經濟部中央樣準局負工消費合作社印製 A7 B7 五、發明説明（j ) 本發明係有關於影像之壓縮方法，特定而言，本發明係有關 於一種具適應性之靜態影像之壓縮方法，可以高壓縮比進行高品 質的影像壓縮儲存與再生處理。 發明之背景 習用技術之中，例如1994年1月4日所頒發的，H. Charavi等人 的美國專利第 5,276,525號，題為"Two-Dimensional Block Scanning for Subband Image and Video Coding”一案，其中所揭示的影像恩 縮方法’強調利用區頻编碼器(subband coder)進行不同頻率頻帶的 掃瞄技術，而未強調整體壓縮編碼的架構。該5,276,525號案中的 區頻編碼器將所要壓縮的影像分解成為四個不同的的區頻。此 外’該5,276,525號案中的编碼器，在其區頻编碼器將差異影像的 信號區分割成為不同區塊之後，仍將粗影像的信號保持為其原有 的格式。再者，5,276,525號案中，使用PCM量化器來將差異影像 的信號加以量化，而未使用純量量化器來進行量化。最後，該 5,276,525號案中的編碼器並未具有適應量化步驟能力。 習用技術之中，再加1993年10月12日所頒發的，H. Charavi等 人的美國專利第 5,253,058號，題為"Efficient Coding Scheme for Multilevel Video Transmission" —案所揭示的影像壓縮方法，利用 在目前的視訊框中與在先前一個視訊框之中所取得的像素之間的 差異視訊信號來減低計算量，但卻無法以容易的方式來達到一個 預定的影像信號壓縮比。此外，該5,253,058號案中，使用較為複 本紙張尺度適用中國國家橾準（CNS ) A4規格（210X297公釐） 1^--J------ .^-- (請先閲讀背面之注$項再填寫本頁) 訂 A7 B7 五、發明説明（>) 雜的混合式DCT/PCM编碼器來進行粗影像的编碼，並且也需要以 不同的掃描方法來掃描差異影像。 習用技術之中，又如1993年6月1日所頒發的，Sang H. Oh等 人的美國專利第5,216,719號’題為'Subband Coding Method and Eneoding/DecodingSystem"—案所揭示的影像壓縮方法，揭示了 一種區頻编碼的方法，將影像分解成為正交與對角兩個個方向的 部份，而非僅只是正交頻帶方向上的部份而已，並且亦未揭示在 區頻编碼之後再利用到的编碼器之架構。 習用技術之中’又如1993年4月13日所頒發的，p, Tourtier的 美國專利第5,220,760號’題為"Information Compression Method and Device for Compatible Decoding of a Group of Television Signals withlncreasingResolution”一案，其中所揭示的影像壓縮方法，強 調利用區頻編碼的技術應用於不同的用途之上。其方法使用不同 的空間尺度，而非利用相同的空間尺度來供包含HDTV以及一般 TV的用途利用。此外，該5,220,760號案中所揭示的内容亦不具備 適應不同的影像/視訊順序的不同特徵的能力。 經濟部中央樣準局員工消費合作社印製 n·— nn ϋ ml am· —1— n^i • 穿 、ve - - (請先閲讀背面之注f項再填寫本頁) 因此，本發明之一目的，係在於提供一種影像壓縮方法，可 以做為諸如JPEG影像資料壓縮程序的前處理或後處理影像壓縮之 用途，以提供更高影像資料壓縮的比例。 本發明之另一目的，係在於提供一種影像壓縮方法，並藉由 諸如JPEG的影像壓縮之一般處理程序加以改良，而提供更具效率 之高倍率磨縮方法。 本紙張尺度適用中國國家搮準（CNS ) A4規格（210X297公釐） A7 、發明説明（3 本發明之又一目的，係在於提供一種影像壓縮方法，可以處 理靜態影像的恩縮，適於在諸如多媒體相關用途之中提供高效率 的影像壓縮，以適於影像資料的傳輸與儲存之用途。 附圖之簡要説明 本發明之其他目的與特點，在本説明書中將配合附圖在後面 進行詳細説明，其中： 圖1 為依據本發明之影像壓縮方法一較佳實施例之方法架構 方塊示意圖； 圖2為本發明影像壓縮方法中三層式多重解析度影像分解之 區頻分佈示意圖；與 圖3 為依據本發明之影像小波多重解析度分解示意圖。 較佳實施例之詳細説明 -----------:I- * - (請先閲讀背面之注意事項再填寫本頁) 訂 經濟部中央標準局貝工消费合作社印製 首先參考圖1，其中顯示依據本發明之影像壓縮方法—較佳 實施例之方法架構方塊示意圖。本發明之影像壓縮方法架構係利 用小波轉換(wavelet transform)為基礎的一種分枝狀的分解法 (hierarchical decomposition scheme)，來將一幅影像分解成為-個 粗影像(coarse image)部份，以及數個差異(詳細）影像(difference images)部份。例如三層式之小波轉換，原影像被輸送至小波轉換 分解器1〇，分解成如圖2中所顯示的第三階的一個粗影像B1，以 及其餘的九個差異影像B2, B3,......，B10。 本紙張尺度適用中國國家揉準（CNS ) A4規格（210X297公釐） 經濟部中央標準局貝工消费合作社印裝 A7 _____________B7 五、發明説明（+ ) 圖2中的二層式影像分解是以B1為粗f彡像，與相當的影像分 解部份B2，B^’以及取__起，形成整個影像第三層的分解。相同 1 ’第二層分解部份’則是第三層影像雜與相#_像分解部 伤B5, B6 ’以及幻开〉成第二層的分解。第二層分解部份則再與相 當的影像分解部份B8，B9，與⑽形成整個影像第―廣的分解。 如同前面所述的’在本發明的此種影像分解架構之中，除了則被 界疋是為粗影像之外，其餘的九個分解部份，皆被歸為是屬詳細 ，差異影像群。利雌此種三層式的小波轉換她的方式，可進 仃較為簡化的二層式分解，亦可以進行更多層的，例如四層、五 層、甚或更多層的分解，端視需要而定。 小波轉換係利用一對的四次鏡式濾波器(Qjyfp，quadrature mirrorfilteCf^行的。其中，濾波器行為高通滤波器，而 則為低通濾波器，〇與Η合成起來是一個全通濾波器。囷3之中 顯示了依據本發明之影像小波多層解析度分解處理示意圖，其中 每一層的多層解析度影像分解，皆係利用小波轉換來進行。每一 次皆使用該層分紐之姉像(CF)，_τ—層_分解程序的 分解程序的輸入影像。經重覆進行三次的小波轉換，便可以完成 二廣的多重解析度影像分解的動作。 最後一層之粗影像Β1，實質上是為原始影像的一個縮小 (zoom out)影像，其尺寸在本發明此一實施例之中，係為原始影像 的水平與垂直方向尺寸的·二。粗影像之中，多含了孥, 夯像的最低頻率的部份，此部份的頻率，為人類視覺系統中所最 敏感的部份。在另一方面，差異影像群，則包含了影像之中的較 -----^--I--- 表| I I---訂 i {請先聞讀背面之注^^項再填寫本頁) 本紙張尺度適用中國國家橾率（CNS > A4规格（21 Ox297公釐） A7 :---__B7_____ 五、發明説明（5 ) 高頻率的部份，此部份的頻率，是人類視覺系統中較為不靈敏的 邵份。為了要減低計算的複雜性，以及保持處理程序影像的良好 °°質’粗影像以及差異影像群，係以不同的方法來進行壓縮，如 同囷1中所顯示的情形。 粗影像B1的壓縮编碼器是以DCT12，向量量化裝置(veet〇r quantizer) Vq 14，以及熵編碼器16為基礎所構成的。DCT區塊的 大小尺寸可以為4x4或8X8，依應用的用途而定。不過，對於一 個特定的影像而言，此DCT區塊的大小是固定。在由DCT輸出之 後，每一個的DCT係數皆是配合一個適應性量化表(AQT，adaptive quantization table)而加以量化的，此即向量量化裝置之操作。習知 技藝中之原始量化表(QT, quantization table )的數値是為固定的。 而本發明之適應性量化表AQT中的每一個量化數値，則是依據每 一個1)〇1'區塊數値，改良其在原始量化表QT中的對應數値，亦即 將QT中的每一個量化數値，乘以一個量化調變係數(QF， quantization factor)而獲得的。此量化調變係數是為一個整數，是 利用量化步驟計算器(QSC, quantization step calculator)計算而得 的。 經濟部中央標準局貝工消費合作杜印褽 1^1. - m m m —.1 I - - ------ I 1^1 n'-1 1^11^1 , . 匀 、T (請先閲讀背面之注意事項再填寫本頁) 在説明量化細計算器QSC_作之前，首先應界定粗影像 B1的AC能量與DC能量比如下： RB1 =——^~.DC (1)Printed by A7, B7, Central Consumers ’Cooperative Cooperative of the Ministry of Economic Affairs. 5. Description of the Invention (j) The invention relates to a method for compressing images. In particular, the invention relates to a method for compressing adaptive static images. , Can perform high-quality image compression storage and reproduction processing with high compression ratio. BACKGROUND OF THE INVENTION Among conventional techniques, for example, U.S. Patent No. 5,276,525 issued by H. Charavi et al. On January 4, 1994, entitled "Two-Dimensional Block Scanning for Subband Image and Video Coding", The disclosed image shrinking method 'emphasizes the use of a subband coder for scanning in different frequency bands, but does not emphasize the overall compression coding architecture. The zone frequency encoder in the 5,276,525 case will require The compressed image is decomposed into four different regional frequencies. In addition, the encoder in the 5,276,525 case, after the regional frequency encoder divided the signal area of the different images into different blocks, The signal remains in its original format. Furthermore, in the case No. 5,276,525, a PCM quantizer was used to quantize the signal of the difference image instead of using a scalar quantizer to quantize it. Finally, the code in the case No. 5,276,525 The device does not have the ability to adapt to the quantization step. Among conventional techniques, plus U.S. Patent No. 5,253,058, issued by H. Charavi et al. " Efficient Coding Scheme for Multilevel Video Transmission " — The image compression method disclosed in the case uses the difference in video signals between the current video frame and the pixels obtained in the previous video frame to reduce the amount of calculation, but However, a predetermined image signal compression ratio cannot be achieved in an easy way. In addition, in the case No. 5,253,058, the use of a relatively duplicate paper size is applicable to the Chinese National Standard (CNS) A4 specification (210X297 mm) 1 ^-J- -----. ^-(Please read the note on the back before filling in this page) Order A7 B7 V. Description of Invention (>) Hybrid hybrid DCT / PCM encoder to edit coarse image It also needs to use different scanning methods to scan the difference images. Among the conventional techniques, and as issued on June 1, 1993, US Patent No. 5,216,719 to Sang H. Oh et al. Is entitled 'Subband Coding' Method and Eneoding / DecodingSystem " —The image compression method disclosed in the case reveals a method of regional frequency coding, which decomposes the image into two parts in orthogonal and diagonal directions, instead of just just It is only in the direction of the cross-band direction, and it does not reveal the architecture of the encoder that is used after the regional frequency coding. Among the conventional techniques, 'as issued on April 13, 1993, p. Tourtier's U.S. Patent No. 5,220,760, titled " Information Compression Method and Device for Compatible Decoding of a Group of Television Signals withlncreasingResolution ", the image compression method disclosed therein emphasizes the use of zone frequency coding technology for different applications on. Its method uses different spatial scales instead of using the same spatial scale for applications including HDTV and general TV. In addition, the content disclosed in the case No. 5,220,760 does not have the ability to adapt to different features of different video / video sequences. Printed by the Consumer Cooperatives of the Central Procurement Bureau of the Ministry of Economic Affairs n · — nn ϋ ml am · —1— n ^ i • Wear, ve--(Please read the note f on the back before filling out this page) Therefore, the invention One purpose is to provide an image compression method that can be used for pre-processing or post-processing image compression such as a JPEG image data compression program to provide a higher compression ratio of image data. Another object of the present invention is to provide an image compression method, which is improved by a general processing program such as JPEG image compression, so as to provide a more efficient high magnification grinding method. This paper size is applicable to China National Standards (CNS) A4 (210X297 mm) A7, the description of the invention (3 Another object of the present invention is to provide an image compression method, which can handle the shrinkage of still images, suitable for Provide high-efficiency image compression for multimedia related applications, suitable for the transmission and storage of image data. Brief description of the drawings Other objects and features of the present invention will be described later in conjunction with the drawings in this specification In detail, of which: FIG. 1 is a schematic block diagram of a method architecture of a preferred embodiment of an image compression method according to the present invention; FIG. 2 is a schematic diagram of a region frequency distribution of a three-layer multi-resolution image decomposition in the image compression method of the present invention; and Figure 3 is a schematic diagram of multi-resolution decomposition of image wavelet in accordance with the present invention. Detailed description of the preferred embodiment -----------: I- *-(Please read the precautions on the back before filling this page ) Printed by the Central Standards Bureau of the Ministry of Economic Affairs, printed by the Bayong Consumer Cooperative, first refer to FIG. 1, which shows the image compression method according to the present invention—the method architecture method of the preferred embodiment. Block diagram. The image compression method architecture of the present invention uses a wavelet transform-based branching decomposition method to decompose an image into a coarse image section. Copies, and several difference (detail) image parts. For example, a three-layer wavelet transform, the original image is sent to a wavelet transform resolver 10, which is decomposed into a third-order as shown in Figure 2. One coarse image B1, and the remaining nine difference images B2, B3, ..., B10. This paper size is applicable to the Chinese National Standard (CNS) A4 specification (210X297 mm). Printed by a consumer cooperative A7 _____________B7 V. Description of the invention (+) The two-layer image decomposition in Figure 2 is based on B1 as a coarse f image, which is equivalent to the image decomposition parts B2, B ^ ', and __. The third layer of the entire image is decomposed. The same 1 'second layer of the decomposed part' is the third layer of image miscellaneous and phase #_image decomposition part hurts B5, B6 'and magic> into the second layer of decomposition. The second The layer decomposition part is equivalent to The image decomposition parts B8, B9, and ⑽ form the widest decomposition of the entire image. As described above, in the image decomposition structure of the present invention, except that the boundary is a coarse image, the rest The nine decomposed parts are all classified as detailed and different image groups. Li Li's three-layer wavelet transforms her method, which can be carried out into a more simplified two-layer decomposition, and more layers can be performed. , Such as the decomposition of four layers, five layers, or even more, depending on the needs. The wavelet transform system uses a pair of quadratic mirror filters (Qjyfp, quadrature mirrorfilteCf ^). Among them, the filter acts as a high-pass filter, but is a low-pass filter. 0 and Η are combined to form an all-pass filter. Figure 3 shows a schematic diagram of wavelet multi-layer resolution decomposition of the image according to the present invention, where the multi-layer resolution image decomposition of each layer is performed using wavelet transform. The input image of the decomposition program of the layer's sister (CF), _τ_layer_decomposition program is used every time. After repeated wavelet transformation three times, the two-resolution multi-resolution image decomposition can be completed. The coarse image B1 of the last layer is essentially a zoom out image of the original image. Its size in this embodiment of the present invention is the horizontal and vertical dimensions of the original image. In the coarse image, the lowest frequency part of the image is often included. The frequency of this part is the most sensitive part of the human visual system. On the other hand, the difference image group contains the comparison in the image ----- ^-I --- Table | I I --- Order i {Please read the note ^^ item on the back first (Fill in this page) The paper size is applicable to China's national standard (CNS > A4 specification (21 Ox297 mm) A7: ---__ B7_____ V. Description of the invention (5) High frequency part, the frequency of this part is The less sensitive Shao Fen in the human visual system. In order to reduce the computational complexity and maintain the good image quality of the processing program, the coarse image and the differential image group are compressed in different ways, as in Figure 1. The situation shown. The compression encoder for coarse image B1 is based on DCT12, vector quantizer Vq 14, and entropy encoder 16. The size of the DCT block can be 4x4 or 8X8, depending on the application. However, for a specific image, the size of this DCT block is fixed. After being output by DCT, each DCT coefficient is matched with an adaptive quantization table (AQT, adaptive quantization table) and this is the vector quantization device The number of original quantization tables (QT, quantization table) in the conventional art is fixed. Each quantization number in the adaptive quantization table AQT of the present invention is based on each 1). 1 'block number 値, which is obtained by improving its corresponding number 値 in the original quantization table QT, that is, multiplying each quantization Q in QT by a quantization modulation factor (QF). The quantization modulation coefficient is an integer and is calculated using a quantization step calculator (QSC). DuPont 褽 1 ^ 1.-Mmm —.1 I-------- I 1 ^ 1 n'-1 1 ^ 11 ^ 1,. Uniform, T (Please Please read the notes on the back before filling this page.) Before explaining the quantization calculator QSC_, you should first define the AC energy and DC energy of the coarse image B1 as follows: RB1 = —— ^ ~ .DC (1)
σ ^C + χ1 AC 其中RBI為計算QF之參考變數。 本紙張尺度適用中國國家標準（CNS ) ( 210X297公)t ) -- Α7 Β7 五、發明説明（6 ) cr2為DCT(Discrete Cosine Transform)區塊之變異數，· X為DCT(Discrete Cosine Transform)區塊之平均値； AC為DCT區塊之AC能量値； DC為DCT區塊之DC能量値； Ti値是做完DCT處理後之影像區塊内之AC數値，經過排序之 集合(set); 為將彩色影像分解成亮度（luminance)影像與色度 (chrominance)影像之亮度影像，經上述壓縮處理時之乃値； 7^為色度(chrominance)影像，經上述壓縮處理時之Ti値。 例如：一個做％2x2 DCT處理之影像區塊値為 1+2+3+4σ ^ C + χ1 AC where RBI is the reference variable for calculating QF. This paper size applies Chinese National Standard (CNS) (210X297) t)-Α7 Β7 V. Description of the invention (6) cr2 is the variation number of the DCT (Discrete Cosine Transform) block, and X is the DCT (Discrete Cosine Transform) Block average; AC is the AC energy of the DCT block; DC is the DC energy of the DCT block; Ti is the number of ACs in the image block after DCT processing, and the sorted set (set ); In order to decompose a color image into a luminance image and a luminance image of chrominance image, it is 値 when the above compression process is performed; 7 ^ is a chrominance image, Ti when it is subjected to the above compression process value. For example: an image block processed by% 2x2 DCT is 1 + 2 + 3 + 4
DcAc Ac Ac => 3 4 則X= 4DcAc Ac Ac = > 3 4 then X = 4
In ί *^^1 n I m —J ，^n -...... 1 n^i In---eJ (請先閲讀背面之注意事項再填寫本頁) σ 經濟部中央樣準局员工消費合作社印製 (\-Χ)2 + (2-Χ)2 ^ (3-Χ)2 + (Α-Χ)2 7 DC = l2 ; AC = 22+32+42 ；In ί * ^^ 1 n I m —J, ^ n -...... 1 n ^ i In --- eJ (Please read the notes on the back before filling this page) σ Central Bureau of Standards, Ministry of Economic Affairs Printed by employee consumer cooperatives (\-×) 2 + (2-Χ) 2 ^ (3-Χ) 2 + (Α-Χ) 2 7 DC = l2; AC = 22 + 32 + 42;
Ti=〔2 3 4〕； QT的每一個數値乘以QF即可得到AQT 〇 令Ti為做完DCT處理之影像區塊ac部份之數値，將這些數値 依漸次升高的次序排定，其中i=l，2,.......，n。量化係數QF便可 以依據下列方法來取得： 本紙張尺度適用中國國家標準（CNS ) A4規格（2丨Οχ297公麓） A7Ti = 〔2 3 4〕; Each number of QT is multiplied by QF to get AQT. Let Ti be the number of the ac part of the image block after DCT processing. These numbers are gradually increased in order. Schedule, where i = l, 2, ........., n. The quantization coefficient QF can be obtained according to the following methods: This paper size applies the Chinese National Standard (CNS) A4 specification (2 丨 〇297297 feet) A7
I 令i=l, QF=1 ; 2. 若即停止；否則即進行至3 ; ----^------氣— (請先閲讀背面之注#^項再填寫本頁) 3·若1，則i=i+l ;若Ti^RBiCTi，則 QF==2 1並停止；否則即行4 ; 4. QF=2n ’ 停止。 燦編碼疋連續長度編碼(run-length coding)與霍夫營蛙成 (Huffinancoding)兩者的組合，可以進一步地將粗影像經過處 ，的資料加以壓縮。這—部份的编碼，可利贿JPEG編碼之中的 7〇全相_编碼部份來進行，或者以更有效的二維連續I Let i = l, QF = 1; 2. If stop, otherwise proceed to 3; ---- ^ ------ qi — (Please read the note # ^ on the back before filling this page) 3. If 1, i = i + l; If Ti ^ RBiCTi, then QF == 2 1 and stop; otherwise, go to line 4; 4. QF = 2n 'stop. The combination of can-encoding, run-length coding, and Huffinan coding can further compress the data passed through the coarse image. This part of the encoding can be performed using the 70% phase_encoding part of the JPEG encoding, or in a more efficient two-dimensional continuous
技術來進行。 W 對於差異影像群（圖1中的B2至ΒΙΟ影像)而言，其純量量化 SQ(scalar quantization)可以下列的方法取得· 1. 將小波轉換後之差異影像群之各個像素數値，依漸次升高的次 序排定，77。 2. 選擇其中之一個數値當臨限値(Threshold)。 3. 差異影像之像素數値若小於臨限値，則將此數値設定為〇，否則 即將此數値除以臨限値後取整數。 經濟部中央標準局貝工消費合作社印製 此臨限値可依各個差異影像之特性做調變。例如，在B2差異影像 之中，新的臨限値，是將原始臨限値乘以一個量化係數QF而得到 的。AC能量與DC能量的比例被界定為： AC能量 σ1 R= _ = _ (2) DC能量 X2 其中R為經小波轉換後，差異影像之AC與DC能量比。 本紙張尺度適用中國國家揉準（CNS ) A4规格（210X297公釐） ' ---—- 發明説明（？ A7 __B7 _ 除了利用R取代Rbi之外，QF的計算是利用與向量量化表調變 係數相似的方法進行的。 1. 令 i=l，QF=2 ; 2. 若Ref即停止，•否則即進行至3 ; 3. 若i$n，則i=i+l ;若Ti.fRcTi，則 QFd並停止；否則即行4 ; 4. QF=2n+1，停止。 經濟部中央標準局貝工消費合作社印製 一種可變長度编碼VLC(variable length coding)方法被發展出 來以進一步地壓縮純量量化SQ所產生出來的資料。亦即，純量量 化SQ的輸出由可變長度编碼VLC程序進一步地加以壓縮。可變長 度編碼VLC的設計是依據兩個因素：（1)在純量量化Sq之後會出現 一長串的連續0 ; (2)在It或ΊΤ進行量化之後，會有可以辨別的許 多個非零數值出現。資料的掃描是依由左至右，由上而下的方式 進行的。就一個可變長度编碼VLC的程序而言，首先是要產生對 應於適當的非零串列的資料的一個霍夫曼表，以及對應 於連續度的適當串列的資料的一個霍夫曼表〇(HuffO)。令A'，ri與r〇 分別表示非零數値，非零數値的連續長度，以及零數値的連續長 度，則可變長度編碼VLC便可以依據下列的方式來進行，其亦可 使編碼較短： 1.連續非零數値的對應可變長度為编碼VLC係為 1 +Huffl〔A1〕+HuffO〔 r】〕， 其中+，1，Huffl〔〕，與〔〕分別代表梯級連結，等於 1個第一個位元，霍夫曼表〇，以及霍夫曼1 ; 本紙張尺度適用中國國家標準（CNS ) Α4规格（210X297公釐） ^1.1 - I I...... I - - - I J. ^ I I - - I I. 丁 i ..-¾ 、τ - {請先Μ讀背面之注意事項再填寫本頁) 393856 A7 B7 五、發明説明（ ----— 2.連續零數値的對應可變長编碼vlc係為 O + Huffl〔q〕+HufTO〔 Γ〕， 其中的q與r分別為r〇/256的商數與餘數0 最後，如圖1中顯示的，多工器18將細 ， 可適於傳輸與儲存的形式加以重新組合，^壓_資料以一種 亦即，可以由多工器18 輸出經過壓騎影像資料。如此-來，需要較低品質影像服務的 用户，便可以只接收粗影像，而需要較高品質影像服務的用户， 則可以接收到粗影像以及至少一個差異影像，此兩個影像在組合 在一起之後，便可以重建高品質的影像信號。 .-- - - J 1 I » - - --- Λ - - - - - -I !- II .....- ------ 丁 . * '驾 i (請先閲讀背面之注拿項再填寫本頁) 經濟部中央標準局員工消费合作社印製 瓣Technology. W For the difference image group (B2 to BIO image in Figure 1), its scalar quantization SQ (scalar quantization) can be obtained by the following methods: 1. The number of pixels of the difference image group after wavelet transformation is The ascending order is scheduled, 77. 2. Select one of the numbers as Threshold. 3. If the number of pixels of the difference image is less than the threshold, set this number to 0, otherwise divide the number by the threshold and take an integer. Printed by the Central Bureau of Standards, Ministry of Economic Affairs, Shellfish Consumer Cooperative, this threshold can be adjusted according to the characteristics of different images. For example, in the B2 difference image, the new threshold value is obtained by multiplying the original threshold value by a quantization coefficient QF. The ratio of AC energy to DC energy is defined as: AC energy σ1 R = _ = _ (2) DC energy X2 where R is the AC to DC energy ratio of the difference image after wavelet transformation. This paper size is applicable to Chinese National Standard (CNS) A4 (210X297mm) '------ Description of the invention (? A7 __B7 _ Except R instead of Rbi, the calculation of QF is made by adjusting with vector quantization table Coefficients are similar. 1. Let i = 1, QF = 2; 2. If Ref, stop; otherwise proceed to 3; 3. If i $ n, then i = i + l; if Ti.fRcTi QFd and stop; otherwise, go to line 4; 4. QF = 2n + 1, stop. A method of variable length coding (VLC) was developed by Shelley Consumer Cooperative of the Central Standards Bureau of the Ministry of Economic Affairs to further develop The data generated by compressing the scalar quantization SQ. That is, the output of the scalar quantization SQ is further compressed by the variable length coding VLC program. The design of the variable length coding VLC is based on two factors: (1) in After scalar quantization Sq, there will be a long series of continuous 0; (2) After It or Ίquantization, there will be many non-zero values that can be discerned. The data is scanned from left to right and from top to bottom In the following way. As far as a variable-length encoding VLC program, the first is Generate a Huffman table corresponding to the appropriate non-zero series of data, and a Huffman table 0 (HuffO) corresponding to the appropriate series of continuous data. Let A ', ri, and r0 denote respectively Non-zero number 値, continuous length of non-zero number 値, and continuous length of zero-number 値, variable-length encoding VLC can be performed according to the following methods, which can also make the encoding shorter: 1. Continuous non-zero number The corresponding variable length of 値 is 1 + Huffl [A1] + HuffO [r]], where +, 1, Huffl [], and [] respectively represent a step connection, which is equal to the first bit , Huffman table 0, and Huffman 1; This paper size applies the Chinese National Standard (CNS) A4 specification (210X297 mm) ^ 1.1-I I ...... I---I J. ^ II --I I. 丁 i ..- ¾, τ-{Please read the precautions on the back before filling this page) 393856 A7 B7 V. Description of the invention (----— 2. The correspondence of continuous zero number 値 may be The variable-length code vlc is O + Huffl [q] + HufTO [Γ], where q and r are the quotient and remainder 0 of r0 / 256 respectively. Finally, as shown in FIG. 1, the multiplexer 18 By recombining the thin, suitable forms for transmission and storage, the data can be output in one form, that is, the multiplexed image data can be output by the multiplexer 18. In this way, users who require lower-quality image services, It is possible to receive only coarse images, and users who need higher-quality image services can receive coarse images and at least one difference image. After the two images are combined together, a high-quality image signal can be reconstructed. .----J 1 I »----- Λ------I!-II .....- ------ Ding. * 'Drive i (Please read the note on the back first Take the item and fill out this page) Printed by the Consumer Cooperatives of the Central Standards Bureau of the Ministry of Economic Affairs
A A4 一嘈 7IV 9 2A A4 7IV 9 2