1335522 九、發明說明： 【發明所屬之技術領域】 本發明疋有關於一種浮水印之嵌入方法，特別是有關 於一種具有壓縮功能之浮水印之嵌入方法。 【先前技術】 隨著科技的進步與發展，網路已成為人們日常生活中 必備的工具。但隨著網路的日漸發達，數位多媒體也很容 易透過網路來傳送或拷貝，因此保護數位多媒體的擁有權 及内容的完整性是一項非常重要的課題。為了避免有心人 士對數位多媒體的蓄意變造竄改或轉售圖利，因而出現了 數位浮水印技術，以便應用於保護數位多媒體的完整性及 智慧財產權。所謂的浮水印嵌入技術是一種把特殊訊息加 進數位多媒體資訊的一個技術。一般而言，根據其應用的 目的，浮水印技術可分成二種主要的類型：強韌型（r〇bust) 浮水印技術以及易碎型（fragile)浮水印技術。強韌型浮水 印技術是利用偵測浮水印來核對數位多媒體的智慧財產 權的合法性，其至少需符合二種基本需求：透通性 (transparency)及強韌性（robustness);亦即所敬入的浮水印 即使受到訊號處理過後，亦或是遭受到蓄意的竄改，都需 保有原始所嵌入的浮水印的特徵；浮水印既不可以輕易被 移除和察覺，也不可以使數位多媒體内容產生過多的失 真。相對於強韌型浮水印技術，易碎型的浮水印技術，是 用來針對多媒體的内容作認證（authenticati〇n)的應用，換 5 句話說就是偵測數位多媒體的内容是否有受任何的改變 並明確指出數位内容被竄改的位置。 目前的影像或視訊壓縮標準，如JPEG、MPEG系列， 皆著重於以數位餘弦轉換（DCT)編碼的架構。這類型的編 碼，利用到DCT轉換、量化（quantization)、zigzag掃瞎、 量長編碼及熵（entropy)編碼，來移除影像或視訊的冗餘資 訊。雖然這些編碼提供高視覺品質、便利的儲存及傳播等 優點，但由於目前強大的數位多媒體編輯軟體的處理功 能，一般人很容易對數位媒體内容作修改，因而在應用上 出現了數位多媒體完整性的隱憂。近年來，藉由嵌入小量 的數位資訊（即浮水印）至多媒體資訊以進行驗證的技術， 已經受到工商界及學術界很大的注意，透過此一技術可以 達到驗證數位多媒體的内容完整性及確實性。 在有效的影像浮水印技術架構方面，其關鍵的必要條 件是低失真、低計算複雜度、低緩衝需求，以及對竄改的 高敏感性。雖然已經有很多以驗證為目的的易碎型浮水印 技術被提出，但這些易碎型浮水印技術並沒有積極地著重 在編碼效率的議題上。然而編碼效率是會影響網路的傳輸 時間及資料的儲存的空間，對於數位多媒體通信是不容忽 視的。因此，需要一種新的易碎形浮水印技術，既可以達 到驗證數位多媒體的完整性，且同時也能增進編碼效率。 【發明内容】 因此，本發明之一方面係在提供具有壓縮功能的浮水 1335522 印嵌入方法，以產生具有較小資料量之已加浮水印圖像。 根據本發明之一實施例，此方法至少包含：提供原始 圖像，其中原始圖像係包含有複數個原始圖像區塊，每一 原始圖像區塊係具有區塊強度矩陣，此區塊強度矩陣係由 複數個原始圖像像素之強度值所組成，區塊強度矩陣係代 表原始圖像像素的空間分佈情況；提供浮水印圖像，其中 浮水印圖像係包含有複數個浮水印像素，此些浮水印像素 係以一對一之方式對應至上述原始圖像區塊；對每一原 始圖像區塊之區塊強度矩陣進行離散餘弦轉換（Discrete Cosine Transform : DCT)步驟，以獲得DCT係數矩陣，其 中DCT係數矩陣係包含複數個DCT係數，每一 DCT係數 係代表上述之區塊強度矩陣於頻率域之各頻率的加權 值；利用量化矩陣來對DCT係數矩陣進行量化處理，以獲 得已量化係數矩陣，其中已量化係數矩陣係包含複數個已 量化DCT係數；以預設掃描順序對已量化係數矩陣進行掃 描，以依序取出已量化DCT係數，並獲得一維圖像資料； 對一維圖像資料進行量長編碼（Run-length coding，RLC) 步驟，以獲得複數個量長（Run-length);利用量長 (Run-length)來計算已量化DCT係數之奇數個數，以提供 奇數個數總數目值；以及對上述浮水印像素之一者和上述 原始圖像區塊之一者進行嵌入判斷步驟，以將上述浮水印 像素之該者嵌入至上述原始圖像區塊之該者，其中浮水印 像素之該者具有一顏色，上述浮水印像素之該者係對應至 上述原始圖像區塊之該者，上述之嵌入判斷步驟至少包 7 1335522 含，當上述之顏色為第一顏色時，進行第一判斷步驟，以 判斷奇數個數總數目值是否為奇數，若奇數個數總數目值 為偶數則進行一係數校正步驟；以及當上述之顏色為第二 顏色時，進行第二判斷步驟，以判斷奇數個數總數目值是 否為偶數，若奇數個數總數目值為奇數則進行上述之係數 校正步驟； 又，上述之已量化DCT係數係包含複數個非零已量 化DCT係數。 又，上述之係數校正步驟至少包含：進行第三判斷步 驟，以利用量長來判斷上述之非零已量化DCT係數之最後 非零者是否大於0並提供第三判斷結果；以及進行奇偶性 變更步驟，以根據第三判斷結果來變更最後非零者之奇偶 性。 又，右上述之第二判斷結果為是：上述之奇偶性變更 步驟係將最後非零者之值減1β若最後非零者之值剛好為 1，則最後非零者之值變更為0,而上述量長之個數減少一 個。 又，若上述之第三判斷結果為否：上述之奇偶性變更 步驟係將最後非零者之值加1。若最後非零者之值剛好為 -1’則最後非零者之值變更為0，而上述量長之個數減少 一個。 由上述說明可知，本發明之特徵係在於：當獲得原始 圖像區塊之已量化DCT矩陣後，記錄已量化DCT矩陣所 包含之已量化DCT係數的奇數個數，並利用此奇數個數總 8 1335522 數目值之奇偶性來表示被嵌入的浮水印圖像像素的顏 色。在本發明之實施例中，奇數個數總數目值為奇數係代 表被嵌入的浮水印圖像像素的顏色為黑色；奇數個數總數 目值為偶數係代表被嵌入的浮水印圖像像素的顏色為白 色。 【實施方式】 請參照第1圖和第2圖，第1圖係繪示根據本發明之 第一實施例之浮水印圖像嵌入示意圖，第2圖係繪示根據 本發明之第一實施例之原始圖像區塊之構造示意圖，其中 浮水印圖像110係包含8乘8個浮水印像素丨12，每一浮 水印像素112可例如為黑色或白色。原始圖像丨14係包含 有64乘64個原始圖像區塊116.，每一原始圖像區塊116 係包含有8乘8個原始圖像像素118，其中每一原始圖像 區塊116係具有區塊強度矩陣’此區塊強度矩陣係由原始 圖像像素118之強度值（例如：灰階值或色度值）所組成。 本發明之第一實施例係將浮水印像素112以一對一之方式 嵌入至原始圖像區塊116中，其中浮水印像素U2a係用以 嵌入原始圖像區塊116a，浮水印像素n2b係用以嵌入原 始圖像區塊116b，以此類推。由上述之說明可知本發明之 第一實施例之浮水印圖像110和原始圖像114所包含之像 素並不受限，此為熟悉本領域之具有通常知識者所能輕易 理解。 請參照第3圖，其係繪示根據本發明之第一實施例之 9 1335522 浮水印之嵌入方法的流程示意圖。由於每一浮水印像素 112的嵌入方法皆相同，在以下的說明中僅以浮水印像素 112a和原始圖像區塊ii6a來舉例說明。1335522 IX. Description of the Invention: [Technical Field] The present invention relates to a method of embedding a watermark, and more particularly to a method of embedding a watermark with a compression function. [Prior Art] With the advancement and development of technology, the Internet has become an indispensable tool in people's daily lives. However, with the development of the Internet, digital multimedia is also easily transmitted or copied over the Internet. Therefore, it is a very important issue to protect the ownership of digital multimedia and the integrity of content. In order to avoid the deliberate alteration or resale of digital multimedia by digital people, digital watermarking technology has emerged to protect the integrity and intellectual property rights of digital multimedia. The so-called watermark embedding technology is a technique for adding special information to digital multimedia information. In general, watermarking technology can be divided into two main types according to the purpose of its application: r〇bust watermarking technology and fragile watermarking technology. The tough watermarking technology is to verify the legitimacy of the intellectual property rights of digital multimedia by detecting the watermark, which must meet at least two basic requirements: transparency and robustness; that is, respected Even if the watermark is processed or deliberately falsified, it needs to retain the original embedded watermark features; the watermark can neither be easily removed and perceived, nor can the digital multimedia content be generated. Excessive distortion. Compared with the tough watermarking technology, the fragile watermarking technology is used to authenticate the content of multimedia (authenticati〇n). In other words, it is to detect whether the content of digital multimedia is subject to any Change and clearly indicate where the digital content has been tampered with. Current video or video compression standards, such as the JPEG and MPEG series, focus on architectures that use digital cosine transform (DCT) coding. This type of encoding uses DCT conversion, quantization, zigzag broom, volume encoding, and entropy encoding to remove redundant information from images or video. Although these codes provide high visual quality, convenient storage and propagation advantages, due to the processing functions of the current powerful digital multimedia editing software, it is easy for people to modify the digital media content, and thus digital multimedia integrity appears in the application. Hidden worry. In recent years, technology for embedding small amounts of digital information (ie, watermarking) to multimedia information for verification has received great attention from the business community and academia. This technology can be used to verify the integrity of digital multimedia content. And authenticity. In terms of an effective image watermarking technology architecture, the key requirements are low distortion, low computational complexity, low buffering requirements, and high sensitivity to tampering. Although many fragile watermarking techniques for verification purposes have been proposed, these fragile watermarking techniques have not actively focused on the issue of coding efficiency. However, coding efficiency is a space that affects the transmission time of the network and the storage of data. It is not negligible for digital multimedia communication. Therefore, there is a need for a new fragile watermarking technique that both verifies the integrity of digital multimedia and at the same time improves coding efficiency. SUMMARY OF THE INVENTION Accordingly, one aspect of the present invention is to provide a floating water 1335522 imprinting method having a compression function to produce a watermarked image having a smaller amount of data. According to an embodiment of the present invention, the method at least includes: providing an original image, wherein the original image includes a plurality of original image blocks, each original image block having a block intensity matrix, the block The intensity matrix is composed of the intensity values of a plurality of original image pixels, and the block intensity matrix represents the spatial distribution of the original image pixels; the watermark image is provided, wherein the watermark image system includes a plurality of watermark pixels The watermark pixels are corresponding to the original image block in a one-to-one manner; a Discrete Cosine Transform (DCT) step is performed on the block intensity matrix of each original image block to obtain a DCT coefficient matrix, wherein the DCT coefficient matrix comprises a plurality of DCT coefficients, each DCT coefficient represents a weighting value of each of the above-mentioned block intensity matrices in each frequency domain; and the quantization matrix is used to quantize the DCT coefficient matrix to Obtaining a matrix of quantized coefficients, wherein the matrix of quantized coefficients comprises a plurality of quantized DCT coefficients; the quantized coefficients are in a preset scan order The array is scanned to sequentially extract the quantized DCT coefficients and obtain one-dimensional image data; and perform a Run-length coding (RLC) step on the one-dimensional image data to obtain a plurality of quantities (Run- Length); using Run-length to calculate an odd number of quantized DCT coefficients to provide an odd number of total number values; and one of the above-described watermark pixels and one of the original image blocks Performing an embedding determining step of embedding the one of the watermark pixels to the one of the original image blocks, wherein the one of the watermark pixels has a color, and the one of the watermark pixels corresponds to the original image In the case of the block, the embedding determination step includes at least 7 1335522. When the color is the first color, the first determining step is performed to determine whether the odd number of total numbers is an odd number, if the odd number is When the total number value is an even number, a coefficient correction step is performed; and when the color is the second color, the second determining step is performed to determine whether the odd number of total number values are even numbers, if the odd number Total number is an odd number of coefficients of the above-described correction step is performed; and the quantized DCT coefficients of the lines comprising a plurality of non-zero quantized DCT coefficient is. Moreover, the coefficient correction step includes: performing a third determining step of determining whether the last non-zero of the non-zero quantized DCT coefficients is greater than 0 and providing a third determination result by using the length; and performing a parity change Steps to change the parity of the last non-zero according to the third determination result. Moreover, the second judgment result on the right side is: the above-mentioned parity change step is to reduce the value of the last non-zero by 1β. If the value of the last non-zero is just 1, the value of the last non-zero is changed to 0. The number of the above quantities is reduced by one. Further, if the third determination result is negative: the above-described parity change step is to increase the value of the last non-zero one. If the value of the last non-zero is just -1', the value of the last non-zero is changed to 0, and the number of the above-mentioned quantity is decreased by one. As can be seen from the above description, the present invention is characterized in that after obtaining the quantized DCT matrix of the original image block, the odd number of the quantized DCT coefficients included in the quantized DCT matrix is recorded, and the odd number is used. 8 1335522 The parity of the number value to represent the color of the embedded watermark image pixel. In an embodiment of the present invention, the odd number of total number values is an odd number representing that the color of the embedded watermark image pixel is black; the odd number of total number values is an even number representing the embedded watermark image pixel. The color is white. [Embodiment] Please refer to FIG. 1 and FIG. 2, FIG. 1 is a schematic diagram of watermark image embedding according to a first embodiment of the present invention, and FIG. 2 is a first embodiment of the present invention. A schematic diagram of the construction of the original image block, wherein the watermark image 110 comprises 8 by 8 watermark pixels 丨12, and each of the watermark pixels 112 can be, for example, black or white. The original image 丨 14 contains 64 by 64 original image blocks 116. Each original image block 116 contains 8 by 8 original image pixels 118, with each original image block 116 There is a block strength matrix 'This block strength matrix is composed of the intensity values (eg, grayscale values or chrominance values) of the original image pixels 118. The first embodiment of the present invention embeds the watermark pixels 112 into the original image block 116 in a one-to-one manner, wherein the watermark pixel U2a is used to embed the original image block 116a, and the watermark pixel n2b is Used to embed the original image block 116b, and so on. It is to be understood from the above description that the pixels included in the watermark image 110 and the original image 114 of the first embodiment of the present invention are not limited, and can be easily understood by those having ordinary knowledge in the art. Please refer to FIG. 3, which is a flow chart showing the embedding method of the 9 1335522 watermark according to the first embodiment of the present invention. Since the embedding method of each of the watermark pixels 112 is the same, in the following description, only the watermark pixel 112a and the original image block ii6a are exemplified.
在離散餘弦轉換步驟210中，係對原始圖像區塊116 之區塊強度矩陣進行離散餘弦轉換（Discrete Cosine Transform : DCT)以獲得DCT係數矩陣。在量化處理步驟 220中，係利用量化矩陣對DCT係數矩陣進行量化處理， 以獲得已量化係數矩陣’其中已量化係數矩陣係包含複數 個已量化DCT係數。在掃描步驟230中，以預設掃描順序 對已量化係數矩陣進行掃描，以依序取出已量化DCt係 數，並獲得原始圖像區塊1163之一維圖像資料。在量長編 碼步驟240中，對一維圖像資料進行量長編碼 C〇ding ; RLC)，以獲得複數個量長（Run-length)。在計算 步驟250中’利用量長來計算已量化dct係數為奇數的總 數目’並將其記錄為奇數個數總數目值。錢人判斷步帮In the discrete cosine transform step 210, Discrete Cosine Transform (DCT) is performed on the block intensity matrix of the original image block 116 to obtain a DCT coefficient matrix. In the quantization processing step 220, the DCT coefficient matrix is quantized using a quantization matrix to obtain a quantized coefficient matrix 'where the quantized coefficient matrix contains a plurality of quantized DCT coefficients. In a scanning step 230, the matrix of quantized coefficients is scanned in a predetermined scan order to sequentially extract the quantized DCt coefficients and obtain one-dimensional image data of the original image block 1163. In the quantity length encoding step 240, the one-dimensional image data is subjected to a length coding C〇ding; RLC) to obtain a plurality of run-lengths. In calculation step 250, 'the amount of utilization is used to calculate the total number of odd-numbered dct coefficients to be odd' and record it as an odd number of total number values. Money people judge step by step
26〇中係、對浮水印像素U2a和原始圖像區塊USa進行 嵌入判斷步驄。 °月參照第4圖’其係繪示根據本發明之第-實施例之 嵌入判斷步驟260的流鞋千立图木一 多A里多拉、社 流程不思圖。當洋水印像素112a之顏 :為，時，進行第_判斷步驟2仏 數目值是否為奇數，抬担糾仿 ^ Λ ^ ^ ^ 獒供第一判斷結果；若第一判斷結 果為否則進仃係數校正牛 g(| .u * « ^ _ 步驟264，若第一判斷結果為是， 則代表原始圖像區塊 β巧疋 * , ΤΓ ^ . J衣不黑色之洋水印圖像後 素，不進仃任何動作。 ^ I園诼像 田年夂印像素112a之顏色為白色 10 I335522 =進行第二判斷步驟262b中，以判斷奇數個數總數目 否疋：偶Γ ’並提供第二判斷結果;若第二判斷結果為 代表原始圖像區塊u6a可表示白色之料印 動作。在係數校正步驟樹，係將奇數個數： _之奇偶性質變更’以使奇數個數總數目值正確地表 示浮水印像素112a之顏色。In the middle, the embedding decision step is performed on the watermark pixel U2a and the original image block USa. Referring to Fig. 4, there is shown a flow shoe in the embedding judging step 260 according to the first embodiment of the present invention. When the color of the ocean watermark pixel 112a is YES, the _threquisition step 2 仏 the number value is an odd number, and the lifting correction ^ ^ ^ ^ 獒 is provided for the first judgment result; if the first judgment result is otherwise The coefficient correction cow g(| .u * « ^ _ step 264, if the first judgment result is YES, it represents the original image block β Qiao 疋*, ΤΓ ^ . J clothing is not black ocean watermark image, Do not enter any action. ^ I 诼 田 田 田 田 112 112 112 112 112 112 112 112 112 112 112 112 112 112 112 112 112 112 112 112 112 112 112 112 112 112 112 112 112 112 112 112 112 112 112 112 112 112 112 112 If the second judgment result is that the original image block u6a can represent the white printing action, in the coefficient correction step tree, the odd number: _ the parity property is changed to make the odd number total value correct. The ground indicates the color of the watermark pixel 112a.
凊同時參照第5圖，其係繪示根據本發明之第一實施 例之離散餘弦轉換212的計算示意圖。原始圖像區塊U6a 所包含之原始圖像像素118的強度值214(例如·灰階）可構 成區塊強度矩陣216，區塊強度矩陣216係代表原始圖像 像素118於空間域的分佈情況。接著，利用離散餘弦轉換 公式來將區塊強度矩陣216轉換成DCT係數矩陣218，其 中離散餘弦轉換公式為： m=0 λ=〇 16Referring to Fig. 5, there is shown a schematic diagram of the calculation of the discrete cosine transform 212 in accordance with the first embodiment of the present invention. The intensity value 214 (eg, gray scale) of the original image pixels 118 included in the original image block U6a may constitute a block strength matrix 216, and the block intensity matrix 216 represents the distribution of the original image pixels 118 in the spatial domain. . Next, the block strength matrix 216 is converted to a DCT coefficient matrix 218 using a discrete cosine transform formula, where the discrete cosine transform formula is: m = 0 λ = 〇 16
其中，c(〇 = < if ι = 0 ;c(/) 72 if i> 16 if i = 0 if i>\ F(*，0代表DCT係數矩陣218之每一元素之值； /(m，”)代表區塊強度矩陣216之每一元素之值； DCT係數矩陣218係包含複數個DCT係數219，DCT係 數219係代表原始圖像區境116a於頻率域之各頻率的加權 值。 11 1335522 請參照第6圖，其係繪示根據本發明之第一實施例之 • DCT係數矩陣的頻率分佈圖，其中位於矩陣DCT係數矩 • 陣218之第一列第一行之係數219a係代表區塊強度矩陣 216的直流分量係數。以係數219a為出發點向下、向右的 其它DCT係數，離係數219a越遠，其代表的頻率越高。 請參照第7圖，其係繪示根據本發明之實施例之量化 •處理222的計算示意圖。量化矩陣224係用以量化原始圖 . 像區塊116a於各頻率區域的分量，並以較大的量階來將代 • 表高頻分量之DCT係數的值縮小。利用量化矩陣224來對 DCT係數矩陣218進行計算可得到已量化係數矩陣226。 已量化係數矩陣226係包含複數個已量化DCT係數228。 請參照第8圖，其係繪示根據本發明之第一實施例之 掃描步驟230的示意圖。在本發明之第一實施例中，預設 掃描順序係為：從已量化係數矩陣226之左上元素232， 以Z字型方式（zigzag)掃描至右下元素234，其中左上元素 232係為已量化係數矩陣226之第一列第一行之元素，右 • 下元素234係為已量化係數矩陣226之最後一列最後一行 之元素。掃描步驟230係用以將已量化係數矩陣226轉換 為一維圖像資料。 請再參照第8圖，量長編碼步驟240係對一維圖像資 料進行量長編碼，以獲得複數個量長，其中每一量長係以 (x，y)來表示。若已量化DCT係數228包含有至少一個非零 • 已量化DCT係數，則y係代表每一非零已量化DCT係數 之值，X係代表每一非零已量化DCT係數之前的0的個 12 1335522 數。舉例而言，已量化係數矩陣226之一維圖像資料為： 56'12'7'3'4、〇、〇、2、1、〇、〇、........0。則量長為 (0’56) ' (0’12) ' (〇,7)、（〇,3)、（〇,4)、（2,2)和（〇，1)，並加上 EOB(end of block)於這些量長後端，以代表已量化係數矩 陣226編碼的標記。若已量化DCT係數228全為零，則直 接以EOB表示。 在計算步驟250中，係利用量長來計算已量化DCT 係數228的奇數個數，並將此數目記錄為奇數個數總數目 值。例如已量化DCT係數228之奇數為7、3和！，則奇 數個數總數目值為3« 在嵌入判斷步驟260中，若浮水印像素丨12a為黑色， 則進行係數校正步驟264。若浮水印像素112a為白色，則 不進行任何動作，因為奇數個數總數目值為3係代表原始 圖像區塊116a可顯示白色之浮水印圖像像素。 請同時參照第8圖、第9圖和第1〇圖，第9圖係繪 :根據本發明之第一實施例之係數校正步驟264的流程示 、圖第1 〇圖係繪示根據本發明之第一實施例之奇偶性 變更步驟268的流程示意圖。在係數校正步驟264中係 進行第二判斷步驟266，以利用量長來判斷已量化DCT係 數228之最後非零者是否大於〇，並提供第三判斷結果， 接著進行奇偶性變更步驟268，以根據第三判斷結果來變 更已量化DCT係數228之最後非零者的奇偶性質。在奇偶 性變更步驟268中，若第三判斷結果為否則進行第一奇偶 性變更步驟268a;若第三判斷結果為是則進行第二奇偶性 13 變更步驟268b，其中在第一奇偶性變更步驟268a中，係 將已量化DCT係數228之最後非零者之值加1 ;在第二奇 偶性變更步驟268b中，係將已量化DCT係數228之最後 非零者之值減1 ^在本實施例中，最後非零者3〇〇之數值 為1，因此將最後非零者30〇之值減i，以變更其奇偶性 質°又’因為最後非零者300之值剛好為1，減1之後可 變更為0，因此可減少量長的個數之個數。 另外’根據本發明之第二實施例，若最後非零者3〇〇 之值剛好為_1，將最後非零者300之值加1後，最後非零 者300之值亦變為〇，因此量長之個數也可減少。 請同時參照第11圖和第12圖，第11圖係繪示根據本 發明之第三實施例之原始圖像區塊之已量化係數矩陣，第 12圖係繪示根據本發明之第三實施例之嵌入判斷步驟66〇 的流程示意圖，其中原始圖像區塊116a之已量化係數矩陣 526之已量化DCT係數528全為〇。嵌入判斷步驟66〇係 類似於嵌入判斷步驟260,但因為已量化係數矩陣526之 已量化DCT係數528全為0，因此嵌入判斷步驟660更包 含直流（DC)係數變更步驟664。在直流係數變更步驟664 中，係將已量化係數矩陣526之直流係數變更為i，如此 已量化係數矩陣526之量長可表示為（oj)再加上E〇B來做 為編碼的標記，其中直流係數係為已量化係數矩陣526之 左上兀素532，左上元素532係位於已量化係數矩陣526 之第一列第一行。由於已量化係數矩陣526之量長可表示 為（〇，1)，因此可得到奇數個數總數目值為1，並據此進行 1335522 後續步驟。 由上述說明可知’本發明對於已量化0(：11係數全為〇 的原始圖像區塊亦可正確地嵌入其對應的浮水印圖像。 對於使用者而言，若想從已嵌入浮水印之圖像中取出 浮水印圖像時，使用者只需讀出奇數個數總數目值的奇偶 性，即可將浮水印圖像取出。下表係列示利用本發明之浮 水印嵌入方法所製作的已加浮水印圖像的資料壓縮比例。Where c(〇= < if ι = 0 ;c(/) 72 if i> 16 if i = 0 if i>\ F(*, 0 represents the value of each element of the DCT coefficient matrix 218; /(m , ") represents the value of each element of the block strength matrix 216; the DCT coefficient matrix 218 includes a plurality of DCT coefficients 219 representing the weighting values of the original image region 116a at frequencies in the frequency domain. 1335522 Please refer to FIG. 6 , which is a diagram showing the frequency distribution of the DCT coefficient matrix according to the first embodiment of the present invention, wherein the coefficient 219 a of the first row of the first column of the matrix DCT coefficient matrix 218 represents The DC component coefficient of the block strength matrix 216. Other DCT coefficients which are downward and rightward with the coefficient 219a as the starting point, the farther away from the coefficient 219a, the higher the frequency represented by it. Please refer to Fig. 7, which is based on A quantization diagram of the quantization/processing 222 of the embodiment of the invention. The quantization matrix 224 is used to quantize the original image. The image block 116a is a component of each frequency region, and the higher frequency component is used to represent the high frequency component. The value of the DCT coefficient is reduced. The quantization matrix 224 is used to input the DCT coefficient matrix 218. The calculated quantized coefficient matrix 226 is obtained. The quantized coefficient matrix 226 includes a plurality of quantized DCT coefficients 228. Referring to Figure 8, a schematic diagram of a scanning step 230 in accordance with a first embodiment of the present invention is shown. In the first embodiment of the present invention, the preset scanning order is: from the upper left element 232 of the quantized coefficient matrix 226, scanning in zigzag mode to the lower right element 234, wherein the upper left element 232 is quantized. The elements of the first row of the first column of the coefficient matrix 226, the right and lower elements 234 are the elements of the last row of the last column of the quantized coefficient matrix 226. The scanning step 230 is used to convert the matrix of quantized coefficients 226 into a one-dimensional map. For example, please refer to Fig. 8. The length encoding step 240 is to length encode the one-dimensional image data to obtain a plurality of lengths, wherein each length is represented by (x, y). The quantized DCT coefficients 228 contain at least one non-zero • quantized DCT coefficients, then y represents the value of each non-zero quantized DCT coefficient, and X represents the 12 of the zero before each non-zero quantized DCT coefficient. number. For example, the one-dimensional image data of the quantized coefficient matrix 226 is: 56'12'7'3'4, 〇, 〇, 2, 1, 〇, 〇, ........ 0. The quantity is (0'56) ' (0'12) ' (〇,7), (〇,3), (〇,4), (2,2) and (〇,1), plus EOB ( The end of block is at the back end of these quantities, with the marks encoded by the matrix of quantized coefficients 226. If the quantized DCT coefficients 228 are all zero, they are directly represented by EOB. In a calculation step 250, the odd number of quantized DCT coefficients 228 is calculated using the length, and this number is recorded as an odd number of total number values. For example, the odd number of quantized DCT coefficients 228 is 7, 3, and! Then, the odd number of total number values is 3 « In the embedding decision step 260, if the watermark pixel 丨 12a is black, the coefficient correction step 264 is performed. If the watermark pixel 112a is white, no action is taken because the odd number of total number values of 3 represents that the original image block 116a can display white watermark image pixels. Please refer to FIG. 8 , FIG. 9 and FIG. 1 simultaneously. FIG. 9 is a flow chart showing the coefficient correction step 264 according to the first embodiment of the present invention, and FIG. 1 is a diagram showing the present invention. A schematic diagram of the flow of the parity change step 268 of the first embodiment. In the coefficient correction step 264, a second determination step 266 is performed to determine whether the last non-zero of the quantized DCT coefficients 228 is greater than 〇, and provide a third determination result, followed by a parity change step 268 to The parity property of the last non-zero of the quantized DCT coefficients 228 is changed according to the third determination result. In the parity change step 268, if the third determination result is otherwise, the first parity change step 268a is performed; if the third determination result is yes, the second parity 13 change step 268b is performed, wherein the first parity change step is performed. In 268a, the value of the last non-zero of the quantized DCT coefficient 228 is incremented by one; in the second parity change step 268b, the value of the last non-zero of the quantized DCT coefficient 228 is reduced by one. In the example, the value of the last non-zero 3〇〇 is 1, so the value of the last non-zero 30〇 is reduced by i to change its parity property. And because the value of the last non-zero 300 is exactly 1, minus 1 Then it can be changed to 0, so the number of the number of lengths can be reduced. In addition, according to the second embodiment of the present invention, if the value of the last non-zero 3 刚 is just _1, and the value of the last non-zero 300 is increased by 1, the value of the last non-zero 300 is also changed to 〇. Therefore, the number of lengths can also be reduced. Please refer to FIG. 11 and FIG. 12 simultaneously. FIG. 11 is a diagram showing a matrix of quantized coefficients of an original image block according to a third embodiment of the present invention, and FIG. 12 is a third embodiment of the present invention. For example, a flow diagram of the embedding decision step 66〇, wherein the quantized DCT coefficients 528 of the quantized coefficient matrix 526 of the original image block 116a are all 〇. The embedding decision step 66 is similar to the embedding decision step 260, but since the quantized DCT coefficients 528 of the quantized coefficient matrix 526 are all zero, the embedding decision step 660 further includes a direct current (DC) coefficient changing step 664. In the DC coefficient changing step 664, the DC coefficient of the quantized coefficient matrix 526 is changed to i, and thus the length of the quantized coefficient matrix 526 can be expressed as (oj) plus E〇B as the encoded mark. The DC coefficient is the upper left pixel 532 of the quantized coefficient matrix 526, and the upper left element 532 is located in the first row of the first column of the quantized coefficient matrix 526. Since the length of the quantized coefficient matrix 526 can be expressed as (〇, 1), an odd number of total number values of 1 can be obtained, and the subsequent steps 1335522 are performed accordingly. It can be seen from the above description that the present invention can correctly embed the corresponding watermark image for the original image block which has been quantized 0 (the 11 coefficients are all 。. For the user, if the user wants to embed the watermark from the watermark When the watermark image is taken out from the image, the user only needs to read the parity of the odd number of total number values, and the watermark image can be taken out. The following series shows the watermark embedding method of the present invention. The data compression ratio of the added watermark image.
浮 水印嵌入前 浮水印嵌入後 — 原始圖像 資料量 峰值信號雜訊比 (PSNR) 資料量 峰值信號雜訊比 (PSNR) 壓縮率 圖樣1 19.4kb 37.52 18.4kb 37.02 5.15% 圓樣2 19.4kb 24.24 17.8kb 23.97 8.24% 圓樣3 19.4kb 33.78 18.2kb 33.60 6.19% 圖樣4 19.4kb 36.16 18.6kb 35.64 4.12% 圖樣5 19.4kb 39.11 18.1kb 38.65 6.70% 圖樣6 19.4kb 35.50 17.9kb 35.11 7.73% (表一） 由表一可知，雖然本發明之浮水印嵌入方法雖會造成 些微的失真，卻可得到不低的壓縮率。 本發明之浮水印嵌入方法係為易碎型浮水印技術，因 此本發明並不受限圖像上之使用，其它種類的資料（例如視 訊）皆可使用本發明來防偽，並壓縮其資料量。 雖然本發明已以實施例揭露如上，然其並非用以限定 15 1335522 本發明，任何熟習此技藝者，在不脫離本發明之精神和範 圍内，當可作各種之更動與潤飾，因此本發明之保護範圍 當視後私之申請專利範圍所界定者為準。 【圖式簡單說明】 為讓本發明之上述和其他目的、特徵、和優點能更明After the watermark is embedded, the watermark is embedded. The original image data volume peak signal noise ratio (PSNR) data volume peak signal noise ratio (PSNR) compression rate pattern 1 19.4kb 37.52 18.4kb 37.02 5.15% circle sample 2 19.4kb 24.24 17.8kb 23.97 8.24% circle 3 19.4kb 33.78 18.2kb 33.60 6.19% pattern 4 19.4kb 36.16 18.6kb 35.64 4.12% pattern 5 19.4kb 39.11 18.1kb 38.65 6.70% pattern 6 19.4kb 35.50 17.9kb 35.11 7.73% (Table 1) It can be seen from Table 1 that although the watermark embedding method of the present invention causes slight distortion, a low compression ratio can be obtained. The watermark embedding method of the present invention is a fragile watermarking technology, so the invention is not limited to the use of images, and other kinds of materials (such as video) can use the invention to prevent anti-counterfeiting and compress the amount of data. . While the present invention has been described above by way of example only, it is not intended to limit the scope of the present invention, and the present invention may be modified and modified without departing from the spirit and scope of the invention. The scope of protection shall be subject to the definition of the patent application scope. BRIEF DESCRIPTION OF THE DRAWINGS The above and other objects, features and advantages of the present invention will become more apparent.
顯易懂，上文特舉一較佳實施例，並配合所附圖式，作詳 細說明如下： 第1圖係繪不根據本發明之第一實施例之浮水印嵌入 示意圖。 第2圖係緣示根據本發明之第一實施例之原始圖像區 塊之構造示意圖。 第3圖係繪示根據本發明之第一實施例之浮水印之後 入方法的流程示意圖。 第4圖係繪示根據本發明之第一實施例之嵌入判斷步 驟的流程示意圖。It is to be understood that the preferred embodiment of the invention has been described in detail with reference to the accompanying drawings in which: FIG. 1 is a schematic diagram of a watermark embedding not according to the first embodiment of the present invention. Fig. 2 is a view showing the construction of an original image block according to the first embodiment of the present invention. Fig. 3 is a flow chart showing the method of the watermarking back-in method according to the first embodiment of the present invention. Fig. 4 is a flow chart showing the embedding judging step according to the first embodiment of the present invention.
第5圖係繪示根據本發明之第一實施例之離散餘弦轉 換的計算示意圖。 第6圖係繪轉據本發明之第—實施狀dct係數矩 陣的頻率分佈圖。 第7圖係緣示根據本發明之實施例之量化處理的計算 示意圖。 第8圖料示根據本發明之第—實施例之掃描示意 圖0 16 1335522 第9圖係繪示根據本發明之第一實施例之係數校正步 驟的流程示意圖。 第10圖係繪示根據本發明之第一實施例之奇偶性變 更步驟的流程示意圖。 第11圖係繪示根據本發明之第三實施例之原始圖像 區塊之已量化係數矩陣。 第12圖係繪示根據本發明之第三實施例之嵌入判斷 步驟的流程示意圖。 【主要元件符號說明】 110 :浮水印圖像 112 :浮水印像素 112a :浮水印像素 112b :浮水印像素 114 :原始圖像 116 :原始圖像區塊 116a :原始圖像區塊 116b :原始圖像區塊 118 :原始圖像像素 210 :離散餘弦轉換步驟 212 :離散餘弦轉換 216:區塊強度矩陣 218 : DCT係數矩陣 219 : DCT 係數 219a :係數 220 :量化處理步驟 222 :量化處理 224 :量化矩陣 226 :已量化係數矩陣 228 :已量化DCT係數 230 :掃描步驟 232 :左上元素 234 :右下元素 240 .莖長編碼步驟 250 :計算步驟 260 :嵌入判斷步驟 262a :第—判斷步驟 262b :第二判斷步驟 17 1335522 264 :係數校正步驟 268 :奇偶性變更步驟 268b :第二奇偶性變更步驟 526 :已量化係數矩陣 532 :左上元素 660 :嵌入判斷步驟 266 :第三判斷步驟 268a :第一奇偶性變更步驟 300 :最後非零者 528 :已量化DCT係數 664 :直流係數變更步驟Fig. 5 is a view showing the calculation of discrete cosine transform according to the first embodiment of the present invention. Fig. 6 is a diagram showing the frequency distribution of the dct coefficient matrix according to the first embodiment of the present invention. Fig. 7 is a schematic diagram showing the calculation of the quantization process according to an embodiment of the present invention. Figure 8 is a schematic view showing the scanning of the first embodiment of the present invention. Figure 9 is a schematic diagram showing the flow of the coefficient correcting step according to the first embodiment of the present invention. Figure 10 is a flow chart showing the steps of the parity change according to the first embodiment of the present invention. Figure 11 is a diagram showing a matrix of quantized coefficients of an original image block in accordance with a third embodiment of the present invention. Figure 12 is a flow chart showing the embedding judging step according to the third embodiment of the present invention. [Main component symbol description] 110: Watermark image 112: Watermark pixel 112a: Watermark pixel 112b: Watermark pixel 114: Original image 116: Original image block 116a: Original image block 116b: Original image Image block 118: original image pixel 210: discrete cosine transform step 212: discrete cosine transform 216: block strength matrix 218: DCT coefficient matrix 219: DCT coefficient 219a: coefficient 220: quantization process step 222: quantization process 224: quantization Matrix 226: Quantized Coefficient Matrix 228: Quantized DCT Coefficient 230: Scanning Step 232: Upper Left Element 234: Lower Right Element 240. Stem Length Encoding Step 250: Calculation Step 260: Embedding Decision Step 262a: Step - Judgment Step 262b: Second determination step 17 1335522 264: coefficient correction step 268: parity change step 268b: second parity change step 526: quantized coefficient matrix 532: upper left element 660: embedding decision step 266: third decision step 268a: first parity Sex Change Step 300: Last Non-Zero 528: Quantized DCT Coefficient 664: DC Coefficient Change Step
1818