1300311 九、發明說明： 【發明所屬之技術領域】 本發明係關於一種壓縮編碼器（compression encoder )，尤指一種漸進式差值（progressive differential motion )的 JPEG(Joint Photographic Coding Expert Group)編 / 解碼器 (codec) 〇 【先前技術】1300311 IX. Description of the Invention: [Technical Field] The present invention relates to a compression encoder, and more particularly to a progressive differential motion JPEG (Joint Photographic Coding Expert Group) encoding/decoding (codec) 〇 [prior art]
隨著電腦、通訊及多媒體裝置的普及，影像壓縮（video compression)技術已經變得日趨重要。影像壓縮技術通常 用來將視訊影像（來自照相機、卡式錄放影機、雷射唱片 4)轉換為數位編碼資料。這樣的數位編碼資料很容易地 利用諸如網際網路（Internet )等網路架構進行資料傳輸。 當需要時，壓縮的影像資料可以執行解壓縮 (decompressing)動作，以便於電腦螢幕或其他顯示裝置 上顯示出來。基於網際網路頻寬的限制，以及降低藉由網 路傳輸圖形資料的存取或者下載時間的需求，對數位影像 進行壓縮處理是有其必要的。數位影像資料的高壓縮率意 味著更多數位影像資料可被儲存於一記憶裝置（例如^ 碟、軟碟或記憶卡）上，以及這些數位影像資料可更快地 於頻寬有限的傳輸線（例如電話線、網際網路等）上進 然而，上述壓縮技術的問題在於壓縮品質以及壓 供m“ 被兼顧。此外，現有的網際網路 =/避免頻寬過小的問題，使得能夠透過網 影像品質仍麸偏你。與也丨冰％ 得狗的 你1舉來說’一個每秒僅能下載-百萬 位凡、，且（1 Mps)資料的通道（例如數 触ScHber Line)即限制了用戶端進行;;二十用戶線：㈣恤 輸的能力，而上述效能為该數彳次 rea time)傳 在實際的資料用戶線的最佳數據， 際的狀況下，該數位資料用戶的傳輸能力甚至較^秒 1300311 一百萬位元組為低，這樣的傳輸能力以及品質，自然不足 以支持高品質影像的傳輸。因此，兼具效率以及效果的影 像壓縮/解壓縮方法是相當重要並且被需要的。 JPEG(Joint Photographic Coding Expert Group)標With the popularity of computers, communications and multimedia devices, video compression technology has become increasingly important. Image compression technology is commonly used to convert video images (from cameras, video cassette recorders, and compact discs 4) into digitally encoded data. Such digitally encoded data can be easily transferred using a network architecture such as the Internet. When needed, the compressed image data can be decompressed for display on a computer screen or other display device. It is necessary to compress digital images based on the limitations of the Internet bandwidth and the need to reduce the access or download time of graphics data transmitted over the network. The high compression ratio of digital image data means that more digital image data can be stored on a memory device (such as a disc, floppy disk or memory card), and these digital image data can be faster on a transmission line with limited bandwidth ( For example, the telephone line, the Internet, etc. are advanced. However, the problem with the above compression technology is that the compression quality and the pressure supply m are taken into consideration. In addition, the existing Internet =/ avoids the problem of too small a bandwidth, enabling the transmission of the network image. The quality is still bran to you. With the one that also has a dog, you can only download - one million per second, and (1 Mps) data channel (such as the number of touch ScHber Line) is limited The user terminal performs; 20 user lines: (four) shirts to lose the ability, and the above performance is the best data of the actual data user line for the number of times rea time, under the condition of the digital data user The transmission capacity is even lower than that of 1300311 and 1 million bytes. This transmission capability and quality are naturally insufficient to support the transmission of high-quality images. Therefore, image compression/decompression with efficiency and effect is effective. The method is quite important and is needed. JPEG (Joint Photographic Coding Expert Group)
準是被普遍且廣泛應用的壓縮技術之一。jPEG標準係利用 離政餘弦轉換（DCT，Discrete Cosine Transform)，將影 像分割映射（mapping)為8x8像素（pixel)的複數影像 區塊（blocks )’並將該等影像區塊由空間域轉成頻域，而 後對該等經由離散餘弦轉換所獲得的係數進行量化 (Quantizing )並藉由一斜向掃描模型（zig—zag scan pattern)轉換為一維向量，該一維向量利用長度編碼 (run-length encode )法以及霍夫曼編碼法（Huffman encode)的組合方法進行編碼。 一第1A圖以及第1B圖係為一高階方塊圖，該方塊圖顯 示JPEG標準對一原始影像進行壓縮、轉換以及重建的基本 運作方式。該原始影像藉由一個或多個顯示單元表示，每 一該等單元包含一個具有多位元像素的矩陣。一灰階影像 包含一單一顯示單元，而一個彩色影像可以包含三個該顯 ，單元。第1A圖以及第^圖所顯示的操作方式被應用在 母一該顯示單元中。 原始衫像 > 料11 〇被傳送到一壓縮編碼器丨以輸 一壓縮影像資料160。該壓縮影像資料16〇可以 一檔案以進行後續的修補（retrieval)以及^ (reconstruction)動作，該壓縮影像檔案16〇 某些傳輸媒介被傳送至遠端（remQtel咖iQn) ^ :或是後續的資料重建動作。在任何事件下，該壓縮J像 貝枓160視為傳达至一解壓縮解碼器（dec〇歐e =二」3,處理，使該解壓縮解碼器130得以提供 重建的衫像貝枓。該壓縮編碼器12〇使用複數種特殊的資It is one of the commonly used and widely used compression technologies. The jPEG standard uses a DCT (Discrete Cosine Transform) to map image segments into 8x8 pixels of multiple image blocks and convert these image blocks from spatial domain to In the frequency domain, the coefficients obtained by the discrete cosine transform are then quantized (Quantizing) and converted into a one-dimensional vector by a zig-zag scan pattern, which uses length coding (run The combination of the -length encode method and the Huffman encode method is performed. A 1A and 1B diagram is a high-order block diagram showing the basic operation of the JPEG standard for compressing, converting, and reconstructing an original image. The original image is represented by one or more display units, each of which contains a matrix of multi-bit pixels. A grayscale image contains a single display unit, and a color image can contain three such display units. The operation modes shown in Fig. 1A and Fig. 2 are applied to the mother-side display unit. The original shirt image > is sent to a compression encoder to convert a compressed image data 160. The compressed image data 16 can be archived for subsequent re-rendering and reconstruction operations, and the compressed image file 16 is transmitted to the remote end (remQtel iQn) ^: or subsequent Data reconstruction action. In any event, the compressed J image is considered to be conveyed to a decompression decoder (dec ei e = two) 3, processing, so that the decompression decoder 130 can provide a reconstructed shirt image. The compression encoder 12 uses a plurality of special funds
Cs 7 1300311 -料結構以進行壓縮’料資料結構的相關部分被視為必須 的次要貧訊（sub lnf〇rmati〇n)被傳送至該壓縮編碼器 12〇’以供該壓縮編碼器12〇進行影像重建。前述之壓縮技 術視次要資料的單一集合為一完整影像顯示單元。 習知的JPEG標準中，該壓縮編碼器12〇包含一離散餘 弦轉換（DCT)單元121、一量化器ι22以及一赫夫曼 (Huffman)編碼器123。該解壓縮解碼器13〇包含一赫夫 曼（Huffman)解碼器131、一解量化器ι32及一逆離散餘 弦轉換（IDCT)單元133。該次要資訊包含一量化表14〇 .以及一個赫夫曼編碼表集15〇，該量化表14〇為該量化器 122或該解篁化器132所使用，該赫夫曼編碼表集15〇為 該赫夫曼編碼器123或該赫夫曼解碼器ι31所使用。 該影像資料被分割為8像素乘以8像素的複數影像區 塊，每一該等影像區塊係被分別處理。一個8χ8影像區塊 中的位置係藉由底標（subscript )標示，其標示方法為第 一底標為列（row)位置座標，第二底標為行（c〇lumn)位 置座標。 ^ 再者’該原始影像資料可為灰階影像資料或彩色影像 f料。有數種方法得以分割一彩色影像為複數顯示單元。 標準的監視器使用r、G、b參數，其中該R、G、B係各別 代表紅色顯示單元、綠色顯示單元與藍色顯示單元。標準 電視廣播（NTSC)標準係使用γϋν參數，其中γ參數係指 輝度（luminance)顯示單元，而u參數及V參數係指色度 顯不單元（近似紅色及藍色）。印表機所使用的標準為CMYK ，數標準，其中C、Μ、Y及κ係各別指青綠色顯示單元、 =紅色顯示單元、黃色顯示單元及黑色顯示單元。CCIR601 標準揭露了一種r、G、Β參數以及γυν參數間的線性轉換。 如第1A圖所示，該原始影像資料11 〇係由R、G、β座標域 線性轉換至YUV座標域。如第1 β圖所示，該解壓縮影像資 1300311 ~ 料係自YUV座標域被線性轉換至R、G、B座標域中。 ,雖然JPEG標準是一種普遍且被廣泛使用的壓縮技 術，JPEG標準仍有幾個缺點。舉例來說，JpEG的缺點之一 在於低位兀率（l〇w bit rates)狀態下，離散餘弦轉換產 、^的不規則（irregularities )性以及重建影像中的不連 續性，上述的缺點形成了眾所皆知的圖像片（tiUng)邊 界假象以及方塊假象（blocking artifacts)問題。。邊界 叙象使得該等8x8像素影像區塊群間的邊界於重建影像中 出現荊述方塊假象使得影像品質不被希望地降低。jpeg • 的f二個缺點為JpEG無法超越特定精確度（fideHty)地 進行影像重建。換言之，假如一個影像係以某種精確度進 行編碼，而其後需要一較低的編碼精度（在受傳輸頻寬或 者儲存旎力所限的狀況），則該影像必須先被解碼還原，再 進行重新編碼以符合需求。 ,為了克服上述Jpeg的缺點，大部份的先進影像壓縮技 術於里化以及熵編碼程序後，使用小波轉換（waveid transform )技術對資料進行處理。基於小波轉換（打）不 需使用編碼方塊以及允許影像以不同精確度重建的特性， 與離散餘弦轉換相較，小波轉換（WT )更適合被使用於 • f理上。除此之外，小波轉換亦具有更好的信號能量壓縮 月匕^以及失真率表現。大部份現存的的以小波轉換為基礎 的壓縮技術將一影像資料分解成係數並且使用某種熵編碼 格式胃(例如赫夫曼【Huffman】編碼或算術編碼等）以進一 步地壓縮該影像資料。然而，根據資料型態的不同，這些 編碼格式可說是相當地複雜並且需要使用各式複雜的真^ 表（例如赫夫曼【Huffman】編碼），或複雜的資料結構（例 如zerotree資料結構）。因此，大部份現有的以小波轉換 為基礎的壓縮技術係複雜且難以實現的。 為了能夠獲得一個於壓縮以及解壓縮程序時，具有延 1300311 遲時間較低、網路負裁平抝4主从μ > β # 發明摇#卜、+、+ ?千均特的尚質解壓縮影像，本 發明棱供對於上述需求的一個解決方法。 【發明内容】 及月=ϊ” 一種資料壓縮/解壓縮裝置 ;方i用以獲侍一尚品質解壓 3壓縮裝置及、以法之_/解壓縮料料;時 a 時間較短，且具有網路負載平均的特性 本毛月之另一目的係提供一種使用 rn==程序r料壓縮/解壓縮裝置及其方= 3旦：ίί 低量化品質之量化步驟，以使得該 以：；ΪΓΓ:速地透過該赫夫曼編碼器進行網編碼。 提供;=早元得以快速地重建該壓縮影像資料流並 衫像輸出。該第二量化程序執行一較高量化品質之 ^縮二Γ。’對現·有？像輸出進行補償以獲得更高品質的解 φ 1 ^ 。上述壓縮/解壓縮技術的組合提供更好的效率、 ^的影像品質、更好的壓縮效率並且明顯n =線料:僂縮/解壓縮裝置及其方法可被應用在任= Μ及無線的傳輸通道技術上。 八 =本發明提供一種壓縮編碼器以便 的，包含下列元件·· 1曰 料中捕捉單元’用以接收一影像資料並將該影像資 之W ”、、員不單兀分割為複數個影像資料區塊； 衫像捕捉暫存器，用以儲存該影像資料區塊； 薄雷^測單元，與該影像捕捉單元以及該影像捕捉暫存 j連接，用以摘測由該影像捕捉單元輸入之該影像 及由該影像捕捉暫存器取出之該影像資料區塊間 η谷疋否不同； 〜 一壓縮單元，電連接該偵測單元，以一預先決定 縮品質對該影像捕捉暫存器取出之該影像資料區塊= 1300311 縮’其中該預先決定之壓縮品質係藉由該影像捕捉單元輸 入之該影像資料區塊以及由該影像捕捉暫存器取出之該影 像資料區塊間之内容差異決定；以及 一品質參數暫存器，用以指定及儲存一壓縮品質指定 參數，當該壓縮顯示單元壓縮由該影像捕捉暫存器取出之 該影像資料區塊時，該壓縮品質指定參數指出該品質參數 暫存器儲存之壓縮品質資料。 本案得藉由以下列圖示與詳細說明，俾得一更深入之 了解。 【實施方式】 一立明參閱第2A圖係為本案一較佳實施例之壓縮編碼器 示w圖^亥壓縮編碼器2 0 0包含一影像捕捉單元2 〇 1、一 偵測單元202、一影像捕捉暫存器203、一壓縮單元204及 一畫質參數暫存器205。原始影像資料被傳送到該壓縮編 碼器200以輸出一壓縮影像資料流。該壓縮影像資料流可 以被儲存為一檔案以進行後續的修補（retrieval)以及重 建（reconstruction)動作，該壓縮影 某些傳輸媒介被傳送至遠端（r⑽telQGatiQn) Ϊ或疋後_的資料重建動作。在任何事件下，該壓縮影像 貝料流視為被傳送至一解壓縮解碼器3〇〇進行處理，使爷 Ϊ = ^器3GG得以輸出重建的影像資料。該壓縮編^ !像捕捉單元201捕捉原始影像以分割該原 ::像為隱像素大小的影像區塊並對每一該 進仃分別處理。如第？「園故- _ 兄 ^中”底標標示，其標示方法為第一底標為列 等辽標’第二底標為行（C〇lumn)位置座標。該 兮：上：二：左至右’自頂端至底端依序形成，因此， ϊίΐ , (°’°)、該右上角之座標為（U-1)、 该左下角之座標為（IM，0)、該較右下角之座標為 1300311 上開始時，對應於影像區塊（m，n+"的資 該影像捕捉201所接收，隨後該债測單元202接收； 塊u’ n+1)，並將影像區塊（m，n+1)的内容“ 像區塊（m，n)的内容進行比對，其中該影像區塊 ，内容係由該影像捕捉暫存器2Q3中取得。如輸人; ===與該影像捕捉暫存器m中儲存的影像以 貝”所不同，則該壓縮單元2〇4壓縮該影像捕捉暫存器 所儲存的影像區塊，並將該畫質參數暫存器2〇5 ;Cs 7 1300311 - The material structure is compressed to the relevant part of the material data structure is regarded as a necessary secondary information (sub lnf〇rmati〇n) is transmitted to the compression encoder 12〇' for the compression encoder 12 〇 Image reconstruction. The foregoing compression technique considers a single set of secondary data to be a complete image display unit. In the conventional JPEG standard, the compression coder 12A includes a discrete cosine transform (DCT) unit 121, a quantizer ι22, and a Huffman encoder 123. The decompression decoder 13A includes a Huffman decoder 131, a dequantizer ι32, and an inverse discrete cosine transform (IDCT) unit 133. The secondary information includes a quantization table 14A. and a Huffman coding table set 15〇, which is used by the quantizer 122 or the decimation device 132, the Huffman coding table set 15 〇 is used by the Huffman encoder 123 or the Huffman decoder ι31. The image data is divided into a plurality of image blocks of 8 pixels by 8 pixels, and each of the image blocks is processed separately. The position in an 8χ8 image block is indicated by a subscript, which is marked by the first bottom mark as the row position coordinate and the second bottom mark as the line (c〇lumn) position coordinate. ^ Again, the original image data can be grayscale image data or color image material. There are several ways to split a color image into a plurality of display units. The standard monitor uses the r, G, and b parameters, where the R, G, and B systems each represent a red display unit, a green display unit, and a blue display unit. The standard television broadcast (NTSC) standard uses the γϋν parameter, where the γ parameter refers to the luminance display unit, and the u parameter and the V parameter refer to the chrominance display unit (approximate red and blue). The standard used by the printer is CMYK, the number standard, where C, Μ, Y, and κ are each a cyan display unit, a red display unit, a yellow display unit, and a black display unit. The CCIR 601 standard exposes a linear transition between r, G, Β parameters and γ υ ν parameters. As shown in Fig. 1A, the original image data 11 is linearly converted from the R, G, and β coordinate domains to the YUV coordinate domain. As shown in the 1st graph, the decompressed image 1300311~ material is linearly converted from the YUV coordinate domain to the R, G, and B coordinate domains. Although the JPEG standard is a widespread and widely used compression technique, the JPEG standard still has several drawbacks. For example, one of the shortcomings of JpEG is the discrete cosine transform, the irregularities of the ^, and the discontinuities in the reconstructed image. The well-known tiUng boundary artifacts and blocking artifacts. . The boundary narrative causes the boundary between the 8x8 pixel image block groups to appear in the reconstructed image to cause the image quality to be undesirably reduced. The two disadvantages of jpeg • f are that JpEG cannot perform image reconstruction beyond a certain degree of precision (fideHty). In other words, if an image is encoded with some precision and then requires a lower encoding accuracy (in the case of limited transmission bandwidth or storage power), the image must be decoded and restored first. Recode to meet your needs. In order to overcome the shortcomings of Jpeg mentioned above, most advanced image compression techniques use wavelet transform technology to process data after the internalization and entropy coding procedures. Wavelet transform (WT) is more suitable for use in wavelet transforms than for discrete cosine transforms, which does not require the use of coded blocks and allows the image to be reconstructed with different precision. In addition, wavelet conversion also has better signal energy compression and distortion rate performance. Most existing wavelet-based compression techniques decompose an image into coefficients and use some sort of entropy coding format (such as Huffman coding or arithmetic coding) to further compress the image data. . However, depending on the type of data, these encoding formats can be said to be quite complex and require the use of various complex tables (such as Huffman encoding) or complex data structures (such as the zerotree data structure). . Therefore, most of the existing compression techniques based on wavelet transform are complex and difficult to implement. In order to be able to obtain a compression and decompression program, it has a delay of 1300311, a low latency, a network negative 拗4 master slave μ > β # invention shake #卜, +, + ?千均特Compressed images, the present invention provides a solution to the above needs. [Summary of the invention] and month = ϊ" a data compression / decompression device; the party i is used to obtain a quality decompression 3 compression device and the method of decompressing the material; time a is shorter, and has Network Load Average Characteristics Another purpose of this month is to provide a quantization step using rn== program r material compression/decompression device and its =3: ίί low quantization quality, so that the ::ΪΓΓ : Quickly encode the net through the Huffman encoder. Provide; = early element to quickly reconstruct the compressed image data stream and output the image. The second quantization program performs a higher quantization quality. 'Compensate the output with the image output to obtain a higher quality solution φ 1 ^. The combination of the above compression/decompression techniques provides better efficiency, image quality, better compression efficiency and significantly n = line Material: The collapsing/decompressing device and its method can be applied to any transmission channel technology of wireless network. Eight= The present invention provides a compression encoder, including the following components: To receive an image data and The image resource W", the member is not only divided into a plurality of image data blocks; the shirt image capture register is used to store the image data block; the thin mine detection unit, the image capture unit and the image capture a temporary j connection for extracting whether the image input by the image capturing unit and the image data block extracted by the image capturing buffer are different; ～ a compression unit, electrically connecting the detection The image data block that is taken out by the image capture buffer by a predetermined predetermined quality = 1300311, wherein the predetermined compression quality is the image data block input by the image capture unit and Determining the content difference between the image data blocks taken by the image capture register; and a quality parameter register for specifying and storing a compression quality specification parameter, when the compressed display unit is compressed by the image capture register When the image data block is taken out, the compression quality designation parameter indicates the compression quality data stored by the quality parameter register. In this case, we can get a deeper understanding by using the following illustrations and detailed explanations. [Embodiment] FIG. 2A is a compression encoder of a preferred embodiment of the present invention. The compression encoder 200 includes an image capturing unit 2, a detecting unit 202, and a The image capture register 203, a compression unit 204, and an image quality parameter register 205. The raw image data is transmitted to the compression encoder 200 to output a compressed image data stream. The compressed image data stream can be stored as a file for subsequent retrieving and reconstruction actions that are transmitted to the remote end (r(10)telQGatiQn) or subsequent data reconstruction actions. . In any event, the compressed image stream is considered to be transmitted to a decompression decoder 3 for processing, so that the 3 = ^ 3GG can output the reconstructed image data. The compressed image capturing unit 201 captures the original image to divide the original image image as a hidden pixel size and processes each of the images separately. As the first? The logo of the "Yuan- _ brother ^zhong" is marked as the first base mark as the column, and the second base mark is the position coordinate of the line (C〇lumn). The 兮: top: two: left to right 'formed from the top to the bottom, so ϊίΐ , (°'°), the coordinates of the upper right corner are (U-1), and the coordinates of the lower left corner are (IM) , 0), the coordinate of the lower right corner is 1300311 at the beginning, corresponding to the image block (m, n+" received by the image capture 201, and then the debt measurement unit 202 receives; block u' n+ 1), and compare the content of the image block (m, n+1) "image block (m, n), wherein the image block, the content is obtained by the image capture register 2Q3 If the input image is different from the image stored in the image capture register m, the compression unit 2〇4 compresses the image block stored by the image capture register, and the image block is Image quality parameter register 2〇5;
之壓縮貝彳日示參數儲存為較低壓縮值。該壓縮單元2 輸出該壓縮影像資料流。 此外，§輸入的影像區塊被偵測到與該影像 器203中儲存的資料並無差異，且該影像捕捉暫存器暫^ 中儲存的資料之該壓縮品質參數指示為低時，該壓縮單元 Μ4即對該輸出影像區塊以及重建資料内容間的差異進行 壓縮以產生一壓縮差值資料，並將該畫質參數暫存器 中之該壓縮品質指示參數更改為較高壓縮值，其中該重建 資料係由該壓縮影像資料流中取得。 ”—請參閱第2Β圖所示，係為根據本案較佳實施例之一解 壓縮解碼器示意圖。該解壓縮解碼器3〇〇包含一解壓縮單 元=01及一重建暫存器3〇2。該解壓縮3〇1接收該壓縮影 像資料流以重建該壓縮影像資料並提供一影像輸出。此 外，該解壓縮單元301接收該壓縮差值資料並重建該壓縮 差值資料以產生一補償資料。該補償資料係被補償至具有 較低畫貝的輸出影像資料以獲得較高晝質的解壓縮影像。 晴參閱第3圖所示，係為第2Α圖所示之該壓縮單元之 二變化實施例。該壓縮單元400包含一離散餘弦轉換元件 單元401、一反離散餘弦轉換單元4〇2、一離散餘弦轉換單 元403、一第一量化單元4〇4、一解量化單元4〇5、一第二 量化元件406、一第一赫夫曼編碼器407、一第二赫夫曼編 12 1300311 】器408及戴爾他（Delta)單元409。該離散餘弦轉換 早元401接收一影像資料區塊並進行一低品質量化動作以 產生複數離散餘弦轉換參數。該第一量化單元404量化來 自一該離散餘弦轉換單元401之該等離散餘弦轉換參數並進 仃Γ低品質量化動作。該等量化參數利用該第一赫夫曼編 ，态j07進行熵編碼。然而，該解量化單元405對來自該 第曰里化單元404之量化參數進行反量化動作以產生複數 解里化參數。該反離散餘弦轉換單元4〇2執行反離散餘弦 轉換。該反離散餘弦轉換過程輸出mxri個影像區塊。該戴 爾t ( Delta)單元409接收該原始影像資料區塊並且重建 該等影像資料區塊，隨後執行影像轉換以及旋積 (convolution)以產生一影像差值資料，該影像差值資料 包含該原始影像資料區塊以及該重建影像資料間之差值資 料。該離散餘弦轉換單元4〇3接收該影像差值資料並且執 ^ 一離散餘弦轉換以產生複數離散餘弦參數。該第二量化 單元406執行一尚品質量化並量化由該離散餘弦轉換單元 403輸出的該等離散餘弦參數。該等量化參數藉由該第二 赫夫曼編碼器408進行熵編碼。 當上述之該壓縮方法被應用於色彩影像上時，相同的 該壓縮方法可以分別地使用於每一色彩頻帶上。然而，由 於YUV色彩空間集中其處理能量在γ參數處理，因而較適 於JPEG壓縮的特性，假如該3個用以顯示色彩的色彩頻帶 並非該yuv色彩空間（例如可使用該RGB【紅綠藍】色彩 ，間）時，先將該等色彩空間轉換為γϋν色彩空間的方法 是具有優勢的，其中Υ參數係指輝度（luminance)顯示單 元，而U參數及v參數係指色度顯示單元（近似紅色及藍 色但亦不以上述方法為限，任何適合jPEG壓縮特性的 色彩空間，均可用來進行色彩空間轉換。請參閱第4圖所 13 1300311 '示，係為第2A圖所示壓縮單元之另一實施方塊圖。該壓縮 單元包含一離散餘弦轉換單元501、一量化單元5〇27 一多 工器503、一赫夫曼編碼器504、一戴爾他單元5〇5、一解 量化單元506及一第二量化單元507。該離散餘弦轉換單 、 元501接收一影像資料區塊，並進行一離散餘弦轉換以產 生複數離散餘弦轉換參數。該量化單元5〇2進行一低品質 “ 量化動作以量化來自該離散餘弦轉換單元501之該等 餘弦轉換參數。該解量化單元506對該等由該量化單元5〇2 輸出之量化參數進行解量化以產生複數解量化參數。該戴 % 爾他單元505接收該等量化參數以及該等解量化參數並執 行影像轉換以及旋積以產生一影像差值資料，該影像^值 資料包含該原始影像資料區塊以及該重建影像資料間之差 值資料。該量化單元507執行一高品質量化動作並量化經 由該戴爾他單元505輸出之該影像差值資料。該多工器5〇3 f換，由該量化單元502輸出之該等量化參數以及經由該 里化單το 507輸出之量化差值資料。該等量化參數藉由該 赫夫曼編碼器504進行熵編碼。 曰 ^ 當W述之該壓縮方法被應用於色彩影像上時，該壓縮 方法可以分別地使用於每一色彩頻帶上。然而，由於γυν 色，空間集中其處理能量在γ參數處理，目而使其較適於 JPEG壓縮的特性，假如該3個用以顯示色彩 =渭色彩空間（例如可使用該叫⑽ 二i二：將二等色彩空間轉換為，色彩空間的方法是 ^ ^ ' Y參數係指輝度（luminance)顯示單元， ♦、及、V參數係指色度顯示單元（近似紅色及藍色）， 二V :以ΐ ί方法為限，任何適合JPEG壓縮特性的色彩空 例的舲^别田進/丁色彩空間轉換。前述對本發明較佳實施 | w , 離散餘弦轉換單元產生離散餘弦轉換參 14 1300311 Ϊ二弦轉換單元以實現-反離散餘弦轉 ΐ換二Ϊ:Γ產生轉換參數的轉換單元，例如小波 為第2Β圖所-用於本發明中。請參閱第5圖所示，係 圖所示，口-解^壓1 解碼器之一變化實施方塊圖。如第5 ^ 解碼态6 01可以接收—姑^ 數LQ，該低品質量化射 被、、扁馬的低w貝篁化係 提供，以及_被 、σ ^由該赫夫曼編碼器407所 係數ΗΟ禆拉ϋ 回0口貝篁化係數HQ,該高品質量化 ”數Q係猎由该赫夫曼編碼器4〇8 接收該被編碼的低品曾旦彳卜仫叙ΤΛ杈仏名、、扁碼器6〇1 品質詈化數LQ，並重建該被編碼的低 以/生一影像輸出621。該影像輸出62i 碼的V f ㈣。該解碼器咖亦接收該被編 ^ HQ J^ . f * ^ 不兩1員貞料6 2 2。该影德齡山β + 由該多工器_multiple…進可；^接經 處理並輸出。經由一加牛哭r : 丁f路设合（multlPlex) 旦〈德介盔 / D ( Adder) 630的加總，該重建 〜出621以及該補償資料622之總和。 〜…來祝，本發明提供一種使用雙重量化 ❿ Γ-t Trl 7ir 等量化失教π Ζ較低里化品質之量化步驟，以使得該 以此，速，Γ赫夫曼編碼器進行熵編碼。 提供-=ίΓ :Γ速地重建該I縮影像資料流並 象輸出。該第二量化程序執行一較 7巧影像輸出進行補償以獲得更高品質二解 更以像縮技術的組合提供更好的效率、 失。該資料㈣/解Μ縮裝置及其方法可 21 以及無線的傳輸通道技術上。 後應用在任何有線 本發明較佳具體實施例的前述說明係用於示範及說明 15 1300311 【主要元件符號說明】 11 0原始影像資料 1 21離散餘弦轉換單元 123赫夫曼編碼器 130解壓縮解碼器 132解量化器 133逆離散餘弦轉換單元 1 5 0赫夫曼編碼表集 200壓縮編碼器 202偵測單元 203影像捕捉暫存器 205晝質參數暫存器 301解壓縮單元 400解壓縮單元 401離散餘弦轉換單元 402反離散餘弦轉換單元 403離散餘弦轉換單元 404第一量化單元 406第二量化元件 407第一赫夫曼編碼器 408第二赫夫曼編碼器 501離散餘弦轉換單元 503多工器 505戴爾他單元 507量化單元 6 01編碼器 621影像輸出 630加法器 HQ高品質量化係數 120壓縮編碼器 122量化器 131赫夫曼解碼器 140量化表 160壓縮影像資料 201影像捕捉單元 204壓縮單元 300解壓縮解碼器 302重建暫存器 405解量化單元 409戴爾他單元 502量化單元 5 0 4赫夫曼編碼器 506解量化單元 620影像暫存器 622補償資料 640多工器 〔8The compressed shellfish date parameter is stored as a lower compression value. The compression unit 2 outputs the compressed image data stream. In addition, if the input image block is detected to be indistinguishable from the data stored in the imager 203, and the compression quality parameter of the data stored in the image capture buffer is temporarily low, the compression is performed. The unit Μ4 compresses the difference between the output image block and the reconstructed data content to generate a compressed difference data, and changes the compression quality indication parameter in the image quality parameter register to a higher compression value, wherein The reconstructed data is obtained from the compressed image data stream. As shown in FIG. 2, it is a schematic diagram of decompressing a decoder according to one of the preferred embodiments of the present invention. The decompression decoder 3 includes a decompression unit=01 and a reconstruction register 3〇2 The decompressed 3〇1 receives the compressed image data stream to reconstruct the compressed image data and provides an image output. In addition, the decompression unit 301 receives the compressed difference data and reconstructs the compressed difference data to generate a compensation data. The compensation data is compensated to the output image data with a lower image to obtain a higher quality decompressed image. As shown in Fig. 3, it is the second change of the compression unit shown in Fig. 2 The compression unit 400 includes a discrete cosine transform component unit 401, an inverse discrete cosine transform unit 4〇2, a discrete cosine transform unit 403, a first quantization unit 4〇4, and a dequantization unit 4〇5. a second quantization component 406, a first Huffman encoder 407, a second Huffman code 12 1300311 device 408 and a Dell unit 409. The discrete cosine transform early element 401 receives an image data area Block and A low quality quantization action is performed to generate a complex discrete cosine transform parameter. The first quantization unit 404 quantizes the discrete cosine transform parameters from a discrete cosine transform unit 401 and performs a low quality quantization operation. The first Huffman is compiled, and the state j07 is entropy encoded. However, the dequantization unit 405 performs an inverse quantization operation on the quantization parameter from the dimerization unit 404 to generate a complex de-living conversion unit. The inverse discrete cosine transform is performed by 4〇2, and the inverse discrete cosine transform process outputs mxri image blocks. The Dell t (Delta) unit 409 receives the original image data block and reconstructs the image data blocks, and then performs image conversion. And convolution to generate an image difference data, the image difference data comprising the difference between the original image data block and the reconstructed image data. The discrete cosine transform unit 4〇3 receives the image difference value Data and a discrete cosine transform to generate a complex discrete cosine parameter. The second quantization unit 406 performs a still product The discrete cosine parameters output by the discrete cosine transform unit 403 are quantized and quantized. The quantization parameters are entropy encoded by the second Huffman encoder 408. When the compression method described above is applied to a color image The same compression method can be used separately for each color band. However, since the YUV color space concentrates its processing energy in the γ parameter processing, it is more suitable for the characteristics of JPEG compression, if the three are used to display colors. When the color band is not the yuv color space (for example, the RGB [red, green, and blue] colors can be used), it is advantageous to first convert the color space into a γϋν color space, where the Υ parameter refers to luminance (luminance) The display unit, and the U parameter and the v parameter refer to the chromaticity display unit (approximate red and blue but not limited by the above method, any color space suitable for the jPEG compression characteristic can be used for color space conversion. Please refer to Fig. 4, Fig. 13 1300311', which is another block diagram of the compression unit shown in Fig. 2A. The compression unit comprises a discrete cosine transform unit 501, a quantization unit 5〇27-multiplexer 503, a Huffman encoder 504, a Dell unit 5〇5, a dequantization unit 506 and a second quantization unit. 507. The discrete cosine transform unit 501 receives an image data block and performs a discrete cosine transform to generate a complex discrete cosine transform parameter. The quantization unit 〇2 performs a low quality "quantization action to quantize the cosine transform parameters from the discrete cosine transform unit 501. The dequantization unit 506 solves the quantization parameters output by the quantization unit 5 〇 2 Quantizing to generate a complex dequantization parameter. The Dynacell unit 505 receives the quantization parameters and the dequantization parameters and performs image conversion and convolution to generate an image difference data, the image data comprising the original image The data block and the difference data between the reconstructed image data. The quantization unit 507 performs a high quality quantization operation and quantizes the image difference data outputted through the Dell unit 505. The multiplexer 5〇3 f changes, The quantization parameters output by the quantization unit 502 and the quantized difference data outputted via the refinement unit το 507. The quantization parameters are entropy encoded by the Huffman encoder 504. When the compression method is applied to a color image, the compression method can be used separately for each color band. However, due to the γυν color, the space concentrates its processing energy. The amount is processed in the gamma parameter, so that it is more suitable for the characteristics of JPEG compression, if the three are used to display the color = 渭 color space (for example, the call can be used (10) two i two: convert the second color space to, color The space method is ^ ^ 'Y parameter refers to the luminance display unit, ♦, and, V parameters refer to the chromaticity display unit (approximate red and blue), and the second V: is limited to the ΐ ί method, any suitable The color space of the JPEG compression feature is 别^别田进/丁色空间转换. The foregoing is a preferred implementation of the present invention | w , the discrete cosine transform unit generates a discrete cosine transform reference 14 1300311 Ϊ two-string transform unit to achieve - anti-discrete cosine The conversion unit is: the conversion unit that generates the conversion parameter, for example, the wavelet is the second picture - used in the present invention. Please refer to the figure 5, as shown in the figure, the port-decompression 1 decoder A variation implementation block diagram. For example, the 5th decoding state 6 01 can receive - the number LQ, the low-quality quantized shot, the low-ma beidization system of the flat horse, and the _ is, σ ^ from the Fuman encoder 407 coefficient ΗΟ禆 拉ϋ Back to 0 mouth Beihua system HQ, the high-quality quantification "number Q" hunting by the Huffman encoder 4〇8 receives the encoded low-grade Zengdan 彳 仫 仫 ΤΛ杈仏 、,, flat code 〇 詈 詈 詈LQ, and reconstructing the encoded low-in-one image output 621. The image outputs Vf (4) of the 62i code. The decoder coffee also receives the edited HQ J^. f * ^ not two members 6 2 2. The shadow of the age of the mountain β + by the multiplexer _multiple ... into the; ^ received by the process and output. Through a plus cow crying r: Ding f road set (multlPlex) The sum of D (Adder) 630, the reconstruction ~ out of 621 and the sum of the compensation data 622. ~... I wish, the present invention provides a quantization step using a quantized 失 t t t t t t Ζ 里 里 t t t , , , , , , , , , , , , , , , , 以此 以此 以此 以此 以此 以此 以此 以此 以此 以此 以此 以此 以此 以此 以此. Provide -=ίΓ : Quickly reconstruct the I image stream and output it like an image. The second quantization program performs a more accurate image output for compensation to obtain a higher quality two solution. The combination of the image reduction techniques provides better efficiency and loss. The data (4)/unwinding device and its method can be used as well as wireless transmission channel technology. The foregoing description of the preferred embodiment of the present invention is for exemplary and illustrative purposes. 15 1300311 [Major component symbol description] 11 0 original image data 1 21 discrete cosine transform unit 123 Huffman encoder 130 decompression decoding Counter 132 dequantizer 133 inverse discrete cosine transform unit 1 H 0 Huffman coding table set 200 compression encoder 202 detection unit 203 image capture register 205 quality parameter register 301 decompression unit 400 decompression unit 401 Discrete cosine transform unit 402 inverse discrete cosine transform unit 403 discrete cosine transform unit 404 first quantization unit 406 second quantization element 407 first huffman encoder 408 second huffman encoder 501 discrete cosine transform unit 503 multiplexer 505 Dell other unit 507 quantization unit 6 01 encoder 621 image output 630 adder HQ high quality quantization coefficient 120 compression encoder 122 quantizer 131 Huffman decoder 140 quantization table 160 compressed image data 201 image capture unit 204 compression unit 300 Decompression decoder 302 reconstruction register 405 dequantization unit 409 Dell other unit 502 quantization unit 5 0 Huffman Solutions of 506 quantization unit 620 registers 622 the compensation image data multiplexer 640 [8