200849135 九、發明說明： 【發明所屬之技術領域】 本發明係一種加入與取出浮水印的 於高動態範圍影像中之加人與取出浮水印的方^'疋一種 【先前技術】 1旦1真貫世界中某一景物於某一特定時刻的輻射分佈鱼 L物於该時刻所拍攝而得的數位相 度之-像素值存在著一個非線性的對 =色… == 圍（dynamlcrange)有限，當幅射低於 /、、Η , %在數位相片中即無法顯現其差異，而只能 白以取Θ日狀悲表不之；同樣地，當輻射強過某_臨界值時， 數位，片也只能以最亮狀態來呈現該像素的強纟。所謂 “動態範圍”京尤是指一個景或影像中所能表現出之最Α : 最暗的對比值。遂有學者提出藉由結合多張相同景物^ 同曝光時間的照片來產生所謂的『高動態範圍（Hlgh200849135 IX. Description of the Invention: [Technical Field] The present invention is a method for adding and removing watermarks in a high dynamic range image by adding and removing watermarks. [Prior Art] 1 The distribution of the radiation of a certain scene in a certain scene in a certain moment at a particular moment. The pixel value of the pixel L at that moment has a nonlinear pair = color... == dynamlcrange is limited, When the radiation is lower than /, Η, % can't show the difference in the digital photo, but can only be used to take the sorrow and grief; likewise, when the radiation is stronger than a certain _threshold, the number, The slice can only present the power of the pixel in the brightest state. The so-called "dynamic range" Jing especially refers to the most ambiguous in a scene or image: the darkest contrast value. Some scholars have proposed to create a so-called "high dynamic range (Hlgh by combining multiple photos of the same scene with the same exposure time).
Dynannc Range)影像』’來解決—般數位相片動態範圍不 足的問題。 數位浮水印在多媒體版權的保護、鑑定及竄改驗證上 已成為一個具潛力且有效的工具。當數位浮水印加在一數 :相片時，最佳的效果是讓使用者無法察覺該數位相片内 含有洋水印；當該數位相片遭冒用而產生數位相片的版權 爭議時，為了確認該數位相片的版權所屬，即可利用特定 的項取手段，取出隱藏於該數位相片中的浮水印，以驗證 200849135 該數位相月的版權，·也就是說，由該數 印必須清楚且可辨認，才出的净水 曰 令Ά L丁、上的使用效益。纨 =Γ相片者”會利用剪裁、縮放、幾何變：影 Ε、、、“色㈣應寻手段企圖破壞該數位相片内藏的心 卜以減弱該浮水印料辨 、吁水 . 口此在η際應用屏 ρ的強動度則是一個關鍵的議題。-般㈣、、心， 兩種不同類型的方法來對抗幾 =水印有 ： ）以及非盲的演算法。所謂非盲的演瞀， :疋在㈣浮水印時，需先取得該數位相片的 加入浮水印）來做計算，因此該方法可以在：：(未 影像和未被攻擊的影像中t 在被成何攻擊的 而所謂盲的演算法，就是在浮水印决問喊。 像，I ft痄^ 一 y t不月匕使用原始 /、難度杈南，但是也因為盲的演瞀味τ兩 影像而在使用上較A ^ /而要使用原始 用。 車乂為方便’因此該類演算法受到廣泛的應 由於高動態範圍影像的合成 高，因此，比一般低動r ^ R Α 易衫像價值相對較 若能將… 圍的影像具有更高的保護價值， 右此將剧述的浮水印技術應用於一 、值 。動4靶圍影像的保護具有相當 ’、“亥 提出有很多不π M ^ y > 良的效果。惟目前雖然 ’很夕不同的數位洋水印技 適用於一船侦叙…-闻取 仁疋该些技術都是僅 般低動怨乾圍影像，例如灰皇 全彩影像等，係因為目前低 =或2 4位元之 之間的敕叙 国象的像素值為0〜255 間的整數，而高動態範圍影像 '55 不固定，因此低動態範圍影像專用^^序點數且範圍 用的汗水印技術直接嵌入 200849135 高動態範圍影像並不適宜且不適用，戶斤以目前尚未存有適 :方、：動悲乾圍影像的浮水印嵌入及取出的方法。另外， :動f範圍影像與低動態範圍影像最大的差異，就是瀏覽 阿動悲乾圍影像的時候一定要經過色調對映(τ· Μ卿㈤) 、力作才可將5亥咼動態範圍影像展示於一般的顯示輸出裝 $如果將現有的浮水印技術直接藏入該高動態範圍影像 大’、只要一經過色調對應的破壞之後，浮水印勢必會消失， ，為現有的低動態範圍影像之浮水印技術均是針對抗雜 _又形裁切等破壞加以實作的，所以無法抵抗高動態 軌圍影像特有的色調對應之破壞。 【發明内容】 為解决既有/于水印加入與取出方法不適用於高動態範 象㈣H H種高動態範圍影像之強健浮水 印的H其包含-嵌人浮水印的手段，包含步驟： 弟 色调對應，係制* .一古命》能々々ΓΕΠ ,Λ ^ ^ ^ ^ 回動悲轭圍影像進行色調對應 俊形成一低動態範圍影像； 圍旦^ t〜像比例H該高動態範圍影像與該低動態範 圍-像之像素值比值而形成—比例影像； 今低動^水印’ u —洋水印嵌人運算將—浮水印嵌入至 範圍影像，並形成—藏有浮水印的低動態範圍影 I，U及 比例a :相乘將°亥藏有洋水印的低動態範圍影像以及該 ，像進行相乘後形成一嵌有浮水印的高動態範圍影 200849135 進—步地，該高動態範圍影像 > 含一=出浮水印的手段，其包含步驟㈣以印的方法包 第二色調對應，將該嵌有浮水印的 行色調對應後形成—藏有浮水印的低動;Vf㈣影像進 洋水印取出，以一浮水印取 /已圍衫像’以及 浮水印的低動態範圍影像之該浮水印取异出將内藏於該藏有 藉此，透過該嵌入浮水印的牛驟 影像加入哕、、I 7jc ep . V ”、A，可對該高動態範圍 …二 透過該取出浮水印的步驟，可對該 汗7 p的尚動態範圍影像取出該浮水印。 【實施方式】 發明：動：弟圖、弟二圖、第三圖以及第四圖，係為本 :二動恶範圍影像之強健浮水印的方法以及實施步驟示 思圖，包含-嵌人浮水印⑴）的手段以及—取出浮水印⑽ 的手段之較佳實施步驟。 L· 口月 > 考第一圖及第二圖’該嵌入浮水印（工〇)之步驟包 3 ·色调對應（12 )、計算影像比例（13 )、加入浮水印 (1 4 )以及影像相乘（1 5 )。 別述之色調對應（12 )步驟中，其係先取一高動態範 圍影像（HDR)進行色調對應而形力24位元之一低動態範 圍影像（LDR)，而該色調對應（T〇ne腿卯丨叩）係將該動 恶範圍影像（HDR )之動態範圍以一特定之對應關係轉換形 成該低動態範圍影像（LDR)，其中，該特定之對應關係可 依據需求為線性或非線性之對應。 200849135 前述之計算影像比例（13)步驟中，係計算該高動態 摩巳圍影像（HDR )以及該低動態範圍影像（LDR )之比例影 像（ratio image)，亦即該高動態範圍影像（hdr)與經 過色調對應後產生之該低動態範圍影像（LDR )之影像像素 值之比值，而形成之一比例影像（RI )。 月ίι述之加入浮水印（14 )步驟中，係將該低動態範圍 影像加入一浮水印（W )，其中加入該浮水印（w )的方式 係採用簡單的浮水印嵌入技術，先將該低動態範圍影像 (DR)作难放餘弦轉換（Discrete Cosine Transform，DC 丁） 之運算，使該低動態範圍影像（LDR )之影像像數值之訊號 由日寸域轉換為頻域，其中，頻域係指訊號變化的程度，越 南頻表示訊號變動越大，在影像方面就是屬於物體邊界的 部份’相對的，越低頻表示訊號越平滑，從影像上來說就 疋平滑的背景。轉換後，即可以分析出該低動態範圍影像 (LDR )的高低頻；之後，選擇該低動態範圍影像（) 之中低頻（即DC7；3 )的位置將該浮水印（w )藏入，即可 以維持該浮水印（W )的強韌性同時保持該低動態範圍影 像（LDR )此一原始影像的資訊，使該浮水印（w )可以抵 抗一般的糢糊化破壞與雜訊的干擾。 另外，為了在取出浮水印時不需要用到來源的影像， 將该浮水印（W )加入離散餘弦轉換後之該低動態範圍影 像（LDR )係採用一鄰近區塊大小判斷法完成，其係先將轉 換後之低動態範圍影像（LDR)預定用來藏入該浮水印（w) 之一區塊DC7；3 ( X )與鄰近一區塊( γ )進行比較， 200849135 若所藏人的浮水印⑴資訊為丨日卜則將原來該區塊 DCJ^u)加入一特定數值後形成—新區塊^^(χ，）， 使該新區塊Z)CT ( X ’）盤該Γ)ΓΤ卩v \ , i 3,3以Π從Γ24 ( Y )的數值狀態為X，〉 Υ，反之，右所藏入的浮水印（w)資訊為〇時，則將 原來該區塊叫3⑴減去一特定數值後形成一新區塊 DCr3,3 (X，），而使該新區塊DC7；3 (X，）與該( γ)的 數值狀態為χ，<γ。最後，再將藏入該浮水印⑴的轉換 广後之低動恶範圍影像（LDR )進行一次離散餘弦轉換之逆轉 換，即可得到一藏有浮水印的低動態範圍影像（LD㈣）。 丽述影像相乘（15)步驟中，其係取該比例影像（RI) 與该藏有浮水印的低動態範圍影像（LDRW )影像相乘後， 即可得到一肷有浮水印的高動態範圍影像（仙⑽）。 請參考第二圖及第四圖，該取出浮水印（2〇)手段包含： 色調對應（22)以及浮水印取出（24)。 該色調對應（22 )步驟中，係將該嵌有浮水印的高動 1；態範圍影像（HDRW )進行色調對應而形成該藏有浮水印的 低動態範圍影像（LDRW )。 該浮水印取出（24 )步驟中，係先將該藏有浮水印的 低動態範圍影像（LDRW )進行離散餘弦轉換後，取得該藏 有净水印的低動態範圍影像（LDRW )轉換後其一結果區塊 DCKXX)與一結果鄰近區塊乃C7^(YY)的值並進行比較， 當XX >= YY時，則判定該結果區塊( χχ )的浮水印 (w)之資訊為1 ;反之，當XX < γγ時，則判定該結果區 塊DC7；3 ( XX )的浮水印（w )的資訊為〇，如此，即可取出 10 200849135 該浮水印（W)。Dynannc Range) "Image" to solve the problem of insufficient dynamic range of digital photos. Digital watermarking has become a promising and effective tool for the protection, identification and tampering of multimedia copyrights. When the digital watermark is added to a number: photo, the best effect is that the user cannot detect that the digital photo contains a foreign watermark; when the digital photo is fraudulently used to generate a copyright dispute of the digital photo, in order to confirm the digital photo The copyright belongs to, and the specific item selection means can be used to take out the watermark hidden in the digital photo to verify the copyright of the digital phase of 200849135, that is, the number must be clear and identifiable. The use of clean water to make Ά L, the use of benefits.纨=ΓPhotographer will use the cropping, zooming, geometrical changes: shadows, and, “color (4) should find ways to destroy the heart of the digital photo to weaken the watermarking material, call the water. The strongness of the η application screen ρ is a key issue. - (4), heart, two different types of methods to combat a few = watermarks have :) and non-blind algorithms. The so-called unblinded deduction, 疋 ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( The so-called blind algorithm is attacked by the watermark. Like, I ft痄^ a yt does not use the original /, difficulty south, but also because of the blind interpretation of the two images In use, it is better to use A ^ / but to use the original. The rut is convenient. Therefore, this kind of algorithm is widely accepted due to the high synthesis of high dynamic range images, so it is lower than the general low motion r ^ R Α Relatively speaking, if the image of the surrounding image has a higher protection value, the watermarking technique of the script will be applied to the value of one. The protection of the target image of the moving 4 has a considerable ', and the number of the proposed image is quite different. ^ y > Good effect. However, although the current digital watermarking technique is suitable for a ship's reconnaissance...--there are some techniques that are only low-moving and screaming, such as the gray emperor. Color image, etc., because it is currently low = or between 2 4 bits The pixel value of the national image is an integer between 0 and 255, while the high dynamic range image '55 is not fixed, so the low dynamic range image is dedicated to the number of points and the range of sweat watermarking technology is directly embedded in 200849135 high dynamic range. The image is not suitable and is not applicable. The method of embedding and removing the watermark of the image is not yet available. In addition, the maximum difference between the image of the dynamic f range and the image of the low dynamic range is It is necessary to go through the tonal mapping (τ· Μ ( (5)) when browsing the image of Ai sorrow and sorrow, and then to display the 5 咼 dynamic range image in the general display output. If the existing watermarking technology is directly Hiding the image in the high dynamic range is large, and as long as the color is correspondingly destroyed, the watermark will disappear. The existing low dynamic range image watermarking technology is aimed at the anti-aliasing and the cutting. In practice, it is unable to resist the damage of the color tone corresponding to the high dynamic track image. [Summary of the Invention] The method for adding/removing existing/adding watermarks is not suitable for high dynamic range. The method of (4) H H high-dynamic range image robust watermarking H contains - embedded watermarking, including the steps: brother color tone, system *. An ancient life can, Λ ^ ^ ^ ^ back The sorrowful yoke image is subjected to a hue correspondence to form a low dynamic range image; the surrounding image is formed by the ratio of the high dynamic range image and the low dynamic range-image pixel value ratio; The watermark 'u—the ocean watermark embedding operation—embeds the watermark into the range image and forms—the low dynamic range shadow I with the watermark, U and the ratio a: multiplying the low dynamic range of the ocean watermark The image and the image are multiplied to form a high dynamic range image embedded in the watermark 200849135. The high dynamic range image includes a method of watermarking, including the step (4) by printing The second tone corresponding to the packet is formed, and the watermark embedded with the watermark is correspondingly formed to form a low-motion with a watermark; the Vf (four) image is taken out into the ocean watermark, and the watermark is taken as a watermark/the watermark is low and the watermark is low. The watermark of the dynamic range image The escaping will be hidden in the possession, and the step of removing the watermark can be performed by adding the 浮, I 7jc ep . V ”, A through the embedded watermark image. The watermark can be taken out from the still dynamic range image of the sweat 7 p. [Embodiment] Invention: Movement: brother diagram, brother second diagram, third diagram and fourth diagram, is based on: the method of robust watermarking of the two motion range image and the implementation step diagram, including - embedded floating The means of watermarking (1)) and the preferred implementation steps of means for extracting the watermark (10). L· 口月> Test the first picture and the second picture 'The step of embedding the watermark (work) 3 · Tone correspondence (12), calculating the image ratio (13), adding the watermark (1 4 ) and the image Multiply (1 5 ). In the color matching (12) step, the high dynamic range image (HDR) is first taken to perform the hue correspondence and the shape is 24 bits low dynamic range image (LDR), and the hue corresponds (T〇ne leg卯丨叩) converting the dynamic range of the moving range image (HDR) into a specific correspondence relationship to form the low dynamic range image (LDR), wherein the specific correspondence may be linear or nonlinear according to demand correspond. 200849135 In the foregoing step of calculating the image ratio (13), calculating the high dynamic surround image (HDR) and the ratio image of the low dynamic range image (LDR), that is, the high dynamic range image (hdr) A ratio image (RI) is formed by a ratio of image pixel values of the low dynamic range image (LDR) generated after the hue is corresponding. In the step of adding the watermark (14), the low dynamic range image is added to a watermark (W), and the method of adding the watermark (w) is a simple watermark embedding technique. The low dynamic range image (DR) is operated by the Discrete Cosine Transform (DC D), so that the signal of the image value of the low dynamic range image (LDR) is converted from the day field to the frequency domain, wherein the frequency The domain refers to the degree of signal change. The frequency of Vietnam indicates that the signal changes more. In terms of image, the part that belongs to the boundary of the object is 'relative. The lower the frequency, the smoother the signal, and the smoother background from the image. After the conversion, the high and low frequencies of the low dynamic range image (LDR) can be analyzed; after that, the position of the low frequency (ie, DC7; 3) in the low dynamic range image () is selected to hide the watermark (w). That is, the toughness of the watermark (W) can be maintained while maintaining the information of the original image of the low dynamic range image (LDR), so that the watermark (w) can resist general blurring damage and noise interference. In addition, in order to extract the watermark without using the source image, the low dynamic range image (LDR) after the watermark (W) is added to the discrete cosine transform is completed by a neighboring block size judgment method. The converted low dynamic range image (LDR) is first reserved for hiding in one of the watermarks (w) blocks DC7; 3 (X) is compared with a neighboring block (γ), 200849135 The watermark (1) information is that the original block DCJ^u) is added to a specific value to form a new block ^^(χ,), so that the new block Z)CT (X ') is the same)卩v \ , i 3,3 is the value of Γ24 ( Y ) as X,> Υ, otherwise, if the watermark (w) information hidden in the right is 〇, then the original block is called 3(1) minus After a certain value is obtained, a new block DCr3, 3 (X,) is formed, and the numerical state of the new block DC7; 3 (X,) and the (γ) is χ, < γ. Finally, the low-motion range image (LDR) hidden in the converted watermark (1) is inversely transformed by a discrete cosine transform to obtain a low dynamic range image (LD(4)) with a watermark. In the image multiplication (15) step, the proportional image (RI) is multiplied by the watermarked low dynamic range image (LDRW) image to obtain a high dynamic of watermarking. Range image (sen (10)). Referring to the second and fourth figures, the means for taking out the watermark (2) includes: a tone correspondence (22) and a watermark extraction (24). In the tone matching (22) step, the watermark-embedded high-range image (HDRW) is subjected to hue correspondence to form the watermark-containing low dynamic range image (LDRW). In the step of taking out the watermark extraction (24), the low dynamic range image (LDRW) with the watermark is first subjected to discrete cosine transform, and then the low dynamic range image (LDRW) with the stored watermark is obtained and converted. A result block DCKXX) is compared with a result neighboring block value of C7^(YY), and when XX >= YY, the information of the watermark (w) of the result block (χχ) is determined as 1; conversely, when XX < γ γ, it is determined that the information of the watermark (w ) of the result block DC7; 3 ( XX ) is 〇, so that the watermark (W) of 10 200849135 can be taken out.
U t證本發明之較佳實施方法確實可行，請參考第五 β弟/、圖’其中第五圖係為-張浮水印⑴，而第六 圖Α則為:張影像解析大小為640x480的低動態範圍影像 LDR ’其係經由-高動態範圍影像（_)經過色調對 應㈣結果，而第六圖⑻則為將第三圖之浮水印⑺依 據前i色調對應、（12)及加入浮水印（⑷步驟加入於該 圍⑤像（LDR)而獲得—藏有浮水印的低動態範圍 :像旦:LDRW ) ’比較第六圖（A ) ( B )即可發現該低動態 範圍:像（LDR)與_藏有浮水印的低動態範圍影像（LDRW) 在視2上幾乎看不出有差異，且該低動態範圍影像⑽" 與該藏有浮水印的低動態範圍影像（LDRW )二者之一峰值 汛 5虎雜。fi 比（peak Slgnal t〇 N〇ise 衍〇，)為 dB /、中，δ亥峰值訊號雜訊比（PSNR )的計算公式如下： PSNR x\og]0(2552 /MSE) ^ 其中 mse=^L^m^\y Μ 乂 Nh、Xi — MXN為影像解析大小， X代表原始影像像素， Y代表經過破壞後的影像像素。 再者’可將第六圖（Β )之藏有浮水印的低動態範圍 影像（LDRW)與該計算影像比例（丨3 )步驟中取得之比例 影像（RI )進行該影像相乘（1 5 )步驟，而可得到該嵌有 浮水印的高動態範圍影像（HDRW)。而為了驗證該嵌有浮 11 200849135 水印的南動態節圖旦/ 。把圍衫像（HDRW)具有高品質、低失真的優 點，將該嵌有浮：火Pn [ ^ p的高動態範圍影像（hdrw )的像素值 線性對應到〇〜2 5 ^$ » 冉鼻出其峰值訊號雜訊比（PSNR )值， 其結果為50,76dft η 因此’證實依據本發明之嵌入浮水印 (1 〇 )手段所形成之山 a ^ _ 肷有洋水印的高動態範圍影像（HDRW ) 具有而品質、你it古，， 貝低失真的特點。 為驗證該取屮、、全^ / 7 f7 ( 20)手段之效果，請再參考第 七圖以及弟八圖，i 古動能μ ^ ，弟七圖（A )為一嵌有浮水印的 冋動悲靶圍影像（阶㈣ 驟執行後所得的藏有/ 由該色調對應（22)步 如第七圖⑻所亍/ 低動態範圍影像（LDRW)即 取出的浮水印^)=前述該浮水印（24)步驟執行後所 印（W)雖有…广弟八圖所示，报明顯地，該浮水 φ, 到輕微的破壞，但是還是可以清楚1別 出浮水印（w)的寺〜 月是識別 ^ )子，此即證明該嵌入浮水印（1(Π Μ 4 段以及一取出滚^ | u u)的手 / 7 p ( 20)的手段之較佳實施步n砝每了 達到發明目的。 貝她步驟確貫可 另外，接著為了驗證該嵌入浮 取出浮水印（2fn主a I ( 手段以及一 手严… 的強健性，對該嵌入浮水印U。） 手奴執订後所得的嵌有浮水印的 像（贿）進行包含裁切、高㈣糊、㈣圍影 色調對應等不同之凑吉 ’、 不同茶數的 手段還原該浮水印（ 取出吁水印（20) 1 u J衫像，其結果如當士回 所示，因此，可试與 圖及第十圖 不同類型的失真破壞之效能。 …、有強健之抵抗 12 200849135 【圖式簡單說明】 第一圖係為本發明一嵌入这 、 ♦、月 予水印的手段之較私每 驟流程圖。 土 a施步 弟一圖係為本發明一取出浮水印 驟流程圖 的手段之較佳實施步U t certificate The preferred implementation method of the present invention is indeed feasible, please refer to the fifth β brother /, the figure 'the fifth picture is - Zhang watermark (1), and the sixth picture is: the image resolution size is 640x480 The low dynamic range image LDR 'is passed the tone-corresponding (four) result via the -high dynamic range image (_), while the sixth figure (8) is the watermark of the third image (7) according to the front i tone correspondence, (12) and the floating Watermark ((4) step is added to the surrounding image (LDR) - low dynamic range with watermark: dan: LDRW) 'Comparing the sixth picture (A) (B) to find the low dynamic range: like (LDR) and _ low watershed image (LDRW) with watermarking are almost indistinguishable from view 2, and the low dynamic range image (10)" and the low dynamic range image (LDRW) with watermarking One of the peaks is 汛5. The fi ratio (peak Slgnal t〇N〇ise 〇,) is dB /, medium, δ Hz peak signal noise ratio (PSNR) is calculated as follows: PSNR x \ og] 0 (2552 / MSE) ^ where mse = ^L^m^\y Μ 乂Nh, Xi — MXN is the image resolution size, X represents the original image pixel, and Y represents the corrupted image pixel. Furthermore, 'the low-dynamic range image (LDRW) with the watermark stored in the sixth picture (Β) can be multiplied by the ratio image (RI) obtained in the step of calculating the image ratio (丨3) (1 5 The step of obtaining a high dynamic range image (HDRW) embedded with a watermark. In order to verify the South Dynamic Section of the watermark embedded in the 200849135 watermark. The Sweater (HDRW) has the advantages of high quality and low distortion, and the pixel value of the high dynamic range image (hdrw) embedded with the float: fire Pn [ ^ p is linearly mapped to 〇~2 5 ^$ » The peak signal to noise ratio (PSNR) value is obtained, and the result is 50, 76 dft η, thus confirming the high dynamic range image of the mountain a ^ _ 洋 with ocean watermark formed by the method of embedding watermark (1 〇) according to the present invention. (HDRW) has the characteristics of quality, you are ancient, and low distortion. In order to verify the effect of the method of taking 屮, 全 ^ / 7 f7 ( 20), please refer to the seventh figure and the buddha figure, i ancient kinetic energy μ ^, and the younger seven figure (A) is a watermark embedded 冋The image of the sorrowful target (the result of the step (4) is stored/received by the color tone (step 22), as shown in the seventh figure (8), the low-dynamic range image (LDRW) is taken out of the watermark ^) = the aforementioned float After the execution of the watermark (24) step (W), although there is a wide picture shown in the broad picture of the younger brother, it is obvious that the floating water φ, to a slight damage, but still clear 1 to the floating watermark (w) of the temple ~ The month is the identification ^), which proves that the method of embedding the watermark (1 (Π Μ 4 segment and a take-out roll | uu) hand / 7 p ( 20) is a preferred implementation step n 砝 every invention The purpose of the beta step can be additional, and then in order to verify the embedded floating watermark (2fn master a I (the means and the strongness of one hand... the embedded watermark U.) A watermarked image (bribe) carries a hand that contains cuts, high (four) pastes, (four) hue tones, and so on. Restore the watermark (take out the watermark (20) 1 u J shirt image, the result is as shown in the back, so you can try different types of distortion damage performance with the map and the tenth map. ..., with strong resistance 12 200849135 [Simple description of the diagram] The first diagram is a relatively private flow chart of the method for embedding this, ♦, and monthly watermarks. The soil a Shi Budi is a flow chart of the invention. Preferred implementation of the means
第二圖係為本發明嵌入浮水印的 之實施示意圖。 手段之較佳 只施步 第四圖係為本發明取出浮水印的手段之較佳 之實施示意圖。 第五圖係為一浮水印示意圖。 第六圖（A )係為一低動態範圍影像。 貫施步驟The second figure is a schematic diagram of the implementation of embedding a watermark in the present invention. Preferably, the fourth embodiment is a preferred embodiment of the method for extracting a watermark from the present invention. The fifth figure is a schematic diagram of a watermark. The sixth picture (A) is a low dynamic range image. Steps
第/、圖（B )係為該第六圖（a )加入浮水印之土 ^ . 小思、圖〇 圖（A )係為一嵌有浮水印之高動態範圍影像。 一第七圖（B)係為第七圖⑴經一色調對應步驟 臧有浮水印的低動態範圍影像。 之 第八圖係為由第七圖（B)所取出之一浮水印。 第九圖為經不同失真破壞後之嵌有浮水印一 ^ 範圍料所水印比較結果圖。 n動態 第十圖為經不同失真破壞後之嵌有浮水印 態範圚旦彡你 < 另—向動 圍〜像所取出之該浮水印比較結果圖。 【主要元件符號說明】(HDR)高動態範圍影像 13 200849135 (HDRW )嵌有浮水印的高動態範圍影像 (LDR)低動態範圍影像 (LDRW )藏有浮水印的低動態範圍影像 (RI)比例影像 (W)浮水印 14The figure / (B) is the watermark of the sixth figure (a). The small picture and the picture (A) are high dynamic range images embedded with watermarks. A seventh figure (B) is a low dynamic range image with a watermark in the seventh picture (1) via a tone matching step. The eighth figure is one of the watermarks taken out by the seventh figure (B). The ninth picture shows the watermark comparison result of the watermark embedded in the watermark after being destroyed by different distortions. n Dynamics The tenth figure shows the comparison of the watermarks with the watermark embedded in the watermarked state after the different distortions are destroyed. [Main component symbol description] (HDR) High dynamic range image 13 200849135 (HDRW) High dynamic range image (LDR) with low watermark image (LDRW) embedded with watermarking Low dynamic range image (RI) ratio with watermark Image (W) watermark 14