those familiar with dead arts, so it will not be repeated here. Virtual = Utilization utilization status (〇CCUpanCy) controls the quantization level, thereby holding down. However, the virtual buffer must specify an initial utilization: the utilization status of the sub-second virtual buffer in the G area also reflects the The encoding level of every ^ m 4 is' so the utilization of the virtual buffer not only controls the perceived quality. Under the condition of the high bit rate, the image quality in the video editing sequence is more consistent with the help of the initial value. 、, ΐ9 Μ · MO from mobile management-〇, signal line WE, image encoder vector and m mode; early = receiving picture data and reducing the moving energy of each MB visible positioning unit amount Next, as far as possible, the encoder 360 determines the quantization level used to encode each rib of the current day based on the average value of the initial quantization level 21 1221739. Based on these data, the image encoder 360 starts to艸] to encode and output the encoded data CD. After the current day surface furnace> encoding, the encoder will average the quantization levels and calculate the actual consumption bits ^^, and then report the current picture's 劭 w and average quantization levels "to the table number updater 340. / 面面 炉) target bit quota of 75 ("and the actual number of bits consumed (0

間的差距,會由參數更新器34〇做累計以便在晝面户0編碼 之後得到超用位元量EB ·· EB^EBx(l- AR)+ UB(i) - TB(i) 其中,AR係既定之攤還率。因此,活動量和實際複雜度之 間的關係,/Ci?⑴,可從以下計算得到 //(〇 ACRU) =—4——-ΑΟω x UB{1) 此jCi?⑴可用來預估同型態的下張晝面之複雜度。為使」^(〇 不會受到富含雜訊晝面的影響,較佳實施例利用了 和The gap between them will be accumulated by the parameter updater 34 to obtain the overused bit quantity EB after the daytime household 0 encoding. EB ^ EBx (l- AR) + UB (i)-TB (i) where, AR is an established amortization rate. Therefore, the relationship between the amount of activity and the actual complexity, / Ci? ⑴, can be obtained from the following calculation // (〇ACRU) = —4 ——- ΑΟω x UB {1) This jCi? ⑴ can be used to estimate the same The complexity of the type of the next day surface. In order to prevent "^ (〇 from being affected by noise-rich daylight, the preferred embodiment uses and

dC/?(7的線性組合,例如: if (I-晝面) ACR^ = ACR(/-l) x (1 ^ CW)+ ACR(i) x CW else if (P-畫面) ACR(;} = ACR(;~l) x (1 - CW)+ ACR(i) x CW else if (B-畫面) ACR(J} = ACR(fX) x(1 ~ CW)+ ACR^ x CW 其中,CW係線性組合既定之加權係數。更新過的、 p及心會被傳送到複雜度估計器320以及量化位階 22 :計器35。,以便為下一張適當型態的晝 度和量化位階之初始值,此外, 开,、妓雜 此外超用位EB則送至配額 为派斋330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^兄明本相之技㈣容,而並非將本發明狭義地限定於 該貝施例 壬何$習此技蟄者,在不脫離本發明之精神和範dC /? (7 linear combination, for example: if (I-day) ACR ^ = ACR (/-l) x (1 ^ CW) + ACR (i) x CW else if (P-picture) ACR (; } = ACR (; ~ l) x (1-CW) + ACR (i) x CW else if (B-picture) ACR (J} = ACR (fX) x (1 ~ CW) + ACR ^ x CW where, CW is a linear combination of the established weighting coefficients. The updated p and xin are transmitted to the complexity estimator 320 and the quantization level 22: counter 35. In order to start the next day of the appropriate type and the quantization level Value, in addition, open, prostitutes and other super-used bits EB are sent to the quota for Paishai 330 as a bit allocation. Although the present invention has been disclosed as a specific embodiment above, it is only for easy The technical content of this phase is not limited to the present invention in a narrow sense to those who learn this technology, without departing from the spirit and scope of the present invention.

圍内$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。The $ within the range can be changed and retouched slightly. Therefore, the scope of protection of the present invention shall be determined by the scope of the attached patent application.

23 1221739 【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(GOP); 第2圖係MPEG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示意圖;以 及 第4圖係本發明之操作流程圖。 【符號說明】 300〜動態影像視訊編碼器; 310〜移動估算單元; 312〜訊號線; 320〜複雜度估計器; 330〜配額分派器; 340〜參數更新器; 350〜量化位階估計器; 360〜影像編碼器; P(7·)〜晝面; 〜全活動量; 〜預估複雜度; iC/?w〜活動量對複雜度之比; 观〜超用位元量; 〜目標位元額度; 〜平均量化位階; 〜實際消耗位元數; 24 1221739 〜平均量化位階初始值; ακ編碼後的資料。23 1221739 [Schematic description] Figure 1 shows a typical day group (GOP) in the order shown; Figure 2 is a large block of MPEG; Figure 3 is a bit of a preferred embodiment of the present invention A block diagram of the rate control device; and FIG. 4 is a flowchart of the operation of the present invention. [Symbol description] 300 ~ Motion video encoder; 310 ~ Motion estimation unit; 312 ~ Signal line; 320 ~ Complex estimator; 330 ~ Quota dispatcher; 340 ~ Parameter updater; 350 ~ Quantization level estimator; 360 ~ Image Encoder; P (7 ·) ~ Day Surface; ~ Full Activity; ~ Estimated Complexity; iC /? W ~ Activity Ratio to Complexity; Viewing ~ Over-utilized Bits; ~ Target Bits Quota; ~ average quantization level; ~ actual number of bits consumed; 24 1221739 ~ average value of average quantization level; ακ-coded data.

2525

Claims (1)

1221739 拾、申請專利範圍: 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·· 為即將被編碼的目前晝面而計算一全活動量,該目前書面 在一視訊序列的一組連續晝面之中; 基於該目前畫面之該全活動量以及該組連續畫面中之一同 型悲的先前編碼晝面其活動量對複雜度之比,估計該目前書 面之一複雜度; 面之一瞬間 以該目前晝面預估的該複雜度更新該組連續晝 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; 又 基於㈣前晝面之該全活動量、該目前晝面之該目標位元額 、及4同型悲的先别編碼晝面其活動量對複雜度之比,決 定用來編碼該目前畫面之一平均量化位階初始值;1221739 Scope of patent application: A method for controlling the bit rate of moving image compression using the initial quantization level estimation, including at least the following steps: Calculate a full activity for the current day and day to be encoded, which is currently written in a Among a set of continuous day-to-day faces of a video sequence; based on the ratio of the amount of activity to the complexity of the previously coded day-to-day face of the same type of sadness in the set of continuous pictures, estimate one of the current written Complexity; one of the faces updates the set of continuous day-to-day complexes with the complexity of the current day-to-day estimate; assigns a target bit quota to the current day-to-day face based on the estimated complexity and the instantaneous complexity; and Based on the ratio of the amount of activity to the complexity of the pre-day diurnal plane, the target bit amount of the current diurnal plane, and the 4 different types of a priori coded diurnal diurnal planes, it is decided to encode one of the current pictures Initial value of average quantization level; 根據該平均量化位階初始值,對該視訊序财之該目前晝面 進行編碼;以及 一 在編碼該目前晝面之後’根據該目前晝面之該全活動量、該 1前畫^之實際消耗位元數和該目前畫面之平均量化位 階,計算該目前晝面其活動量對複雜度之比; β亥目前晝面的該活動量對複雜度之比與該目前畫面之 量成正比,而與該目前畫面之實際消耗位元數以及 Μ目别旦面之平均量化位階成反比。 26 2. 、申明專利範圍第1項所述之動態影像壓縮位元率控制方 法,更至少包含下列步驟: 執订上述目前晝面其活動量對複雜度之比和上述同型態的 J、、扁碼晝面其活動量對複雜度之比兩者間的一線性組合 運算。 ' 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 /、中计异上述目前晝面之上述全活動量係在上述目前書 面進行移動估算時。 4 如由上主 •曱請專利範圍第丨項所述之動態影像壓縮位元率控制方 去’其中上述估計複雜度之步驟係依據一方程式進行計算: cest=Kx」L· 开 ACR 其中’ c…係上述目前畫面預估的上述複雜度,夂係一既定 的加權值,其範圍在0到1之間,」係上述一組連續畫面中 的上述目命晝面之上述全活動量,」Ci?係上述同型態的先前 編碼畫面其活動量對複雜度之比。 5·如申請專利範圍第i項所述之動態影像壓縮位元率控制方 去’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,而與上述一組連續畫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: MQest =—ά_ TBxACR 其中’ 代表上述平均量化位階初始值,j係上述一組連 續晝面中的上述目前晝面之上述全活動量,係上述目前畫 27 1221739 面之上述目前晝面之上述目標位元額度,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匕⑺耶 picture,GOP),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-目前晝面,用來在執行移動估算的區塊匹配時, 汁异該目前畫面之一全活動量; :複雜度估計H,基於該目前晝面之該全活動量以及該組連 續畫面中一同型態的先前編碼畫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; ㈣貝分派益’以該目前晝面預估的該複雜度更新該組連續 晝面之-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; -量化位階估計器,以該目前畫面之該全活動量、該目前主 ,之該目標位域度以及該同型態的先前編碼畫面其活動 二對複雜度之比為基礎,蚊用來編碼該目前畫面之 量化位階初始值;以及 二參數更新器,以該目前畫面之該全活動量、該目前書面之 貫際,肖耗位讀和該目前晝面之平均量化位 該目前晝面其活動量對複雜度之比;&為基礎,計算 28 1221739 目前晝面的絲動量對複雜度之轉該 ::面舌動量成正比’而與該目前晝面消耗之位元數以;該目 刖直面之平均量化位階成反比。 I Ϊ申:Γ利範圍第8項所述之動態影像壓縮位元率控制裝 雜声^上述參數更新$還執打上述目前晝面其活動量對複 "又t和上述同型態的先前編碼畫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。 又Encode the current daytime plane of the video sequence based on the initial value of the average quantization level; and after encoding the current daytime plane, according to the full activity of the current daytime plane, the actual consumption of the previous picture ^ The number of bits and the average quantization level of the current picture calculate the ratio of the amount of activity to the complexity of the current day. Β The ratio of the amount of activity to the complexity of the current day is proportional to the amount of the current picture. It is inversely proportional to the actual number of bits consumed in the current picture and the average quantization level of the M-membrane plane. 26 2. The method for controlling the bit rate of moving image compression as stated in item 1 of the patent scope, further includes at least the following steps: The above-mentioned current day-to-day activity ratio to complexity ratio and the same type of J, A linear combination of the ratio of the amount of activity to the complexity of the flat-day surface. '' As described in Item 1 of the scope of the patent application, the above-mentioned full-time activity of the current daytime plane is controlled by the motion image compression bit rate control method, and the above-mentioned current activity is at the time of the above-mentioned current paper for motion estimation. 4 As described by the above-mentioned owner, please request the bit rate control of the moving image compression as described in item 丨 of the patent scope, where the above steps of estimating the complexity are calculated according to a formula: cest = Kx "L · On ACR where" c ... is the above-mentioned complexity estimated in the above current picture, not a predetermined weighted value, which ranges from 0 to 1, "is the above-mentioned total activity amount of the above-mentioned day-to-day view in the above set of continuous pictures, "Ci? Is the ratio of the amount of activity to the complexity of the previously encoded pictures of the same type. 5. According to the motion image compression bit rate controller described in item i of the scope of the patent application, where the target bit quota allocated above is directly proportional to the above-mentioned complexity of the current daytime surface, and is related to the above-mentioned set of continuous pictures The above-mentioned instantaneous complexity is inversely proportional. 6. The method for controlling the bit rate of moving image compression according to item 1 of the scope of the patent application, wherein the initial value of the average quantization level is determined by the following function: MQest = —ά_ TBxACR where 'represents the initial value of the average quantization level, j is The total activity of the current day surface in the above set of continuous day surfaces is the target bit quota of the current day surface of the current picture 27 1221739, which is the previously encoded day surface activity of the same type. Ratio of complexity. 7. The bit rate control method for dynamic image compression as described in item 1 of the scope of patent application, wherein the above-mentioned group of continuous diurnal planes includes at least one diurnal plane group (picture (GOP)), and the diurnal plane group conforms 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 set of consecutive daylight planes in a video sequence to be encoded-the current daylight plane , Used to perform a motion estimation block matching, which is different from the full activity of the current picture;: complexity estimate H, based on the current activity of the current day and the same pattern in the group of continuous pictures The ratio of the amount of activity to the complexity of the previously coded picture is used to estimate one of the current day-to-day complexity; ㈣ 贝 分派 益 'updates the set of continuous day-to-day complexes with the complexity estimated from the current day-to-day. Degree, used to allocate a target bit quota to the current day surface based on the estimated complexity and the instantaneous complexity;-a quantized level estimator, based on the full activity of the current picture, the current master, and the target The bit field degree and the activity-to-complexity ratio of the previously encoded picture of the same type are used as a basis for the mosquito to encode the initial value of the quantization level of the current picture; and a two-parameter updater that uses the The total amount of activity in the previous screen, the current written date, the reading of the Shaw bit and the average quantization of the current day. The ratio of the current day's amount of activity to the complexity; based on & calculation 28 1221739 The degree of complexity of the diurnal silk momentum is: 'the tongue-momentum momentum is directly proportional to the number of bits consumed by the current diurnal surface; the average quantization level of the straight face of the objective is inversely proportional. I: Application of the motion image compression bit rate control described in item 8 of the Γ Lee range. The above-mentioned parameter update $ also performs the above-mentioned current day and its activity on the complex. A linear combination of the ratio of the activity to the complexity of the previously coded picture to reduce the effect of noise. also 爪如申請專利範圍第8項所述之動態影像壓縮位元率控制裳 置’更至少包含一影像編碼器,才艮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畫面編碼所用之平均量化位階回報 給上述參數更新器。 U.如申請專利範圍帛8 $所述之動態影像屢縮位力率控制震 置,其中上述複雜度估計器所估計得到的上述目前畫面之上 述複雜度,係依據一方程式: Cest = Kx —The dynamic image compression bit rate control device described in item 8 of the scope of the patent application further includes at least an image encoder, so as to encode the current day surface according to the initial value of the average quantization level, and The actual number of bits consumed by the current day and the average quantization level used in the current picture coding are reported to the parameter updater. U. As described in the scope of patent application 申请 8 $, the dynamic image repeated shrinkage rate control vibration setting, wherein the above complexity of the current picture estimated by the complexity estimator is based on a formula: Cest = Kx — ACR 其中,C…係上述目前晝面預估的上述複雜度,足係一既定 的加權值,其範圍在〇到丨之間,义係上述一組連續畫面中 的上述目前畫面之上述全活動量,係上述同型態的先前 編碼畫面其活動量對複雜度之比。 12··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,而與上述一組連續畫面之上述瞬間複雜度成 反比。 29 izzi/jy 13·如申請專利範圍第 置’其中上述量化位 位階初始值: 項所述之動態影像I縮位 階估計器以下列函數決定上述= MQest = A TBxACR =面7:代表上述平均量化位階初始值係上述-組連 ,旦、中;^述目前晝面之上述全活動量,π係上述目前畫 面之上述目刖畫面之上述目標位元額度,仏及係上述同型態 的先前編碼晝面其活動量對複雜度之比。 14·如申清專利範圍第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(gr〇up 〇『 picture ’ GOP),且該晝面群組符合mpeg視訊標準。ACR Among them, C ... is the above-mentioned complexity of the current day-to-day estimation, which is a predetermined weighting value, which ranges from 0 to 丨, meaning the above-mentioned full activity of the above-mentioned current picture in a set of continuous pictures. The quantity is the ratio of the activity quantity to the complexity of the previously coded picture of the same type. 12 ·· The dynamic image compression bit rate control device 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 and is proportional to the above-mentioned set of continuous pictures. The above-mentioned instantaneous complexity is inversely proportional. 29 izzi / jy 13. As set in the scope of the patent application, where the initial value of the quantization level mentioned above: The dynamic image I scale estimator described in the above item determines the above with the following function: MQest = A TBxACR = surface 7: represents the average quantization The initial value of the rank is the above-group connection, once, middle, and ^ is the above-mentioned total activity of the current day, π is the above-mentioned target bit quota of the above-mentioned target screen of the above-mentioned current picture, and is the previous previous type Coding the ratio of the amount of activity to the complexity of the day. 14. The dynamic image compression bit rate control device as described in item 8 of the scope of patent application, wherein the above-mentioned set of continuous day-planes includes at least one day-plane group (gr0up 〇 "picture 'GOP), and The day-to-day group meets the mpeg video standard. 3030
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 true TWI221739B (en) 2004-10-01
TW200428880A 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
TW200428880A (en) 2004-12-16

Similar Documents

Publication Publication Date Title
TW464831B (en) Image processing circuit and method for reducing a difference between pixel values across an image boundary
US10013746B2 (en) High dynamic range video tone mapping
CN105264888B (en) Coding strategy for adaptively switching to color space, color samples rate and/or bit-depth
CN105230023B (en) The adaptive switching of color space, color samples rate and/or bit-depth
CN102986211B (en) Speed control in Video coding
TW571589B (en) Method of performing video encoding rate control using bit budget
US20060133493A1 (en) Method and apparatus for encoding and decoding stereoscopic video
US7095784B2 (en) Method and apparatus for moving picture compression rate control using bit allocation with initial quantization step size estimation at picture level
TW201545545A (en) Projected interpolation prediction generation for next generation video coding
US8737485B2 (en) Video coding mode selection system
CN106031177A (en) Host encoder for hardware-accelerated video encoding
CN1965585B (en) Method of storing pictures in a memory
JPH0974565A (en) Controller for video signal coding system
CN101984665A (en) Video transmission quality evaluating method and system
US7373004B2 (en) Apparatus for constant quality rate control in video compression and target bit allocator thereof
JP2003517796A (en) How to reduce the "uneven picture" effect
JP3633204B2 (en) Signal encoding apparatus, signal encoding method, signal recording medium, and signal transmission method
TWI247546B (en) A video encoding method which carries out the encoding of P frame or B frame by utilizing I frame
Erdem et al. Compression of 10-bit video using the tools of MPEG-2
Wang et al. View synthesis distortion model based frame level rate control optimization for multiview depth video coding
US7804897B1 (en) Method for implementing an improved quantizer in a multimedia compression and encoding system
JP4670085B2 (en) Method for determining reference picture block in direct coding mode
TWI498005B (en) Methods for coding and decoding a block of picture data, devices for coding and decoding implementing said methods
TWI221739B (en) Method and apparatus for moving picture compression rate control using bit allocation with initial quantization step size estimation at picture level
US20110317758A1 (en) Image processing apparatus and method of processing image and video

Legal Events

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