200524406 玖、發明說明 【發明所屬之技術領域】 本發明係有關於一種影像輸出裝置，特別是與一種具影 像補償功能之影像輸出裝置有關。 【先前技術】 電荷麵合裝置（Charge Coupled Device，CCDs)和互補式 金氧半場校電晶體（CMOS)，常使用在掃瞄器的掃瞄頭1 是數位式相機上作為受光體。其中CCDs在表面感受到光線 時，會將電荷反應在元件上，當整個CCDs上的所有感光元 件所產生的訊號，就構成了一個完整的畫面。而cm〇s主 要是利用朴鍺這兩種元素所做成的半導體，使其在〇刪 上共存著帶負電和帶正電級的半導體，這兩個互補效應所產 生的電流即可被處理晶片紀錄並解讀成影像。因此當這些受 光體本身具有缺陷，亦極具有損壞晝素（— 會影響到影像品質。 專充上的解决方法是先將損壞畫素的位址記錄下來，在 fs或CMOS上所記錄的影像要被讀出時，再將損壞畫素 =錄值由與其相鄰晝素的平均記錄值所取代，亦即使用 平均之方法進行損壞畫素之補償。 由* 所示為傳統上影像輸出處理系統，當-資料 ==時’此資料會先經過FPN校正1。2,先將雜 再經過PRNU校正1〇4,消除受光體間之差 Ή象輪出。其中畫素補償記憶體1GG係用來同步 200524406 紀錄於晝素檢測後損壞晝素之位址。當影像資料傳入時，損 壞晝素補償區塊110會將所輸入之影像資料，不論是壞的畫 素或是好的畫素資料均進行補償，補償後之影像資料再傳送 到輸出選擇區塊120，而另一方面未經補償之影像資料亦會 經由一延遲模組130傳送到輸出選擇區塊120。此時輸出選 擇區塊120即會根據晝素補償記憶體1〇〇中所記錄之損壞畫 素之f址，選擇輸出由延遲模、组13()傳送之未經補償影像資 料或是由損壞晝素補償區塊丨1〇補償後之影像資料。在上述 影像輸出處理中，需要一額外之晝素補償記憶體ι〇〇來記錄 才貝壞畫素之位置，如第1 Hi 0^ ^ 里划弟1圖所不傳統方式之缺點是需要複雜 程序及額外記憶体。 而另-種影像輸出方式収將所輸人之影像資料，不論 是壞的晝素或S由好的畫素之f料均進行補償後輸出，雖然 此方式並不需要-額外之畫素補償記憶體，但由於所有之影 像資料均經過補償，雖然解決指 ’ m 貝壞晝素問題，也造成所輸出 之衫像貝料與原有之影像資料間存在一定之差異。 因此’如何在不增加額外雜 者去，廿斜甘> μ㈢力顯外忑隐體之情況下，記錄損壞之 旦素並對其§己錄之影傻杳祖隹 y像貝枓進仃補償，即成為改進之重點。 【發明内容】 ::明的主要目的即是在提供一種影像輸出裝置，i有 相壞旦素之仙功能，可提升輸出影像之品質。 本發明另一目的為接·/此_ n ^ - Γ ” —種衫像輸出裝置，並不需要使 用額外之5己憶體來§己錄損壞晝素，因此可降低成本。 200524406 本發明再-目的為提供—種影像輸出裝置 外記憶體來記錄損壞畫素之情況下，可針對損奎 額 償來提升輸出影像品質。 、、1素進行補 根據上述之目的，本發明提供—種影像輸出裝置 包括一卿校正表，儲存有在建立卿校正 二 壞=之資料，:PRNU校正表，健存有在建立二= 正表呀判斷為損壞畫素。當一資料由 ^ 會先藉由FPN h τ主』η 70 貝取4 ’此資料 先糟由FPN&正表和PRNU校正表進行校 示），接著如第3圖所示一損壞畫素補償區塊會將 象資料進行補償，一輸出選擇區塊會根據prnu父校 表和FPN校正表中所記錄之損壞畫 經補償之影像資料或是補償後之資料。 輪出未 【實施方式】 在不限制本發明之精神及應用II圍之τ，以下即以一 二:本Γ月之實施;熟悉此領域技， 影像_置-是要== 全面性後續補償’要不就是對所有晝素進行 像品質，因此本i:i:r額外記憶裝置，一影響輸出影 像輸出梦署^月即疋要改進上述缺點。根據本發明之影 ^糸利用傳統在進行影像輸出時，通常會& __ =⑽校正表(corre—^ 月即疋在對受光元件進行校正之同時，將偵測 200524406 到異常之受光元件一併記錄在FPN和PRNu校正記錄表中， 如此，在做影像補償時，即可根據校正記錄表對異常查素之 記錄來進行補償，而不需額外使用一記憶體。 每一個受光體在相同的光線照射下，應該有相同之輪 出，然而由於受光體本身於生產、製造、或與產品之組裝2 產、使用過程及材料之差異都可能造成bad pixel產生，若未 適當處理將影響影像品質會造成對光反應彼此不同，個別受 光體對光之反應與整體平均之差異稱為像素反應不一2 (Photo Response Non-Uniformity，PRNU)，而另一方面， 受光體亦會受到來自外界之影響而造成讀取影像時產生雜 訊’此稱為固定型式雜訊（Fixed pixel Noise，FPN)，一般而 言，上述兩值應該越低越好。 一般為了解決光源、鏡頭及受光體本身特性等因素所 造成之固定型式雜訊或像素反應不一致之情形，會進行如第 圖所示之平面校正法（Fiat fieid Correction )，利用減法和 乘法來對每一個畫素進行FPN和PrnU校正。在進行校正 前，會先建立FPN和PRNU校正表，其中FPN*PRNU校正表 是儲存於一記憶體中，此記憶體可為一隨機存取記憶體，例 如可藉由掃描一張黑紙或遮蔽鏡頭使受光體處於不受光狀 態，建立每一個受光體受到雜訊之影響之FPN表，在正常情 形下，黑紙之掃描值應為〇，但受雜訊影響造成每個受光體 值介於0〜127。另外藉由掃描一張白紙，建立每一個受光 體與正常輸出值之PRNU表，在正f情形下白紙之掃描^應 為255,但由於受光體本身材料之差異，造成每個受光體之 200524406 PRNU值介於128〜255。因此當一資料由受光體讀取時’此 資料會先經過FPN校正202，先將雜訊影響過渡掉’再經過 PRNU校正204，消除受光體間之差異。 本發明是在做受光元件校正之同時，增加受光元件之 判斷，並將偵測到異常之受光元件一併記錄在FPN和PRNU 校正記錄表，便不需額外之記憶體。例如，在正常情況下， FPN校正表是藉由掃描一張黑紙所建立，因此表中每一個值 應介於0〜127間，若有一值超過127，即可斷言此為損壞晝 素，因此本發明之方法，會在原來之FPN校正記錄表記錄下 此損壞畫素。同樣的，PRNU校正記錄表是藉由掃描一張白 紙所建立，因此表中之每一值應介於128〜255間’若有一 值低於128，即可斷言此為損壞畫素，因此本發明之方法’ 即會在原來之PRNU校正記錄表中記錄下此損壞畫素’供後 續作畫素補償。藉由將損壞畫素之位址記錄在PRNU* FPN 校正表中，即可不使用額外之記憶裝置。值得注意的是’上 述判斷損壞畫素之關鍵值，如FPN值之127和PRNU值之128 可由使用者根據所要求之影像品質更改，而記錄損壞畫素之 方法，例如可將原來掃描到之FPN值和PRNU值置換為〇儲 存，來代表此畫素為損壞，然，亦可使用其他之值代表。 參閱第3圖所示為本發明根據FPN和PRNU校正表之内 容決定其輸出之影像輸出裝置。當一資料由受光體讀取時， 此資料會先作一平面校正，藉由一 FPN校正表302，將雜訊 過濾掉，再經過PRNU校正表304，消除受光體間之差異， 其中FPN校正表中儲存有在建立FPN校正表時判斷為損壞 200524406 二，RNU校正表中同樣儲存有在建立™校 入壞畫素之資料。當做完校正後之影像資料輸 不:ϊ是區塊310會將所有校正後之影像資料， 是好的晝素資料均進行補償，補償後之 3J= 延遲模組330傳送到輪出選擇區塊 通，、讀Λ 33G之作㈣將未經補償之影像資料進行延 遲’以讓未經補償之寻彡彳复眚 叶出…1 償後之影像資料同時傳輸 此時輸出選擇區塊320會根針_校 正表和FPN杈正表中所記錄之損壞晝素資料 遲模組330傳送之未經補償影像資珣 塊3H)補償後之影像資料像貝#或疋由損壞畫素補償區 藉由本發明之方法，並不需要額外記憶體來記錄損壞畫 ^ 且根據PRNU校正表和卿校正表中所記錄: 拍壞：素貝料來決定需不需要做補償，因此不會如傳技蓺 Ϊ成二減二記憶體之使用而將所有資料均進行補償，反： 像失真。再加以’本發明是使用原來叹而校 和FPN校正表進行記錄，並不會改變原本之架構。 雖然本發明已以一較佳實施例揭露如上，然其並 限定本發明，任何熟習此技藝者，在不脫離本發明之 範圍内’當可作各種之更動與潤飾，因此本發明之保‘範圍σ 當視後附之申請專利範圍所界定者為準。 【圖式簡單說明】 10 200524406 為讓本發明之上述和其他目的、特徵、和優點能更明顯 易懂，下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圖為傳統影像處理裝置； 第2圖為根據本發明之影像資料校正裝置；以及 第3圖為依據本發明較佳實施例的影像處理裝置。 【元件代表符號簡單說明】 100畫素補償記憶體 110和310損壞晝素補償區塊 120和320輸出選擇區塊 130和330延遲模組 102、202 和 302 FPN 校正表 104、204 和 3 04 PRNU 校正表 • 11200524406 (1) Description of the invention [Technical field to which the invention belongs] The present invention relates to an image output device, and particularly relates to an image output device with an image compensation function. [Prior technology] Charge Coupled Devices (CCDs) and complementary metal-oxide-semiconductor half-field correction crystals (CMOS) are often used as the light receiver on the scanning head 1 of the scanner, which is a digital camera. Among them, when the CCDs feel the light on the surface, they will reflect the charge on the element, and the signals generated by all the photosensitive elements on the entire CCDs constitute a complete picture. And cm〇s is mainly a semiconductor made of two elements, such as pak germanium, so that negative and positively charged semiconductors coexist on the chip, and the current generated by these two complementary effects can be processed. The chip is recorded and interpreted as an image. Therefore, when these photoreceptors have defects themselves, they also have great damage to the image quality (which will affect the image quality. The solution for the special charge is to first record the address of the damaged pixel and record the image on fs or CMOS. When it is to be read out, the damaged pixel = the recorded value is replaced by the average recorded value of the adjacent pixel, that is, the average pixel is used to compensate for the damaged pixel. The traditional image output processing is shown by * System, when -data == 'This data will be corrected by FPN 1.2 first, and then the noise will be corrected by PRNU 104, which eliminates the difference between photoreceptors in rotation. The pixel compensation memory 1GG is It is used to synchronize the 200524406 record to the location of the damaged daylight after the detection of the daylight. When the image data is imported, the damaged daylight compensation block 110 will input the image data, whether it is a bad pixel or a good picture. All the raw data is compensated, and the compensated image data is transmitted to the output selection block 120, while the uncompensated image data is also transmitted to the output selection block 120 through a delay module 130. At this time, the output selection Block 120 is coming According to the f address of the damaged pixel recorded in the day pixel compensation memory 100, it is selected to output the uncompensated image data transmitted by the delay mode, group 13 () or to be compensated by the damaged day pixel compensation block. Later image data. In the above image output processing, an extra daylight compensation memory ι〇00 is needed to record the position of the bad pixels, as shown in the first traditional method of the first picture. The disadvantage is that it requires complicated procedures and additional memory. Another type of image output method accepts the input image data, whether it is bad daylight or S, and compensates for the f material of the good pixel. This method does not require-additional pixel compensation memory, but because all the image data is compensated, although the problem of 'm's bad daylight is solved, it also causes the output of the shirt material and the original image data. There is a certain difference between them. Therefore, 'How to record the damaged element without adding extra miscellaneous people, and obsessively hide the exogenous invisible body, and record the shadow of its own 杳y Compensation, like Betty, becomes an improvement [Summary of the invention] :: The main purpose of Ming is to provide an image output device, i has the function of phalanx, which can improve the quality of the output image. Another object of the present invention is to connect with this / this_ n ^-Γ ”— a shirt-like image output device, does not need to use an additional 5 hexadecimal memory to § record damage to the day element, so the cost can be reduced. 200524406 The present invention further aims to provide an external memory of an image output device In the case of recording damaged pixels, the quality of the output image can be improved according to the amount of damage. According to the above purpose, the present invention provides an image output device including a correction table, which is stored in the establishment. The data of the correction of the second bad =, the PRNU correction table, there is a establishment of the second = positive table, which is judged to be a damaged pixel. When a piece of data is taken by ^, it will be taken by FPN h τ lord η 70. 4 'This piece of data is corrected by FPN & positive table and PRNU correction table), and then a damaged pixel compensation area is shown in FIG. 3 The block will compensate the image data, and an output selection block will be based on the damaged image recorded in the prnu parent school table and the FPN correction table, or the compensated image data. [Embodiment] Without limiting the spirit of the present invention and its application, the following τ will be taken as follows: the implementation of this month; familiar with the technology in this field, image_set-is to be == comprehensive follow-up compensation 'Otherwise, it is the image quality of all daylight elements, so this i: i: r extra memory device, once it affects the output image output, it is necessary to improve the above disadvantages. According to the present invention, when using the traditional method for image output, it is usually & __ = ⑽ correction table (corre- ^ month). While calibrating the light receiving element, it will detect 200524406 to an abnormal light receiving element. It is recorded in the FPN and PRNu calibration record tables. In this way, when doing image compensation, you can compensate for the abnormal query record according to the calibration record table without using an additional memory. Each photoreceptor is the same Under the light of the light, there should be the same rotation. However, due to the difference in the production, manufacturing, or product assembly of the photoreceptor itself, the difference in production, use process and materials may cause bad pixels. If not processed properly, it will affect the image Quality will cause different responses to light. The difference between the individual light receiver ’s response to light and the overall average is called pixel response non-uniformity (PRNU). On the other hand, the light receiver will also be subject to external light. Noise is generated when reading the image due to the influence of this. This is called fixed pixel noise (FPN). Generally speaking, the above two values should be Generally, in order to solve the situation where the fixed type noise or pixel response is inconsistent caused by factors such as the characteristics of the light source, lens, and the receiver itself, the Fiat fieid Correction method as shown in the figure is used. Multiply to perform FPN and PrnU correction for each pixel. Before performing the correction, FPN and PRNU correction tables will be created. The FPN * PRNU correction table is stored in a memory. This memory can be a random access. Memory, for example, by scanning a piece of black paper or shielding the lens to make the photoreceptor in a light-free state, an FPN table for each photoreceptor affected by noise is created. Under normal circumstances, the value of the black paper scan should be 0. , But affected by noise, the value of each photoreceptor is between 0 ~ 127. In addition, by scanning a piece of white paper, a PRNU table of each photoreceptor and the normal output value is established. In the case of positive f, the scan of white paper should be 255, but due to the difference in the material of the receiver, the 200524406 PRNU value of each receiver is between 128 and 255. Therefore, when a data is read by the receiver, this data will be corrected by FPN first. 202, the effect of noise is first transitioned, and then the PRNU correction 204 is performed to eliminate the difference between the photoreceptors. The invention is to increase the judgment of the photoreceiving element while correcting the photoreceiving element, and to detect an abnormal photoreceiving element. It is recorded in the FPN and PRNU calibration record tables, so no additional memory is required. For example, under normal circumstances, the FPN calibration table is created by scanning a piece of black paper, so each value in the table should be between 0 ~ In 127, if there is a value exceeding 127, it can be asserted that this is a damaged day element, so the method of the present invention will record this damaged pixel in the original FPN correction record table. Similarly, the PRNU correction record table is created by scanning a piece of white paper, so each value in the table should be between 128 and 255. 'If there is a value below 128, it can be asserted that this is a damaged pixel. The method of invention 'will record this damaged pixel in the original PRNU correction record table for subsequent pixel compensation. By recording the address of the damaged pixel in the PRNU * FPN correction table, no additional memory device is required. It is worth noting that the above key values for judging damaged pixels, such as 127 of FPN value and 128 of PRNU value, can be changed by the user according to the required image quality, and the method of recording damaged pixels, for example, can scan the original The FPN value and PRNU value are replaced with 0 and stored to represent that this pixel is damaged, however, other values can also be used to represent it. Referring to FIG. 3, there is shown an image output device that determines its output according to the contents of the FPN and PRNU correction tables. When a piece of data is read by the light receiver, the data is first corrected for a plane. The noise is filtered by an FPN correction table 302, and then passed through the PRNU correction table 304 to eliminate the difference between the light receivers. Among them, the FPN correction The table stores the 2005254406 which was judged to be damaged when the FPN correction table was created. The RNU correction table also stores the data of bad pixels that were entered in the creation ™. When the image data after correction is not input: ϊ Yes block 310 will compensate all the corrected image data, which is good daylight data, and the compensated 3J = delay module 330 will be transmitted to the rotation selection block Pass, read Λ 33G's work, delay the uncompensated image data, so that the uncompensated search will be restored ... 1 the compensated image data will be transmitted at the same time. At this time, the output selection block 320 will have a needle. _Correction table and FPN frame recorded in the damaged daytime data late module 330 uncompensated image data block 3H) after compensation image data # or 疋 from the damaged pixel compensation area by the present invention This method does not require additional memory to record the damaged picture ^ And according to the records in the PRNU correction table and the correction table: Shooting bad: The raw material determines whether compensation is needed or not, so it will not be as successful as transmission technology The use of two minus two memories compensates all the data, and the reverse: image distortion. Furthermore, the present invention uses the original calibration and FPN correction table for recording, and does not change the original structure. Although the present invention has been disclosed as above with a preferred embodiment, but it does not limit the present invention. Anyone skilled in the art will be able to make various modifications and retouches without departing from the scope of the present invention. The scope σ shall be determined by the scope of the attached patent application. [Brief Description of the Drawings] 10 200524406 In order to make the above and other objects, features, and advantages of the present invention more comprehensible, a preferred embodiment is given below in conjunction with the attached drawings to make a detailed description as follows: 1 is a conventional image processing apparatus; FIG. 2 is an image data correction apparatus according to the present invention; and FIG. 3 is an image processing apparatus according to a preferred embodiment of the present invention. [Simple description of component representative symbols] 100 pixel compensation memory 110 and 310 are damaged. Daylight compensation block 120 and 320 output selection block 130 and 330 delay module 102, 202 and 302 FPN correction table 104, 204 and 3 04 PRNU Calibration table • 11