1377510 九、發明說明： 【發明所屬之技術領域】 本發明是一種翻譯裝置與方法，特別是一種影像翻譯裝置與方法。 、 【先前技術】 . 翻譯麵於語言學冑上具有相當的幫助。而-贿纖賴的使用方 式乃是使用者透過鍵盤而輸入欲查詢的單字、片語或句子，然後等待翻譯 0 機運算處理後，即可於翻譯機的螢幕上看到查詢之結果。 傳統的翻_在使肖上’須由使用者透賴盤逐字輸續欲查詢的文 字，最後才能制翻譯騎h此種方式在使用上具有不便之處，例如·· 透過鍵盤逐字輸入，將使得整體的查詢速度較為緩慢；使用者需熟知所欲 查詢文字的輸入方式’若是使用者欲查詢本身不熟悉的文字，便無法正確 的輸入。尤其當所欲查詢的文字數量較多時，更加突顯傳統上逐字輸入的 翻譯機之不便之處。 • 中華·專利申請案號8611娜「攝像式電子翻賴」與中華 民國專利申請案號㈣峨「影像辨識翻譯裝置及方法」，提出利用鏡頭攝 取影像後’翻像㈣文字並齡翻譯出之文字。如此，不僅可避免使 用者因看不懂文字而無法查詢的固擾，更可達到隨拍即用而快速翻譯的功 • 能。 . 上述之f知專利’利用辨識影像中的文字，進而翻譯影像中文字之作 法，改善了傳統上逐字輸入的翻譯方法。然而，卻衍生了另一個間題當 文字所在的看板或是印刷品的背景其底色花樣過於豐富或花偶時，對於文 5 印7510 字辨識系統而言是—個難解之問題。也就是說，當所取得的影像之背景影 像過於複雜#料財彳贱辟，*且準輕 差的問題產生。所以’仍然無法有效的翻譯出影像中之文字。 因此，如何解決習知技術中，影像辨識翻譯所衍生的問題為一虽待 改善的議題。 【發明内容】 有鏗於此本發明提出一種影像翻譯裝置與方法。藉由本發明所提出的 裝置或方法，可大幅提升影像巾文字辨識的能力。，即使賴取的影 像具有過於複雜的背景，或者是⑽豐富的彩色影像，透過本發明所提出 的裝置或方法，皆能有效辨識_中的文字，進而翻譯成制者所需的 語言。 本發明提出-種翻譯裝置，包含：—影像做，用以摘取可 見光影像與紅外線影像卜影像處理模組，比對可見絲像與紅外線影像， 而產生文轉像；—光學文字觸(〇CR)做，顺文字職為屬於第一語 言之文字；一翻譯模組，翻譯文字為第二語言；一輸出模組，輸出翻譯為 第二語言之文字。 本發明亦提tij-錄像翻譯方法，包含下列步驟1取可見光影像與 紅外線影像；比對可見光影像與紅外線影像，而產生文字影像；辨識文字 影像為屬於第之文字；崎文字為第二語言；輸出崎為第二語言 之文字。 有關本發明的較佳實施例及其功效，兹配合圖式說明如后。 6 1377510 【實施方式】 請參照「第^圖」，該圖所示為影像翻譯裝置之一實施例方塊圖。影像 翻譯裝置包含：驗1G、影像纽歡2G、絲文字简啊 模組30、翻譯模組4〇、輸出模組5〇。 影像齡模組10用以操取可見光影像與紅外線影像。影像祿取模組ι〇 可為-般_機_雜等，由於械、攝職—般1377510 IX. Description of the Invention: [Technical Field] The present invention is a translation apparatus and method, and more particularly, an image translation apparatus and method. [Prior Art] The translation is quite helpful in linguistics. The method of using the bribe is to input the words, phrases or sentences to be queried through the keyboard, and then wait for the translation to be processed, and then the results of the query can be seen on the screen of the translator. The traditional 翻 _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ The query speed will be slower; the user should be familiar with the input method of the text to be queried. If the user wants to query the text that is not familiar to the user, the user cannot input it correctly. Especially when the number of words to be queried is large, the inconvenience of the traditional verbatim input machine is more prominent. • China Patent Application No. 8611 Na "Camera Electronic Rewind" and Republic of China Patent Application No. (4) "Image Recognition Translation Device and Method", proposed to use the lens to capture images after the 'mirror image (four) text and translated Text. In this way, not only can the user avoid the stagnation that cannot be queried because they can't understand the text, but also the function of quickly translating with the beat. The above-mentioned patents use the method of recognizing text in an image and then translating the characters in the image, thereby improving the traditional translation method of verbatim input. However, it has created another problem. When the background of the kanban or printed matter of the text is too rich or a couple of background colors, it is a difficult problem for the 710-character recognition system. That is to say, when the background image of the acquired image is too complicated, the problem of quasi-lightness is caused. Therefore, it is still not possible to effectively translate the text in the image. Therefore, how to solve the problem of image recognition translation in the prior art is an issue to be improved. SUMMARY OF THE INVENTION The present invention provides an image translation apparatus and method. With the device or method proposed by the present invention, the ability of the image recognition of the image towel can be greatly improved. Even if the captured image has an overly complex background or (10) a rich color image, the device or method proposed by the present invention can effectively recognize the text in _ and translate it into the language required by the maker. The invention provides a translation device, comprising: an image for extracting a visible light image and an infrared image image processing module, and comparing the visible silk image with the infrared image to generate a text image; CR), Shun text is the text of the first language; a translation module, the translated text is the second language; an output module, the output is translated into the language of the second language. The invention also provides a tij-video translation method, which comprises the following steps: taking a visible light image and an infrared image; comparing the visible light image with the infrared image to generate a text image; recognizing the text image as the first text; and the text is the second language; The output is in the language of the second language. Preferred embodiments of the present invention and their effects are described below in conjunction with the drawings. 6 1377510 [Embodiment] Please refer to "FIG. 2", which shows a block diagram of an embodiment of an image translation apparatus. The image translation device includes: 1G, image 2G, silk text, module 30, translation module 4〇, and output module 5〇. The image age module 10 is used to acquire visible light images and infrared images. Image Locator module 〇 〇 can be - like _ machine _ miscellaneous, due to machinery, job-like
—用™。因此，本發明提出於利用影像：組： 取同-影料’同時可見光影像妹外線影像，作法可為齡可見光 影像時’如-般練祕的模式，也就是紅外魏抑；相對的，在 操取紅外線影料，以碰肢外_如，並概驗Μ讓紅外線通 過的遽光片(IRpass filter)便可榻取到紅外線影像。或者，也可採用如紅外線 夜視功能的方式，擷取紅外線影像。 影像處理模組20比對可見光影像與紅外線影像，進而產生文字影像。 透過办像處理模纟 1 2G而產生文字景彡像後，藉由光學文字辨識(qCR)模組 3〇辨識文子影像為屬於第一語言之文字。之後，便可由翻譯模組翻譯 屬於第一语s之文字為第二語言，例如：將原本文字影像中屬於曰本語言 的曰文翻譯為中文等。最後，由輸出模組5〇輸出翻譯為第二語言之文字， 以提供使用者參考。 本發明為了克服先前技術所提及，當文字所在看板或印刷品的背景其 底色花樣過於豐富，也就是說，所取得的影像之背景過於複雜時，無法有 效辨識該文字之問題’因此本發明提出同時擷取可見光影像與紅外線影像 7 !37751〇 之方式而加以改善。於此，文字影像所對應之實體物可包含一塗科，且該 塗料為可見光不可穿透而紅外線可穿透。其中，實體物可為招牌菜單、 站牌或運輸工具時刻表等。 舉例說明’ -般遊客到國外旅遊，特別是自助旅行時，當在各個觀光 景點遇到看不懂的文字’便可藉由本發明所提出的影像翻譯裝置先娜 影像’而所棟取的影像為可見光影像與紅外線影像。由於文字影像所對應 之實體物(例如：招牌），具有可見光不可穿透而紅外線可穿透的塗料，也就 是說，使用此塗料所印刷的文字是人眼清楚可見，但紅外線卻可穿透。因 此’如果採用-般攝影模式即可將文字影像正常拍攝下來，若採用紅外線 影像攝紛娜卩冑乎看不見敎轉㈣存在。目此，職知模組2〇比 對影像操取模組1G所操取之可見光影像與紅外線影像兩者間的差異，便可 得到文字影像。 於此，為了餅文字影像的觸效果，影像處_組2()可先將可見光 影像與紅外線影轉絲灰祕式後，再進行__作1像處理模組 20應用數位影像處理技術，對兩張影像進行比對及過遽的演算，遽除比對 後之可見光影像與紅外線影像財景，即可產生—張對應於所欲查詢文字 的文字影像i此，較可見絲像為背景複雜的彩色影像，也可清楚獲 得文字影像，如此後續的衫_上，獻幅提升贱及準雜。 述說月可知’本發明實際使用之—實施例，可在觀光景點、交通 設施'餐飲店等設置時，便將招牌、站牌、運輸工具時刻表或菜單等，用 可見光不可穿透巾紅祕可穿透的塗料加。峨光客便可透過本發 1377510 明所提出的裝置或方法’迅速棘影像而翻譯出所熟悉的文字，如此將提 升外國觀光客鱗時的便雛，糾提升觀光旅遊的品質。 上述之#軸譯裝置可選自手機、個人數㈣理pm)、智慧電話㈣时 phone)或攝影機等’卩要具有攝影功能的電子裝置即屬於本發b月之範脅。 請同時參照「第2A圖」與「第2B圖」分別為影像翻譯之示意圖㈠ 與㈡，說明輸出模組5〇的其中一種輸出方式。輸出模組％可顯示第二語 言之文字於可見光影像上。.由「第2A圖」與「第2b圖」所示，左半邊為 可見U象圖中的第-語言之文字為日本語言的日文字經由影像翻譯 後，輸出模組50在可見光影像上顯示翻譯為第二語言之文字也就是中文， 如圖中右半邊所7F如此’可讓使用者清楚得知翻譯後的結果。當然，輸 出模組50也可單純顯示第二語言之文字，也就是只顯示翻譯後的中文予使 用者參考，或韻由料觀第二語言之文字，崎制者得知鱗的結 果。 影像 請參照「第3 ®」，刻所4爾崎之方法流_，包含下列步驟。 步驟SH):餘可見絲像與紅物彡像。其中，可見絲像可為彩色 步卿0 ·比對可見光影像與紅外線影像，而產生文字影像。其十，文 字影像所對狀實_包含-_，且錄料可見料可穿透而紅外線 可穿透。爾可選自招牌、菜單、站牌、運輸工具時刻表及其組合 所構成的群組。 在步驟S20的比對步驟之前， 可包含下列轉：灰可見光影像與 9 1377510 卜線景y像之後再進行比對的動作。此外，比對步驟也可包含下列步驟： _比對後之可見光影像與紅外線影像的背景。 步驟S30 :辨識文字影像為屬於第-語言之文字。 步驟S40 :翻譯文字為第二語言。 步驟S50 :輸出翻譯為第二語言之文字。其中，輸出之步驟可包含下列 步驟：顯示第二語言之文字於可見光影像上。或者，單純顯示第二語言之 文子。又或者，語音撥放第二語言之文字。 雖然本㈣的技_容已經赠佳實關揭露如上，財並非用以限 定本發明’任何熟習此㈣者，在不聽本發明之騎轉些許之更動與 潤飾，皆賴胁本發_範細，因此本㈣之_麵當視後附之申 請專利範圍所界定者為準。 【圖式簡單說明】 第1圖：影像翻譯裝置之一實施例方塊圖。 第2A圖：影像翻譯之示意圖(一）。 第2B圖：影像翻譯之示意圖(二）。 第3圖：影像翻譯之方法流程圖。 【主要元件符號說明】 1〇 :影像擷取模組 20 :影像處理模組 3〇 :光學文字辨識模組 40 :翻譯模組 1377510 50 :輸出模組- Use TM. Therefore, the present invention is directed to the use of images: groups: taking the same - shadow material 'simultaneous visible light image sister outside line image, the method can be the same as the age of visible light image mode, that is, infrared Wei suppression; relative, in Take the infrared shadow material to touch the outside of the body. For example, and check the IRpass filter that allows the infrared rays to pass through, you can take the infrared image. Alternatively, infrared imagery can be captured by means of an infrared night vision function. The image processing module 20 compares the visible light image with the infrared image to generate a text image. After the image scene image is generated by the image processing module 1 2G, the text image is recognized by the optical character recognition (qCR) module 3 as the text belonging to the first language. After that, the translation module can translate the text belonging to the first language s into a second language, for example, translating the essays belonging to the slang language in the original text image into Chinese. Finally, the output module 5 outputs the text translated into the second language to provide a user reference. In order to overcome the problems mentioned in the prior art, when the background of the kanban or printed matter of the text is too rich in the background color, that is, the background of the obtained image is too complicated, the problem of the character cannot be effectively recognized. It is proposed to improve the method of simultaneously capturing the visible light image and the infrared image 7 !37751〇. Herein, the physical object corresponding to the text image may include a coating, and the coating is invisible to visible light and infrared permeable. Among them, the physical object can be a sign menu, a stop sign or a transportation timetable. For example, 'when a tourist travels abroad, especially when traveling by self-help, when encountering an incomprehensible text at each sightseeing spot, the image translation device proposed by the present invention can be used. The images are visible light images and infrared images. Because the physical object corresponding to the text image (for example, the signboard) has a paint that is invisible to visible light and penetrates into the infrared light, that is, the text printed by the paint is clearly visible to the human eye, but the infrared light can penetrate. . Therefore, if you use the normal photography mode, you can shoot the text image normally. If you use the infrared image, you can't see it (4). Therefore, the function module 2 can obtain a text image by comparing the difference between the visible light image and the infrared image captured by the image capturing module 1G. Here, for the touch effect of the pie text image, the image portion _ group 2 () can first convert the visible light image and the infrared ray shadow to the gray image, and then perform the digital image processing technology for the image processing module 20 . By comparing and comparing the two images, the visible image and the infrared image are deleted, and the text corresponding to the text to be queried is generated. Complex color images can also clearly obtain text images, so that the subsequent shirts, the increase in the size and the quasi-mixed. It can be said that the actual use of the present invention - the embodiment can be used in sightseeing spots, transportation facilities, food and beverage stores, etc., and the signboard, the stop sign, the transportation timetable or the menu, etc. Penetrating paint added. Twilight customers can translate the familiar texts quickly and easily through the device or method proposed by the hair of the 1377510. This will enhance the convenience of foreign tourists and improve the quality of sightseeing. The above-mentioned #axis translation device may be selected from a mobile phone, a personal number (four) pm), a smart phone (four) phone, or a camera, etc. The electronic device having a photographic function belongs to the Fan of the present month. Please refer to "2A" and "2B" as the schematic diagrams (1) and (2) of image translation, respectively, to explain one of the output modes of the output module 5〇. The output module % can display the text of the second language on the visible image. As shown in "2A" and "2b", the left half is the U-picture in the U-picture, the Japanese-language text is the Japanese language, and the output module 50 is displayed on the visible image. The text translated into the second language is also Chinese, as shown in the right half of the figure, 7F, so that the user can clearly understand the translated result. Of course, the output module 50 can also simply display the text in the second language, that is, only the translated Chinese is used to refer to the user, or the rhyme is based on the text of the second language, and the rhythm knows the result of the scale. Image Please refer to "3th ®", engraving the method of the 4th Arsaki method, including the following steps. Step SH): I can see the silk image and the red object. Among them, it can be seen that the silk image can be a color step 0. The visible image and the infrared image are compared to produce a text image. Tenth, the text image is ___, and the recorded material can be penetrated and the infrared light can penetrate. It can be selected from the group consisting of signs, menus, stop signs, transportation schedules and combinations. Before the comparison step of step S20, the following rotations may be included: the gray visible light image is compared with the 9 1377510 line image y image. In addition, the comparison step may also include the following steps: _ Aligning the background of the visible light image with the infrared image. Step S30: Recognizing the character image as a text belonging to the first language. Step S40: The translated text is the second language. Step S50: Output the text translated into the second language. The step of outputting may include the steps of: displaying the text of the second language on the visible light image. Or, simply display the text of the second language. Or, voice dial the text of the second language. Although the skill of this (4) has been given by Jiashiguan to expose the above, the money is not used to limit the invention. Anyone who is familiar with this (four), who does not listen to the invention, is a little more versatile and retouched. Fine, so the scope of this (4) is subject to the definition of the scope of the patent application. BRIEF DESCRIPTION OF THE DRAWINGS Fig. 1 is a block diagram showing an embodiment of an image translation apparatus. Figure 2A: Schematic diagram of image translation (1). Figure 2B: Schematic diagram of image translation (2). Figure 3: Flow chart of the method of image translation. [Main component symbol description] 1〇: Image capture module 20: Image processing module 3〇: Optical character recognition module 40: Translation module 1377510 50 : Output module