1267737 1 黍 九、發明說明: 【發明所屬之技術領域】 本發明係一利用視覺模型檢測平面顯示器之方法盘裝 置,藉視覺模型產生檢測圖與檢測值以評斷顯示器品質, 針對整體顯示面板考量’並且能直接藉由檢測方^得知整 體面板之品質優劣。 正 【先前技術】 隨著平面顯示器的普遍應用,如液晶顯示器(lcd) 大量使用在電視、電腦監視器、行動電話、各種家電上, 液曰日顯示态面板的品質在大量生產下也愈需要重視,如色 ,對比、反應日守間、党度等,而關於Lcd背景均勻度則 ,祭 LCD 輝度(luminance)異常(MURA)狀況,mura f指顯示器亮度不均勻造成各種痕跡的現象,最簡單的判 ^方法就是在切換到黑色晝面以及其他灰階晝面,鋏後從 ^不同的角度仔細去看,隨著各式各樣的製程械,液 =示ϋ就有各式各_MURA,各樣瑕絲橫向條紋或 一:五度祕紋、方塊、圓形塊,亦可能是某個角落出現 声^ *此颁瑕疵低對比、不規則顯示,故MURA的嚴重程 度亦為判斷液晶顯示器面板品質的重要參考。 =習知技術中’針對MURA的判斷係直接觀察液晶 不器晝面之瑕疵,其仍有下列缺點: I不易檢測、取像。 2·不易對各種情況之MURA分類,且各家廠商定義 不一致,而欠缺指標。 1267737 3·並無描述MURA嚴重程度的標準,或是太過粗糙。 國際組織 VESA ( Video Electronics Standards Association)有針對各類型MURA作定義(VESA FPDM2 303-8),但也僅止於簡單的定義,並無定義其嚴重程度與 其判斷的解決方案。 而習用相關之技術中有部分提出MURA的偵測裝置 與方法,其中US6,154,561案中有定義MURA類別,如第 一 A圖所示一包括有瑕疵之平面顯示器1〇的線條型 MURA ’其為與鄰近相異且不正常的畫素(pixei),產生如 直線、曲線c、L型線條a、垂直線條b、細線條e、粗線 條f等MURA。又如第一 b圖所示區塊型MURA,包括暗 點區塊i、亮點區塊g等MURA,還有面板邊界產生的邊 緣區塊h與邊緣亮點區塊j等MURA。 而如美國專利US5,917,935案則藉由設定一門檻值, 將MURA瑕疲與背景值作一比較,因而得出MURA瑕疫 的程度,作成統計表格,並藉第二圖所示流程與此統計表 格來分類與定義各種不同的MURA瑕疵狀態。步驟包括: 由顯示面板取樣一原始畫面(步驟201);由此原始畫面產 生複數個子樣本晝面(步驟202);依不同需求過濾每一子 樣本晝面,如依據晝面之各樣特性圖(histogram)作特定 之特徵過濾(步驟203);再對每一晝面設定一門檻值,藉 以產生特徵區塊(Blob)(步驟204);並由特徵區塊之分 析,判斷該原始晝面之MURA瑕疵(步驟205);並特徵 化(characterize)其MURA瑕疵(步驟206);之後,執 行最後調整動作,藉消除錯誤偵測來決定某MURA類別 1267737 (步驟207)。 藉上述專利之步驟僅將各種MURA分類,如線條 MURA( Line Mura )、斑點 MURA( Spot Mura )、填充 MURA (Fill Port Mura)、邊緣 MURA ( Panel Edge Mura)、區塊 MURA (Block Mura)等,並無針對各樣圖像(pattern) 判斷顯示面板之品質。 於2003年國際顯示器研討會(internati〇nai Display Workshops,IDW )中發表一種利用最小可覺差異JND ( Just1267737 1 发明 、, invention description: [Technical field of invention] The present invention is a method for detecting a flat panel display using a visual model, and generating a detection map and a detection value by a visual model to judge the quality of the display, and considering the overall display panel And the quality of the overall panel can be known directly by the detecting party. [Prior Art] With the widespread use of flat panel displays, such as liquid crystal displays (LCDs), which are widely used in televisions, computer monitors, mobile phones, and various home appliances, the quality of liquid-liquid display panels is also needed in mass production. Pay attention to, such as color, contrast, reaction, day-to-day, party, etc., and regarding the background uniformity of Lcd, the brightness of the LCD is abnormal (MURA), and mura f refers to the phenomenon that the brightness of the display is uneven, causing various traces. The simple method of judging is to switch to the black enamel surface and other gray-scale enamel surfaces, and then carefully look at it from different angles. With various kinds of manufacturing tools, liquid = demonstration ϋ has various _ MURA, all kinds of horizontal stripes or one: five-degree secret lines, squares, round blocks, or a corner sound ^ * This is a low contrast, irregular display, so the severity of MURA is also judged An important reference for the quality of LCD panels. In the prior art, the judgment of MURA directly observes the flaws of the liquid crystal, and it still has the following disadvantages: I is difficult to detect and take images. 2. It is not easy to classify MURA in various situations, and each manufacturer's definition is inconsistent, but lacks indicators. 1267737 3. There is no standard describing the severity of MURA, or it is too rough. The international organization VESA (Video Electronics Standards Association) has defined MURA for each type (VESA FPDM2 303-8), but it is only limited to a simple definition, and there is no solution to define its severity and its judgment. Some of the techniques related to the use of the MURA detection device and method, wherein the US6,154,561 case defines the MURA category, as shown in Figure A, a line type MURA including a flat panel display In order to be different from the neighboring and abnormal pixels (pixei), MURA such as a straight line, a curved line c, an L-shaped line a, a vertical line b, a thin line e, and a thick line f are generated. Another example is the block type MURA shown in the first b diagram, including the MURA such as the dark spot block i and the bright spot block g, and the MURA such as the edge block h and the edge bright spot block j generated by the panel boundary. For example, in US Patent No. 5,917,935, by setting a threshold, the MURA fatigue is compared with the background value, and the degree of the MURA plague is obtained, and a statistical form is created, and the flow shown in the second figure is used. Statistical tables to classify and define various MURA瑕疵 states. The step includes: sampling an original picture by the display panel (step 201); thereby generating a plurality of sub-sample faces by the original picture (step 202); filtering each sub-sample face according to different requirements, such as according to various characteristic maps of the facet (histogram) for specific feature filtering (step 203); then setting a threshold value for each facet to generate a feature blob (step 204); and judging the original face by analysis of the feature block MURA瑕疵 (step 205); and characterizes its MURA瑕疵 (step 206); thereafter, a final adjustment action is performed to determine a certain MURA category 1267737 by eliminating error detection (step 207). The above-mentioned patent steps only classify various MURAs, such as line MURA (Line Mura), spot MURA (Spot Mura), filled MURA (Fill Port Mura), edge MURA (Plane Mura), block MURA (Block Mura), etc. The quality of the display panel is not judged for each pattern. Presented in the 2003 Internati〇nai Display Workshops (IDW) with a minimum sensible difference JND (Just
Noticeable Difference)與 SEMU 定義作為 MURA 分析的 方法(Mura Analysis Method by Using JND luminance and the SEMU definition),針對各式MURA計算嚴重程度,但 也過於簡單與僅包含少數參考依據,不足藉以判斷面板品 質。上述最小可覺差異jND係習知表示人眼觀察之差異性 的方法。 而SEMI國際標準亦發表一種平面顯示榮幕晝質 MURA檢查之計量定義(SEMI D31-1102)。但以上針對 Mura定義或是分析皆限於顯示器之面積、對比、背景輝度 來考1,係過於簡單而不能完整呈現平面顯示器的瑕疵。 習知技術有下列缺點: L如用人工檢測,故可信賴度不高,時有爭議。 2·僅得到MURA物理上的描述或分類,並無考慮人本 身的感知,模型太簡單。 3·僅針對MURA的嚴重性加以描述,並無對顯示器品 質作判斷。 夕數自動檢測技術仰賴分段(segmentati〇n )檢測, 1267737 即針對各種不同MURA的圖像檢測,有太多參數 無法決定,若一旦有新的MURA,則需要新的渖算 法。 /、 然本發明之較佳實施例乃引用習知JND圖,i習知應 用可參襲國專利US6,285,797揭露—種藉產生細圖^ 估計影像品質之方法。 本發明係一利用視覺模型檢測平面顯示器之方法與裝 置,利用感光元件取像,如電荷耗合元件(chargeco响d Device, CCD)或金氧互補半導體(CM〇s)等感光元件達 士二!引入一參考圖像’避免人工檢測的缺點,再利用特 ,演算方式與視覺模型產生—檢測圖與—檢測值,如最小 可覺差異圖UND map )及励值,藉以評斷顯示器品質, 即針對整體顯示面板考量,且並不針對MURA定義,而直 接藉由檢測方法得知整體面板之品質優劣。 【發明内容】 本發明係一利用視覺模型檢測平面顯示器之方法與裝 置,藉一取像系統攝取影像,再利用視覺模型產生一檢測 圖(如励map)及-檢測值(如她值),藉以評斷顯 不器品質’針對整體顯示面板考量,並且能直接藉由檢測 方法得知整體面板之品質優劣。 本發賴述之裝置包括有:-差_識產生系統;搞 接差異韻產生系統之一取像系統;輕接差異辨識產生系 統之一背景模擬系統,以產生一參考圖像;一視覺 統,係導入一視覺模型至差異辨識產生系統= = 1267737 賁^產生系統,係祕於差異輯產 芩考資訊。 以產生檢测 其檢測方法步驟包括有··由一取像车 圖像產生待測面板圖像;由背景模擬產生參考;象測= 測面板圖像與該參考圖像經—差異辨識產將待 流程;結合辨識流程產生之測試f訊產生_檢=識 識流程產生之測試資訊取-平均值產生檢測值;7及= 該檢測圖與該檢測值作為該顯示器之品f評斷」用 器之優劣。 /丁卸頌不 【實施方式】 本發明係-利用視覺模型檢測平面顯示器之方法盘壯 置,利用感光元件,如電荷輕合元件(CCD) = 半導體(CMOS)等感光元件達成取像,再利用視覺^ 產生一檢測目(如最小可覺差異圖(JNDmap))及一檢 值卿值),藉以評斷顯示器品質,針對整體顯二面 板考里而直接藉由檢測方法得知整體面板之品質優劣。 本發明至少有下列優點: ' L習用檢測機制仍採用可信賴度不高的人工檢測,而 本發明避免人工檢測平面顯示器的爭議。 2·針對MURA的嚴重程度提供一精確且合理的測度。 3·不僅描述MURA的嚴重程度,亦可藉以得知顯示器 面板之優劣。 4·建立一針對平面顯示器的MURA的測度標準,解決 廠商之間對顯示器品質的爭議。 1267737 5·並不分段(segmentati〇n)檢測,即不針對各種不同 MURA進行分析,而是整體顯示面板考量,省 略因為新的MURA產生而需要新的演算之過 程0 6·利用視覺模型中導入檢測圖(如最小可覺差異圖 (JND map))之產生,於視覺模型中會考慮 形狀、面積、對比等因素,因此,即使未來因 不同的製程而出現不同的MURA時,仍可評 斷品質。The Noticeable Difference and the SEMU Definition Method by Using JND luminance and the SEMU definition calculate the severity for each MURA, but it is too simple and contains only a few references, which is not enough to judge the quality of the panel. The above-mentioned minimum sensible difference jND is a conventional method of expressing the difference in human eye observation. The SEMI International Standard also publishes a measurement definition for the planar display of the glory MURA inspection (SEMI D31-1102). However, the above definition or analysis for Mura is limited to the area, contrast, and background luminance of the display. It is too simple to fully display the flaws of the flat panel display. The prior art has the following disadvantages: L. If it is manually detected, the reliability is not high and it is controversial. 2. Only the physical description or classification of MURA is obtained, without considering the perception of the person himself, the model is too simple. 3. Only describe the severity of MURA and do not judge the quality of the display. The automatic detection technology of the singularity relies on the segmentation detection. 1267737 is the image detection for various MURAs. There are too many parameters that cannot be determined. If there is a new MURA, a new calculation method is needed. The preferred embodiment of the present invention is a reference to a conventional JND diagram, which is disclosed in US Patent No. 6,285,797, the disclosure of which is incorporated herein by reference. The invention relates to a method and a device for detecting a flat panel display by using a visual model, and uses a photosensitive element to take an image, such as a charge-consuming component (charge-corresponding device, CCD) or a gold-oxygen complementary semiconductor (CM〇s), etc. ! Introducing a reference image 'avoiding the shortcomings of manual detection, reusing special, calculus and visual model generation - detection map and - detection value, such as UND map and excitation value, to judge the quality of the display, that is, for the whole The display panel considers and does not define the MURA, but directly detects the quality of the overall panel by the detection method. SUMMARY OF THE INVENTION The present invention is a method and apparatus for detecting a flat panel display using a visual model, taking an image by an image taking system, and then generating a detection map (such as an excitation map) and a detection value (such as her value) using a visual model. In order to judge the quality of the display, it is considered for the overall display panel, and the quality of the overall panel can be directly detected by the detection method. The device of the present invention includes: a difference-aware generation system; an image acquisition system of the difference difference generation system; a background simulation system of the difference identification generation system to generate a reference image; , the introduction of a visual model to the difference identification generation system = = 1267737 贲 ^ production system, the system secrets the difference between the production and reference information. The step of detecting the detection method includes: generating a panel image to be tested from a image of the image capturing vehicle; generating a reference by background simulation; and measuring the image of the panel and the reference image by using the difference between the image and the reference image Waiting for the process; the test generated by the identification process generates the test information generated by the identification process. The test information generated by the identification process takes the average value to generate the detected value; 7 and = the test pattern and the detected value are used as the product f of the display. The pros and cons. [Invention] The present invention is a method for detecting a flat panel display by using a visual model, and a photosensitive element such as a charge-receiving element (CCD) = semiconductor (CMOS) is used for image capturing, and then Use the visual ^ to generate a test object (such as the minimum sensible difference map (JNDmap) and a check value) to judge the quality of the display, and directly determine the quality of the overall panel by means of the detection method for the overall display panel. Good or bad. The present invention has at least the following advantages: The 'L-use detection mechanism still uses manual detection with low reliability, and the present invention avoids the controversy of manually detecting a flat panel display. 2. Provide an accurate and reasonable measure of the severity of MURA. 3. Not only describe the severity of MURA, but also the pros and cons of the display panel. 4. Establish a measurement standard for MURA for flat panel displays to resolve disputes between manufacturers regarding display quality. 1267737 5·not segmentation (segmentati〇n) detection, that is, not for different MURA analysis, but the overall display panel consideration, omitting the process of new calculus due to the new MURA generation 0 6·Using the visual model The introduction of detection maps (such as the minimum sensible difference map (JND map)) will take into account the shape, area, contrast and other factors in the visual model, so even if different MURAs appear in different processes in the future, they can still be judged. quality.
7· t過本發明所使用之檢測圖(如JND map)可突顯 不明顯、不清楚之MURA瑕疵。 壯第三圖所示為本發明利用視覺模型檢測平面顯示器之 農置示意圖,其目的係為藉本發明之裝置#產生檢測圖(其 較佳實施例如JND圖)與檢測值(如JND值)對一平面 顯示器面板31之優劣作—判斷,係歧—取像系統3 2内 之感光元件取像’如電荷耗合元件(CCD)或金氧互補半The detection map (such as JND map) used in the present invention can highlight the MURA瑕疵 which is not obvious and unclear. The third figure is a schematic diagram of the invention for detecting a flat panel display using a visual model, the purpose of which is to generate a detection map (a preferred embodiment thereof such as a JND map) and a detection value (such as a JND value) by the apparatus # of the present invention. The merits and demerits of a flat panel display panel 31 are judged, and the photosensitive elements in the image capturing system 32 are imaged, such as a charge-consuming component (CCD) or a gold-oxygen complementary half.
^體(C^OS) # ’由圖像擷取系統34_取待測顯示器影 像’此時藉-背景模擬系統33依實際環境需要導入相關背 景,使所擷取之待測影像有—對照之背景值,以期得到客 觀之影像檢測。 將擷取之影像與模擬背景值導人__差異辨識產生系袭 35’利用特定演算方式與視覺模型經由輸出系統%產生_ ¥ 檢測圖3:與-檢測值38,此輸出系統%可以一顯示g 幕、圖表等方式實施,檢測者可藉其檢測圖37與檢 38判斷面板31之優劣。 、 11 1267737 然第四圖則揭露本發明各部系統連接概圖,本發明為 突顯平面顯示器中MURA瑕疵,即藉一以硬體或軟體實現 之差異辨識產生系統45,由耦接於差異辨識產生系統45 之取像系統41得到待測顯示器之影像晝面,另外,再由耦 接於差異辨識產生系統45之背景模擬系統43得出一背景 值。此差異辨識產生系統45再導入模擬人眼視覺系統之視 覺模型47 ’其中加入觀點條件(Viewing Condition) 46, 使玎導入由檢測者(如面板廠商)彈性加入之調整值,使 產生出來的檢測值可以適當反映不同測試環境的結果;亦 f導入一模型調校(Model Calibration) 48,係提供一系統 _ 性之模型,藉改變其中輸入之參數以適應各種測試環境; 最後藉耦接於差異辨識產生系統45之檢測資訊產生系統 49產生檢測參考資訊,檢測資訊產生系統49可耦接一輸 出系統(請參閱第三圖),如一顯示器或一圖表方式,如本 發明所應用之檢測圖(如JND圖)與檢測值(如值)。 第五圖係為本發明利用視覺模型檢測平面顯示器之方 法步驟流程,流程包括待測面板摘取與引入背景,並產生 待測圖像與參考圖像,再經差異辨識產生系統產生辨識流⑩ 程,步驟如下: 開始時,由取像系統擷取待測面板圖像(步驟51); 調整擷取圖像,如修正圖樣像差、消除雜訊、調整圖 像角度、比例、大小、轉換成輝度值等(步驟52); 產生待測圖像(步驟53 ); 同時,考慮檢測環境,以模擬背景圖像(步驟54); 產生參考圖像,此參考圖像可以為均勻的背景,亦可 12 1267737 為一般人可以接受的不均勻背景圖像,亦可由此取像系統 本身所產生的不均勻圖像(步驟55); 將待測圖像與翏考圖像皆經上述之差異辨識產生系統 產生辨識流程,包括前置處理、估計圖像資訊與整合各參 數等步驟。此辨識流程並可導入觀點條件(viewing Condition),使檢測者可彈性加入之調整值,使產生出來 的檢測值可以適當反映不同測試環境的結果,亦可導入一 模型調校(Model Calibration),係提供一系統性之模型, 可藉改變其中輸入之參數以適應各種測試環境(步驟56); 削置處理,係包括找出圖像邊界,擷取屬於顯示器的 範圍、圖像大小調整,邊界補償,並可模擬檢測環境之環 境光,再加以修正等(步驟561,564); 估a十圖像資訊,此時導入複數個流程之參數,如人因 參數(Human Factor),包括如下所述之步驟(其詳細内容 請參閱第八圖,在此並不詳述): a· 光適應(Light Adaptation)修正; b· 金字塔解構(Pyramid Decomposition); c· 對比計算(Contrast Computation ); d· 對比敏感度過濾(Contrast Sensitivity^体(C^OS) # 'Image capture system 34_takes the display image to be tested' At this time, the background simulation system 33 introduces the relevant background according to the actual environment, so that the captured image to be tested has The background value is to obtain objective image detection. The captured image and the simulated background value are guided by the __ difference identification to generate the attack 35' using the specific calculation mode and the visual model is generated via the output system % _ ¥ detection Fig. 3: and - detection value 38, the output system % can be one The g screen, the graph, and the like are implemented, and the tester can use the test to determine the advantages and disadvantages of the panel 31 and the check panel 31. 11 1267737 However, the fourth figure discloses the connection diagram of each part of the system of the present invention. The present invention is to highlight the MURA瑕疵 in the flat panel display, that is, by using a hardware or software to implement the difference identification generation system 45, which is coupled by differential identification. The image capturing system 41 of the system 45 obtains the image plane of the display to be tested, and further, a background value is obtained by the background simulation system 43 coupled to the difference recognition generating system 45. The difference identification generation system 45 re-imports into the visual model 47' of the simulated human visual system, in which the Viewing Condition 46 is added, and the 玎 is introduced into the adjustment value elastically added by the detector (such as the panel manufacturer), so that the generated detection is generated. The value can appropriately reflect the results of different test environments; also introduce a Model Calibration 48, which provides a system _ sex model, by changing the parameters input to adapt to various test environments; finally by coupling to the difference The detection information generating system 49 of the identification generating system 45 generates detection reference information, and the detection information generating system 49 can be coupled to an output system (refer to the third figure), such as a display or a chart mode, such as the detection chart applied by the present invention ( Such as JND map) and detection values (such as values). The fifth figure is a flow chart of the method for detecting a flat display by using a visual model. The process includes extracting and introducing a background of the panel to be tested, and generating an image to be tested and a reference image, and generating a recognition stream by the difference identification generation system. The procedure is as follows: At the beginning, the panel image to be tested is captured by the image capturing system (step 51); the captured image is adjusted, such as correcting the pattern aberration, eliminating noise, adjusting the image angle, proportion, size, and conversion Generating a luminance value or the like (step 52); generating an image to be tested (step 53); meanwhile, considering the detection environment to simulate a background image (step 54); generating a reference image, which may be a uniform background, 12 1267737 can also be an uneven background image acceptable to the average person, and can also obtain an uneven image generated by the image system itself (step 55); the image to be tested and the reference image are identified by the difference described above. The generation system generates an identification process, including pre-processing, estimating image information, and integrating various parameters. The identification process can introduce a viewing condition, so that the tester can flexibly add the adjustment value, so that the generated detection value can appropriately reflect the results of different test environments, and can also be introduced into a Model Calibration. Provide a systematic model, which can be adapted to various test environments by changing the parameters input therein (step 56); cutting processing includes finding the image boundary, capturing the range belonging to the display, image size adjustment, and boundary Compensation, and can simulate the ambient light of the detection environment, and then correct it (steps 561, 564); estimate a ten image information, at this time introduce parameters of a plurality of processes, such as human factor, including the following Steps (for details, please refer to the eighth figure, which is not detailed here): a· Light Adaptation correction; b· Pyramid Decomposition; c· Contrast Computation; d· Contrast Sensitivity filtering (Contrast Sensitivity
Filter); e· 傾斜效應(Oblique Effect)修正; f· 孔控(Aperture)修正; g· 頻道過濾(Channel Filter)。 藉上述各參數修正達到由各種不同的參考值來估計圖 像貪訊之目的,以求得更準確的檢測數據(步驟562,565); 13 1267737 整^圖像資訊步驟,係擷取來自步驟562,565中待測 圖像與翏^圖像的圖像資訊,考慮不同的測試資訊(頻道) 契之間的父互影響’將估計圖像資訊步驟中產生各頻道 (Channel)的結果作進一步修正(步驟563,566); /5各’則我資訊(頻道),以第四圖所示之差異辨識產 ^系統產生一檢測圖,如一最小可覺差異(JND)圖,此 檢測圖可為f頻道的檢卿,也可以是職和彩度的檢測 圖’也可以是综合上述各測試參數的檢測圖 (步驟57); 取平句值其一貫施例可以Minkowski pooling平均值 方/去(步驟58)把各檢測圖整合成一個檢測值(如JND 值)(步驟59); ^利用檢測值與檢測圖作為平面顯示器影像品質的評 斷以,斷面板優劣,同時可以由檢測圖中得知人眼感受 到差異最大的區域(步驟60 )。 第六圖則為第五圖所示調整擷取圖像步驟52之詳細 流程,其步驟包括有: &、/周正#1取®像步驟開始,預備—待測顯示器面板與各 榀測儀器(步驟6〇1); 藉取待測面板圖像,可以—電荷辆合元件 ^ ^戈金氧互補半導體(CMC)S)等感光元件達成(步 驟 603 ); 校正擷取圖像,如輕#s_像縣、消除雜訊、調 -圖像角度、比例、大小等(步驟605);Filter); e· Oblique Effect correction; f· Aperture correction; g· Channel Filter. The above parameters are modified to achieve the purpose of estimating image corruption by various reference values to obtain more accurate detection data (steps 562, 565); 13 1267737 The entire image information step is taken from steps 562, 565. The image information of the image to be tested and the image of the image, considering the parental interaction between different test information (channels), the result of the channel generated in the estimation image information step is further corrected (steps) 563, 566); /5 each 'my information (channel), with the difference identification system shown in the fourth figure to generate a detection map, such as a minimum sensible difference (JND) map, this test map can be the f channel check Qing, can also be the test chart of job and chroma 'can also be the test chart of the above test parameters (step 57); take the sentence value of its consistent application can be Minkowski pooling average / go (step 58) The detection maps are integrated into one detection value (such as JND value) (step 59); ^Using the detection value and the detection map as the judgment of the image quality of the flat panel display, the quality of the broken panel is good, and the difference between the human and the eye can be seen from the detection map. Large area (step 60). The sixth figure is a detailed process of adjusting the captured image step 52 shown in the fifth figure, and the steps include: &, / Zhou Zheng #1 take the image like the step start, the preparation - the display panel to be tested and the various test instruments (Step 6〇1); borrowing the image of the panel to be tested may be achieved by a photosensitive element such as a charge-carrying component ^^Golden-oxygen complementary semiconductor (CMC) S) (step 603); correcting the captured image, such as light #s_像县, elimination of noise, tone-image angle, ratio, size, etc. (step 605);
圖像經轉換後取樣,如數位處理,將顯示關像轉換 蜀立於檢測設備之影像,如輝度值(luminanee)或CIE 1267737 國際照明委員會採用的色彩空間(CIELAB),以期得到客 · 觀之檢測結果,不受環境與設備瑕疵影響(步驟607)。 而第七圖為第五圖所示步驟56產生辨識流程中之前 置處理步驟流程,步驟包括有: - 前置處理步驟開始(步驟701) ; 處理步驟包括圖像邊緣偵測,以得到正確的圖像邊 緣’降低檢測誤差(步驟7〇3); 作邊界補償,即修正適當邊界寬度及影像輝度 (luminance),以描述邊界帶來的效應(步驟7〇5) ; φ 修j£圖像為一適當大小,各檢測設備間所需之圖像若 無一致’則需依照其中所需而調整(步驟707); 修正環境參數,如修正環境光源,消弭環境給予的影 響’以期得到準確的檢測結果(步驟709); 第八圖係為本發明第五圖所示產生辨識流程中估計圖 像資訊步驟流程,步驟包括有: 估計圖像資訊步驟開始(步驟801); 依照人眼對於光的適應(light adaptation),依視覺模 Φ 型給予光適應調整,反映出人眼對於不同輝度的反應(步 驟 803 ); 以金字塔解構(Pyramid Decomposition)將一張圖像 解構產生不同尺度的影像(步驟805); 對比計算(Contrast Computation)算出各種尺度影像 的對比值(步驟807); 對比敏感度過濾(Contrast Sensitivity Filter)係依照 每個人對不同環境、空間頻率下有不同的對比敏感度,即 15 1267737 依此修正上述步驟805對比計算的結果(步驟8〇9) · 傾斜效應(Oblique Effect)修正係依照人對斜角的視 覺敏感度不同而設計,即對不同方向的敏感度修正上 驟807的結果(步驟811) ; 孔徑(Aperture)修正係依照人對影像中心及影像周 圍的敏感度不同而修正(步驟813) ; 、 口 頻道過濾(Channel Filter),將上述各參數修正使 圖像與參考圖像解構成不同的頻道(channel)f依照人董1 不同尺度的影像模式有不同的敏感度,如針對圖像解析 度、顯示頻率之差異而修正(步驟815)。 藉上述各參數修正達到由各種不_參考 像資訊之目的,以求得更準確的檢測數據。 。圖 之方’ ί發明為一利用視覺模型檢測平面顯示器 ^及_值’ _評斷顯示器品質,針對整體^ =反考i,並且能餘藉錢财法得 口、 質優劣。 胆叫做< 口口 以上,實為一不可多得之發明物品,及具產業上 新穎性及進步性,完全符合發明專射請要件,爰 =法提^申請詳查並轉本 4 之較佳可行實施例,非二ί :===化故r運用本發_^^ 合予陳明。,句同理包含於本發明之範圍内, 16 1267737 【圖式簡單說明】 第一 A圖與第一 B圖係為習用技術MURA偵測方法中 _ MURA之類別示意圖; 第二圖係為習用技術MURA的偵測方法流程圖; “ 第三圖所示為本發明利用視覺模型檢測平面顯示器之 裝置示意圖; 第四圖所示為本發明各部系統連接示意圖; 第五圖係為本發明利用視覺模型檢測平面顯示器之方 法步驟流程; ❿ 第六圖係為本發明調整擷取圖像步驟流程; 第七圖係為本發明前置處理步驟流程; 弟八圖係為本發明估計圖像貧訊步驟流程。 【主要元件符號說明】 L形線條a 垂直線條b 曲線c ® 細線條e 粗線條f 邊緣區塊h 亮點區塊g 暗點區塊i 邊緣党點區塊j 平面顯示器10 17 1267737 面板31 取像系統32 背景模擬系統33 圖像擷取系統34 差異辨識產生系統35 輸出系統36 檢測圖3 7 檢測值38 取像系統41 背景模擬系統43 差異辨識產生系統45 觀點條件46 視覺模型47 模型調校48 檢測資訊產生系統49The image is converted and sampled, such as digital processing, to display the image of the image, such as luminance (luminanee) or CIE 1267737 International Lighting Commission's color space (CIELAB), in order to get the guest view The test results are not affected by the environment and equipment (step 607). The seventh figure is the flow of the pre-processing step in the identification process in step 56 shown in the fifth figure. The steps include: - the pre-processing step starts (step 701); the processing step includes image edge detection to get the correct The edge of the image 'reduces the detection error (step 7〇3); performs boundary compensation, ie corrects the appropriate boundary width and image luminance to describe the effect of the boundary (step 7〇5); φ 修j£图If it is an appropriate size, if the image required between the detection devices is not consistent, then it needs to be adjusted according to the requirements (step 707); Correct the environmental parameters, such as correcting the ambient light source, eliminating the influence of the environment' in order to get accurate The detection result (step 709); the eighth figure is the flow of the step of generating the image information in the identification process shown in the fifth figure of the present invention, the steps comprising: starting the step of estimating the image information (step 801); Light adaptation, according to the visual mode Φ type, gives light adaptation, reflecting the response of the human eye to different luminances (step 803); Pyramid Decomposition The image deconstruction produces images of different scales (step 805); Contrast Computation calculates contrast values of various scale images (step 807); Contrast Sensitivity Filter is based on each person's different environments and spaces. There are different contrast sensitivities at the frequency, ie 15 1267737. Correct the results of the above steps 805 and compare the calculations (steps 8〇9). · The tilt effect (Oblique Effect) correction is designed according to the different visual sensitivity of the human to the bevel. That is, the result of the above step 807 is corrected for the sensitivity in different directions (step 811); the aperture correction is corrected according to the sensitivity of the person to the image center and the surrounding image (step 813); and channel filtering (Channel Filter) ), the above parameters are corrected so that the image and the reference image are decomposed into different channels. According to the image mode of different scales of the human body, there are different sensitivity, such as the difference between the image resolution and the display frequency. Correction (step 815). The above parameters are corrected to achieve the purpose of various non-reference image information to obtain more accurate detection data. . The figure of the figure ί is a visual model to detect the flat panel display ^ and _ value ' _ judge the quality of the display, for the overall ^ = retest i, and can borrow money and money to get the quality and quality. The daring is called < above the mouth, it is a rare invention item, and has the industry novelty and progressiveness, fully in line with the invention special requirements, 爰=法提^ application detailed investigation and transfer to this 4 Good feasible example, non-two: === The reason is to use this hair _^^ to Chen Ming. The sentence is included in the scope of the present invention, 16 1267737 [Simple description of the drawing] The first A picture and the first B picture are the class diagram of the MURA detection method in the conventional technology MURA detection method; the second picture is the usage The flow chart of the detection method of the technical MURA; "The third figure shows the schematic diagram of the device for detecting the flat display by using the visual model; the fourth figure shows the connection diagram of the various systems of the present invention; The flow of the method for detecting the flat display of the model; 第六 The sixth figure is the flow of the steps of adjusting the captured image of the present invention; the seventh figure is the flow of the pre-processing step of the present invention; the eighth figure is the estimated image poor for the present invention Step flow. [Main component symbol description] L-shaped line a Vertical line b Curve c ® Thin line e Thick line f Edge block h Highlight block g Dark spot block i Edge party block j Flat display 10 17 1267737 Panel 31 Image acquisition system 32 Background simulation system 33 Image capture system 34 Differential recognition generation system 35 Output system 36 Detection Figure 3 7 Detection value 38 Image acquisition system 41 Background simulation system 43 Difference Identification Generation System 45 Perspective Conditions 46 Visual Model 47 Model Tuning 48 Detection Information Generation System 49
1818