九、發明說明： 【發明所屬之技術領域】 一種影像擷取裝置及其方法，特別是指一 域之影像擷取裝置及其方法。 【先前技術】IX. Description of the invention: [Technical field to which the invention pertains] An image capturing device and method thereof, in particular, an image capturing device and method thereof. [Prior Art]
種自動偵測對焦區 〜隨著數位時代絲臨，再加上相硬财術喊勃發展， 使:各種數位裝置已經開始獲得人們的認同，並被廣泛地使用於 ^生活之中。例如數位相機、數位攝顯㈣像擷取裝置，都 制者更便機生柯'態。耻，在如此的料需求下， ，商發應用及魏日新月異的新產品，來滿足不斷變化的需 ’、’於疋乎，各式各樣*咖途_位裝置才能如 們的生活之中。 、 二：’其自動對焦的方式，都是内定對 二t 進行自動對焦，並沒有辦法自動判斷對隹 2作=如數位相鱗所提供的對財法制㈣像的全域中 所魅I ’’、點心成自崎焦，或是湘數位訊號處理器計算 =:=域對比值之平均值作為全域對焦，也有其他如 n隹—53鱗’其便是使請焦區域的對比值來 其$隹^作，雖驗專利案提出焦點可自動偵測的技術，但 其對焦_城需要手動設定，並未完全的自動。 -随然在拍照取景時’有時候會將欲對焦的人物放到畫面之 可能是希錢置的是背景。所以，使用者拍照時，實際上 旦面一旁的人物可以是清晰的，而相片其他的部份 丄JJOJOy .晰的，’監視系統在監控時，在不同·用上， 也而要以移動物體作為 ]愿用上 專利案提出可自動3號 .像產生差值，_產W驗置的技★，其_相鄰的影 並根據移動向量改變對隹n出移動的物體的移動向量， 焦區域以自動祕咖=，’ 編_對 動向量在彡像（ ^專並对考慮到移 【發明内容Γ 遇到模糊的影像時，無法正確的對焦。 本發==:象中不，確的_焦的問題， 及其方法。 〃 |自動_對焦區域之影像韻取裝置 體的像進行分析來__物 如此便可卿物對焦區域， 先前技術所提到之無法對模^ 發明同時也可以對清晰影像進行自_，外，本 射式，_也不喪峨的精用確=象 發明所揭露之裝置===—面達成’本 組、、對焦模組。其中感興趣區域判定模組更包含：模糊t德 1336589 . 本發0綺揭露之方法，包括有下·列步驟：對複數張以上之連 .續的影像進行移動制，藉以產生區塊二元影像，並定義出至少 -移動區塊；當影像之騎度小於第—臨界值時，設定各移_ 塊所組成之最大區域為對焦區域；當影像之解析度大於或等於第 • 4界值時’對影像進行邊緣偵取產生邊緣二元影像，並將邊 緣二/〇影像映射至區塊二元影像以取得包圍移動區塊之移動邊 緣’同時設定移動邊緣骑焦_ ;⑽纽域之解析度 隹。 τ 有關本發明之詳細特徵與實作，_合圖示在實施方式中詳 細說明如下，其内容足以使任何熟習相關技藝者了解本發明之技 術内容並據以實施，且根據本說明書所揭露之内容及圖式，任何 熟習相關技藝者可輕易地轉本㈣_之目的及優點。 【實施方式】 以下先以「第1圖」本發明所提之可自動偵測對焦區域之影 籲像娜裝置之喊元件示意圖來綱本發明的運作。如圖所示， 本發明之裝置含有鷄躺歡11G、感興趣區域判定模組120 及對焦板組130。 其中移動偵測模組110與感興趣區域判定模組12〇電性連 接，負貝對複數張以上之連續的影像進行移動偵測（Ch如辟 Detection) ’藉以產生個別影像之區塊二元影像，並使用該區塊二 凡影像定義出至少-移祕塊；感興趣區域判定模組⑼電性連 接移動制肋m及難肋13G，貞㈣_影像之解析度 與預先設定之第-臨界值T1之_，並依_斷之結果設定對焦 i336589 而依細_度鮮—臨界值τι的嶋，本發明對 焦區域的設定可以分下列兩麵式進行設定： 當影像之解析度小於第一臨界值τι時，可以透過模糊影 為ft:設定單元121負責設定各移動區塊所形成之最大區域 ⑺當影像之解析度大於第—臨界值Tl ^可以透過邊緣偵 d早^3郷像進行邊緣制㈤帅伽㈣以產 =像，以及利用映射單元124將邊緣二元影像映射叫㈣ :::=::r 模㈣ ___ 晰二 最後，對焦模組130與感興趣區域判定模組12〇電性 負責依據對焦區域之解析度進行對焦。 接著以-個實施例來解說本發明的運作方法， 刀 2圖」/發明所提之可自動侧對焦區域之方法流程圖Γ、第 當影像擷取裝置卿所拍攝的物體進行移動時 置100所摘取到的影像會產生變化，意即相鄰的影像之^的、 2會改變’而糊貞測模組⑽便是減相鄰 辛 產生區塊二元影像（步驟聊在本實撕，移動; (Higher〇rderTeS〇 , x月所七之移動偵測之方法並不以4階的高階統計為限。 1336589 一假設Dn代表第n張晝面與第时張晝面的差值影像 以定=Dn裡每個8χ8區塊的平均差值為In。選用㈣區塊之 乃考量效能與運算量之間的平衡而定義的最適合數值，告. =大小可《任意更改’ w符合每相邊長值為二= 或見度的醜即可。接著使用4 _高階統計 _ 又Automatic detection of the focus area ~ With the digital era, coupled with the development of hard financial technology, so that: a variety of digital devices have begun to gain people's approval, and is widely used in life. For example, a digital camera, a digital camera (four), such as a capture device, are more convenient. Shame, under such demand, commercial applications and new products that are constantly changing to meet the ever-changing needs, and in all kinds of things can be like in our lives. . 2: 'The method of auto-focusing is to auto-focus on the two-t, and there is no way to automatically judge the 隹2== as the digital phase scale provides the financial system (four) image of the whole domain of the charm I '' The dim sum is from Saki Jiao, or the Xiang digital signal processor calculates =:= the average value of the domain contrast value as the global focus, there are other such as n隹-53 scale' which is to make the comparison value of the focus area to its $隹^作, although the patent case proposes a technology that can automatically detect the focus, but its focus_city needs to be manually set, not completely automatic. - When taking a picture, it is sometimes the case that the person who wants to focus will be placed on the screen. Therefore, when the user takes a photo, the characters on the side of the photo can be clear, and the rest of the photo is JJOJOy. Clearly, the monitoring system is different in use, but also in moving objects. As a result of the patent case, it is possible to automatically generate the number 3. The difference between the image and the image of the W, the _ adjacent shadow and the movement vector change the moving vector of the moving object to the 隹n, the focus The area is automatically secret coffee =, 'editing _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ The problem of _ focus, and its method. 〃 |Automatic _ focus area of the image of the image of the body of the device for analysis __ so that the object can focus on the area, the prior art mentioned can not be It is also possible to carry out the clear image from _, outside, the ejector, and the _ is not sinful. = The device disclosed in the invention ===-face to achieve 'this group, focus module. Where the region of interest The decision module further contains: fuzzy t de 1336589. The method includes the following steps: moving the continuous image to a plurality of consecutive images, thereby generating a binary image of the block, and defining at least a moving block; when the riding degree of the image is less than the first critical value Set the maximum area composed of each shift _ block to be the focus area; when the resolution of the image is greater than or equal to the 4th limit value, 'edge detection of the image produces an edge binary image and maps the edge 2/〇 image To the block binary image to obtain the moving edge surrounding the moving block' while setting the moving edge riding focus _; (10) the resolution of the new field τ Detailed description and implementation of the present invention, _ The details are as follows, and the content thereof is sufficient for any skilled person to understand the technical content of the present invention and implemented according to the content and the schema disclosed in the specification, and any skilled person can easily transfer the text (4) OBJECTS AND ADVANTAGES [Embodiment] The following is a schematic diagram of the invention of the present invention by the first embodiment of the present invention, which can automatically detect the focus area of the image-recalling device. As shown in the figure, the device of the present invention comprises a chicken lying 11G, a region of interest determination module 120 and a focus plate group 130. The motion detection module 110 is electrically connected to the region of interest determination module 12, Perform motion detection on a plurality of consecutive images (such as Detection) to generate a block binary image of an individual image, and use the block to define at least a secret block; the region of interest is determined The module (9) is electrically connected to the movable rib m and the difficult rib 13G, and the resolution of the image of the 贞(4)_ image and the preset first-threshold value T1_, and the focus i336589 is set according to the result of the _breaking. For the threshold value τι, the setting of the focus area of the present invention can be set in the following two-sided manner: When the resolution of the image is smaller than the first critical value τι, the blur can be used as the ft: setting unit 121 is responsible for setting each moving block. The largest area formed (7) when the resolution of the image is greater than the first critical value Tl ^ can be edge-detected by the edge detection (3), the edge system (5), the gamma (four) to produce the image, and the mapping unit 124 to use the edge unit binary image The map is called (4) :::=::r Module (4) ___ Clearly Finally, the focus module 130 and the region of interest determination module 12 are electrically responsible for focusing according to the resolution of the focus area. Next, an operation method of the present invention will be described in an embodiment, a method of the method of the invention capable of automatically focusing the focus area, and a method for moving the object photographed by the first image capturing device. The extracted image will change, meaning that the adjacent image will change 2, and the paste measurement module (10) will reduce the adjacent symplectic block to generate a binary image (step chat in this real tear) (Higher〇rderTeS〇, the method of motion detection in x7 is not limited to the high-order statistics of 4th order. 1336589 A hypothesis Dn represents the difference image between the nth and the first time The average difference of each 8χ8 block in the fixed=Dn is In. The (4) block is the most suitable value defined by the balance between the performance and the calculation amount, and the size can be changed arbitrarily. The length of each phase is 2 = or the ugliness of the visibility. Then use 4 _ high-order statistics _ again
==以測試各個Μ區塊中的物體是屬二= -或疋屬於靜止的物體。其中4階的高階統計函式為 H(11,12,13,14)=14+13-3^ (I2-i l ^-3^-1^+2114 由上述的高階統計函式可以得知’當連續的影像之間不存在 移動的物體（也就是只有靜止的物體）時，連續的影像之間的差 值經過上述高階統計的運算後，將會趨近於零；反之，當連續的 2之間存在移動的物體時，連續畫面間的差值經過高二二 運算後’將會遠大於零。 外接著，移_測模組110將判斷第二臨界值T2與高階統計的 運异結果的關係’當有8X8區塊的高階辑的運算結果低於第二 臨界值Τ2時，可以定義脉8區塊為非移動區塊，同時設定該㈣ 區塊對應在區塊二元影像的位置為G ;當有㈣區塊的高階統計 的運异結果高於臨界值時，則可以定義該8χ8區塊為移動區塊， 同時設定該8Χ8區塊對應在區塊二元影像的位置為i。如此一來， 在進行移動偵測後將產生-種只存在著兩種值（(mi)的區塊二 兀影像，也就是說，在區塊二元影像中的值為Q，則表示該8χ8 區塊為靜止的物體，而影像中值為丨，則代表該㈣區塊為移動 的物體’其中，值為丨的部份便包含了欲對焦的物體。本發明並 1336589 不以上述之ο與】作為靜止與軸的物__，意即也可 用1來表示靜止的物體所對應的，Sx8 , 尾便用0來表示移動的 物體所對應的8x8區塊。== to test whether the objects in each block are genus = - or 疋 belongs to a stationary object. The fourth-order higher-order statistical function is H(11,12,13,14)=14+13-3^ (I2-i l ^-3^-1^+2114 can be known from the above high-order statistical function 'When there are no moving objects between successive images (that is, only stationary objects), the difference between successive images will approach zero after the above high-order statistics; otherwise, when continuous When there is a moving object between 2, the difference between successive frames will be much larger than zero after the high-two operation. Then, the shift-test module 110 will judge the second critical value T2 and the higher-order statistics. The relationship of the result 'When the operation result of the high-order series of the 8X8 block is lower than the second critical value Τ2, the pulse 8 block can be defined as a non-moving block, and the (4) block corresponding to the block binary image is set. The position is G; when the result of the high-order statistics of the (4) block is higher than the critical value, the 8χ8 block can be defined as the moving block, and the position of the block corresponding to the binary image of the block is set to i. As a result, after the motion detection, there will be only two types of values ((mi) block two shadows That is, the value of Q in the block binary image indicates that the 8χ8 block is a stationary object, and the median value of the image is 丨, which means that the (4) block is a moving object' where the value is The part of the cymbal contains the object to be focused. The invention does not use the above as the object of rest and axis __, which means that 1 can also be used to indicate the stationary object, Sx8, tail. Use 0 to indicate the 8x8 block corresponding to the moving object.
事貫上’上述方式較適合使用在完成對钱影像中，在未士 成對焦的影像中，若使用上述產生的移動區域來進行對隹通^ :夠完整的，亦即，可能有部分的移動區域無法在未完成對焦^ 影像中被上述的難所_出來。這是因為影像相當模糊，所以 間接影_欺移龍域的準雜，此，本發明將進—步侧 感興趣區域’藉以由上述過程產生的飾區域產生相對應之對焦 因此，在感興趣區域判定模組12〇偵測感興趣區域時，首先 需要分析前一個影像的解析度，並判斷影像的解析度與第一臨界 值Τ1的關係（步驟230)，當解析度低於第一臨界值T1時，本發 月將判疋輸入的影像疋模糊影像，而當解析度高於第一臨界值丁 1 時，本發明將判定輸入的影像是清晰的影像，而後，本發明將依 據輸入的影像為模糊影像或清晰影像有不同的處理過程。 以下先說明輸入的影像為模糊影像的處理過程，在感興趣區 域判疋模組120判斷輸入的影像為模糊影像（也就是影像的解析 度小於第一臨界值T1)後，模糊影像對焦區域設定單元121會被 啟動，其會在區塊二元影像中選出值為丨所形成的各個區域中最 大的一個區域（也就是選出移動區塊所組成的最大區域），如「第 3A圖」之區域310所示’接著可以將區域31〇所的包圍的方框 3〇5作為對焦區域（步驟243)，如「第3B圖」所示，在本實施例 11 1336589 中，模糊影像對焦區域設定單元121將先找出區域的最左邊、 最右邊最上面、及最下面的四.個像素（分別為像素3ΐι〜像素 4)接著以通過最左邊與最右邊的像素（像素如、像素犯） 向垂直線（虛線3(Π、麟3G2)、及通過最上面與最下面的像素 (像素313、像素314)的水平線（虛線3〇3、虛線3〇4)圍成— 個方框305，方框3〇5即為該影像的對焦區域。In fact, the above method is more suitable for use in the completion of the image of the money, in the image in which the focus is not focused, if the above-mentioned generated moving area is used to perform the communication, it is complete, that is, there may be part of The moving area cannot be ejected in the unfinished focus image. This is because the image is rather ambiguous, so the indirect shadow _ bullying the quasi-heterogeneous of the dragon domain, this, the invention will advance the step-by-side region of interest' by the decorative region produced by the above process to produce a corresponding focus, therefore, interested When detecting the region of interest, the region determining module 12 first needs to analyze the resolution of the previous image, and determine the relationship between the resolution of the image and the first critical value (1 (step 230), when the resolution is lower than the first criticality When the value is T1, the input image will determine the input image 疋 blurred image, and when the resolution is higher than the first threshold value D1, the present invention will determine that the input image is a clear image, and then the present invention will be based on the input. The images are blurred or blurred and have different processing. The following describes the process of processing the input image as a blurred image. After the region of interest determination module 120 determines that the input image is a blurred image (that is, the resolution of the image is less than the first threshold T1), the blurred image focus area is set. The unit 121 will be activated, which will select the largest area among the regions formed by the 丨 in the binary image of the block (that is, the largest area selected by the moving block), such as "Picture 3A". In the area 310, the surrounding frame 3〇5 of the area 31〇 can be used as the in-focus area (step 243). As shown in FIG. 3B, in the present embodiment 11 1336589, the blurred image focus area setting is set. The unit 121 will first find the leftmost, rightmost, and bottommost four pixels of the region (pixels 3ΐι to 4, respectively) and then pass the leftmost and rightmost pixels (pixels, pixels) A vertical line (dashed line 3 (Π, 麟3G2), and a horizontal line (dashed line 3〇3, dashed line 3〇4) passing through the uppermost and lowermost pixels (pixel 313, pixel 314) encloses a block 305, Box 3〇5 Focus area of the image.
為了避免移動區塊所組成的區域過於零散或破碎因此本發 明在感興趣_狀歡12G中更可以包含有放大單元128，^ 責在模娜像對缝域設定單元121選出最大的區域⑽做為對 3域（步驟243)之前，先將影像中的各個移動區塊所組成的 (麵242)，使得—細雜是零舰碎的區域 〇a與區域现可峻紐域喊結合在 的區域320,如「第3C圖」所示。 〜另外’由她糊的影像在進行移_辦容$發生誤判，使 =際上有移_區频狀絲止，或是靜㈣區塊被判定為 因此本發明錢興趣區域判定模組⑽_射 移 動正規化單元m，編娜彡輪峰辦幻2= =大的區域训做為對焦區域（步驟243)之前，將移動制後 ^娜撕_錢⑽(辣241) ==焦區域更為精確。其中，移動正規化的二 化運算的區塊的二元值與該區塊周圍的8個相鄰 H7C值之間的關係，判斷順序與關係為：先 後的區塊二元影像t各個值為G的移動區塊進行判斷，若移動= 17 塊410的值為〇，其相鄰的周圍存在著4個以上值為】的移動區 塊時’即將該運算中的移動區塊·彻對應的二元值設定為i，如 第4A圖」所示；接著再對經過上述正規化後的區塊二元影像 ^各個值為1的移祕塊進行满，若轉區塊420的值為1， /、相糊ϋ存在著4個以上值為Q的移祕塊時，即將該判斷中 的移動區塊420的二元值設定為〇,如「第4B圖」所示。 此外’當影像中有移動的物體時，在連續的影像中的物體也 ，移動’因此職區域可能會改變，是故模糊影像職區域設定 單=121會不斷的修正練的對焦區域，在本實施例中，將對移 動貝訊缺乏、移動貧訊增加、移動區域的位置改變等三種狀況來 ^是否要更新對域驗置，但本發明更新對顏域的觸發 時機並不以上述三種狀況為限。 富第n+1 謂娜_祕塊所形成的最大(I域被包含在| η張的影像的對焦區域中時，稱之為移動資訊缺乏，則第n+1発 影像的對舰新為第n張㈣像㈣錢域；當物體移動時， 移動資訊將會增加’若移動f訊增加的程度在預期的範圍之内， 發明將以第n+1張的影像的移動區塊所形成的最大區域為依 據來更新對錢域；當第㈣_影像⑽_塊所形成的最 大區域的位置賴時，本發簡重触得第㈣_彡像中的移 動區塊中的最大區域的最左邊、最右邊、最上面、及最 素’藉此定義出新的對焦區域。 、象 1336589 度高於第-臨界值T1 (步驟23〇)時，代表影像已清晰至一定程 度’這時感興趣區域判定模組Π0中的邊緣債測單元123將對該 影像進行邊緣偵測（步驟251)，藉以偵測出二元值的邊緣二元影 像。其中，邊緣债測法採用肯尼邊緣侦測法娜 ―)，其包含五個步驟’分別是高斯模糊、梯度計算、振幅 計算、非最大抑制(nc^axima supp聰iGn)、以及遲滯門檻值 (^ysteresis threshGld)。首祕輸人秘像與冑斯;纽$做摺積運 算’高斯模糊能絲容易造成運算錯誤的雜訊使影像看起來更 ^平滑’β將高纏罐㈣像賴對χ方向及γ方向做梯度運 二曰付到兩個相對的邊界影像梯度值。振幅計算運算出兩邊界 $像梯度關的振幅。非最大抑制_單—響應法則使它消除粗 線造成多餘的振“只齡到最強的振幅，如此將可得到較為 精準的邊界最後以遲滞門檻值提供兩個不同的門捏值來消除多 餘的邊界並修飾物件彼此間的輪廓。在本實施例中，邊緣二元影 像中的值為1的像素代表影像中物體的邊緣部份，而邊緣二元影 像中的值為〇的像素代表非邊緣部份如此便可域出該影像中 f個物體的邊緣。本發明並不以1作為物體的邊緣為限，意即也 …使=1來表不非邊緣的部份；使用G來表示物體的邊緣。 接者]感興趣區域判定模組12〇中的映射單元以會將上述 二、緣疋办像與區塊二元影像進行集合關聯性的運算，當邊緣 ^影像中值為1的像素映射到區塊二元影像的值也為！時（步 二^2) ’則該像素即為移動邊緣，在經過映射的步驟後，可以標 斤有的移動邊緣。在本實施例中’將對邊緣二元影像5⑴依 14 1336589 序執行水平線條擷取(Horizontal line extracti〇n)、垂直線條擷取 (Vertical line extraction)、水平線條擷取後，可以取得由移動的物 體的邊緣所包圍的-個區域540，如「第5A圖」所示，而區域 540的邊界（boundaiy)，也就是移動邊緣541，便是對焦區域， 因此感興舰域判定模組12G中的清晰影像對焦區域設定單元 125 (步驟 254)。 另外為了使對焦區域更為精確，在清晰的影像的對焦過程 中，感興趣區域判定模組120更可以包含精鍊單幻29，負責對 移動邊緣541進行精鍊（Refinement)的動作（步驟253)，、以取 得較為清晰的-健域56〇 ’也就是將上述取得的軸邊緣541 再次與邊界二元影像510比對，藉明除額外產生的部份，如「第 5B圖」所示，如此，將可使得對焦區域更為精確。 經由上述’不論輸人的影像屬於模糊或是清晰，在經過本發 明之感興趣區域判賴組12〇的處理後，都會得到—個對焦區 域’對焦核組130將由對焦區域的解析度來進行對焦（步驟_。 如此，本發卿可鱗絲法自騎絲在_的影財益法正 確的對焦的問題。 本發明之可自動侧對舰域之方法，可實現於硬體 佈於;=之組合中’亦可在轉中方式實現或以獨元件散 佈於右干紐逹接之電子裝㈣分散方式實現。 =本料贿述之触實施·露如上，财麟用以限 :本^2何鮮相像者，在不脫離本_之精神和範圍 所為之更動細飾，觸本㈣之專利賴麵因此 丄 明之專梅職_視本制書_•之申請專利 準0 , 【圖式簡單說明】 第1圖係本發騎提之可自_測對越域之影像 之組成元件示意圖。 第2圖係本發明所提之可自動_對舰域之方法的流程In order to prevent the area formed by the moving block from being too fragmented or broken, the present invention may further include an amplifying unit 128 in the _ _ _ 12G, which is responsible for selecting the largest area (10) in the stencil setting unit 121. Before the 3 fields (step 243), the moving blocks in the image are first formed (face 242), so that the fine-grained area is the zero-broken area, and the area is now combined with the area. The area 320 is as shown in "3C". ~ In addition, the image of her paste is moving _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _Shooting mobile normalization unit m, 编娜彡峰峰幻幻2==large regional training as the focus area (step 243), will move the system after the end of the tear _ money (10) (spicy 241) == focal area More precise. Wherein, the relationship between the binary value of the block of the normalized binary operation and the eight adjacent H7C values around the block, the order and relationship are: the successive block binary image t values are The moving block of G is judged. If the value of the mobile block = 17 block 410 is 〇, and there are four or more moving blocks around the adjacent one, the mobile block in the operation is correspondingly The binary value is set to i, as shown in Fig. 4A; and then the block of the binary image of each of the normalized blocks is filled, and if the value of the block 420 is 1 When there are four or more secret blocks whose value is Q, the binary value of the moving block 420 in the determination is set to 〇, as shown in "Fig. 4B". In addition, 'when there is a moving object in the image, the object in the continuous image also moves, so the job area may change, so the fuzzy image job area setting list = 121 will continuously correct the focus area, in this In the embodiment, whether the mobile broadcast signal is lacking, the mobile poor information is increased, and the location of the mobile area is changed, whether the domain verification is to be updated, but the triggering time of the present invention for updating the color domain is not in the above three situations. Limited. The rich n+1 is the largest formed by the _ secret block (I domain is included in the focus area of the | η image, it is called the lack of mobile information, then the n+1 発 image of the new ship The nth (four) image (4) money field; when the object moves, the mobile information will increase 'if the degree of increase of the mobile f signal is within the expected range, the invention will be formed by the moving block of the n+1th image The maximum area is updated to the money field; when the position of the largest area formed by the (4)th_image (10)_block depends, the local touches the largest area of the moving block in the (4)th image The leftmost, the far right, the top, and the most 'by this define a new focus area. If the 1336589 degree is higher than the first threshold T1 (step 23〇), it means that the image is clear to a certain extent. The edge debt detecting unit 123 in the region of interest determination module Π0 will perform edge detection on the image (step 251), thereby detecting the binary binary image of the binary value. Among them, the edge margin measurement method uses Kenny edge detection Measured Na-), which consists of five steps 'Gass fuzzy, gradient calculation respectively Amplitude calculation, the non-maximum suppression (nc ^ axima supp Cong IGN), and a hysteresis threshold (^ ysteresis threshGld). The first secret is to enter the secret image and the muse; New York to do the convolution operation 'Gaussian fuzzy energy is easy to cause the operation error of the noise to make the image look more smooth 'β will be high cans (4) like the opposite direction and γ direction The gradient is applied to two relative boundary image gradient values. The amplitude calculation calculates the amplitude of the two boundaries $image gradient off. The non-maximum suppression_single-response rule eliminates the extraneous vibration caused by the thick line "only the strongest amplitude, so that a more accurate boundary can be obtained. Finally, two different gate values are provided by the hysteresis threshold to eliminate the excess. The boundary of the object and the contour of the object. In this embodiment, the pixel with a value of 1 in the edge binary image represents the edge portion of the object in the image, and the pixel in the edge binary image has a value of 〇. The edge portion can thus represent the edge of f objects in the image. The invention is not limited to 1 as the edge of the object, that is, also to make = 1 to represent the part of the edge; use G to represent The edge of the object. The mapping unit in the region of interest determination module 12〇 performs the association operation between the second image and the binary image of the block, and the value in the edge ^ image is 1 The value of the pixel mapped to the block binary image is also ! (step 2^2) 'The pixel is the moving edge, and after the step of mapping, the moving edge can be marked. In this embodiment 'will be on the edge binary image 5 (1) 14 1336589 The horizontal line extracti〇n, vertical line extraction, and horizontal line extraction can obtain the area 540 surrounded by the edge of the moving object, such as As shown in Fig. 5A, the boundary ailaiy of the area 540, that is, the moving edge 541, is the focus area, so the clear image focus area setting unit 125 in the shipboard determination module 12G is sensed (step 254). In addition, in order to make the focus area more precise, in the focusing process of the clear image, the region of interest determination module 120 may further include a refinement phantom 29, which is responsible for refining the movement edge 541 (step 253). In order to obtain a clearer - health domain 56', the axis edge 541 obtained above is again compared with the boundary binary image 510, except for the extra generated portion, as shown in "Fig. 5B". Will make the focus area more precise. After the above-mentioned 'whether the input image is blurred or clear, after the processing of the region of interest of the present invention is determined to be 12〇, the focus area will be obtained. The focus core group 130 will be determined by the resolution of the focus area. Focusing (step _. So, the hair of the hair can be used to align the correct focus of the imaginary method. The method of the invention can automatically realize the side of the ship, can be realized in the hard cloth; In the combination of = can also be realized in the transfer mode or in the electronic device (four) scattered mode in which the individual components are scattered in the right-hand side. The implementation of the bribes of the material is as above, and the use of Cailin is limited to: ^2There is no change in the spirit and scope of this _, and it is more precise and fine-cut, and the patent of the (4) patent is based on the esthetic work of the _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ Brief description of the model] The first figure is a schematic diagram of the components of the image of the cross-domain that can be self-tested. The second figure is the flow of the method of the invention that can automatically
範圍所界定者為 擷取裝置 第3A圖係本發明實施例所提之移動區塊示意圖。 第3B圖係本發明實施例所提之對焦區域示意圖。 第冗圖係本發明實施例所提之放大移動區塊示意圖。 第4A圖係本發明實施例所提之移動正規化示意圖。 第4B圖係本發明實施例所提之移動正規化示意圖。 第5A圖係本發明實施例所提之映射過程示意圖。The scope is defined as the capture device. FIG. 3A is a schematic diagram of the mobile block proposed by the embodiment of the present invention. FIG. 3B is a schematic diagram of a focus area proposed by the embodiment of the present invention. The redundant diagram is a schematic diagram of an enlarged moving block proposed by the embodiment of the present invention. FIG. 4A is a schematic diagram of mobile normalization proposed by the embodiment of the present invention. FIG. 4B is a schematic diagram of the mobile normalization proposed by the embodiment of the present invention. FIG. 5A is a schematic diagram of a mapping process according to an embodiment of the present invention.
第5B圖係本發明實施例所提之精鍊過程示意圖。 【主要元件符號說明】 100 影像擷取装置 110 移動偵測模組 120 感興趣區域判定模組 121 模糊影像對焦區域設定單元 123 邊緣偵測單元 124 映射單元 125 清晰影像對焦區域設定單元 127 移動正規化單元 (S >， 16 1336589 - 128 . 129 130 301 302 303 304 • 305 310 311 312 313 314 320 320a 320b 410 420 510 540 560 541 步驟210 放大單元 精鍊單元 ’ 對焦模組 虛線 虛線 虛線 虛線 方框 最大區域 最左像素 最右像素 最上像素 最下像素 區域 區域 區域 移動區塊 移動區塊 邊緣二元影像 區域 區域 移動邊緣 對影像進行移動偵測 17 1336589 步驟241對區塊二元影像進行移動正規化 步驟242放大移動區塊形成之區域 步驟243設定對焦區域 步驟251對影像進行邊緣偵測 步驟252將邊緣二元影像映射至區塊二元影像 步驟253精鍊移動邊緣 步驟254設定對焦區域 步驟260自動對焦Figure 5B is a schematic view of the refining process of the embodiment of the present invention. [Main component symbol description] 100 image capturing device 110 motion detecting module 120 region of interest determining module 121 blur image focusing region setting unit 123 edge detecting unit 124 mapping unit 125 clear image focusing region setting unit 127 mobile normalization Unit (S >, 16 1336589 - 128 . 129 130 301 302 303 304 • 305 310 311 312 313 314 320 320a 320b 410 420 510 540 560 541 Step 210 Amplification Unit Refining Unit ' Focus Module Dotted Line Dotted Line Dotted Line Box Maximum Area leftmost pixel Rightmost pixel Topmost pixel Lowermost pixel area Area moving block Moving block edge Binary image area Moving edge to image motion detection 17 1336589 242. Enlarging the area in which the moving block is formed. Step 243: Setting the in-focus area. Step 251: Performing edge detection on the image. Step 252: Mapping the edge binary image to the block binary image. Step 253: Refining the moving edge. Step 254: Setting the in-focus area.