CN1758021A - Device and method,and recording medium for checking gradation unequal - Google Patents

Abstract

In an unevenness inspecting apparatus (1) for inspecting density unevenness of resist coated on a substrate (9), a picked-up image is compressed and low-pass filtering is performed on the image, and furthermore, high-pass filtering is performed on it by high-pass filtering part (4214). In the high-pass filtering part (4214), in a case where a high-pass window exists near an edge, the high-pass window is contracted and kept away from a region which is not a region including a target pixel whose value is to be obtained by the filtering. By the above operation, in the unevenness inspecting apparatus (1), a group of pixels which are subjected to high-pass filtering are limited to pixels in the high-pass window, in other words, pixels in a region which includes a target pixel, and the accuracy of inspection of unevenness around an edge is improved.

Description

Check the apparatus and method and the recording medium of gradation unequal

Technical field

The present invention relates to a kind of technology of checking the gradation unequal of object.

Background technology

All the time, for the employed various parts of the display surface of display device, by inspectoral visual unequal the detecting of bar shaped that produce the wire inequality exist local uneven of uneven whole inequality, the uneven part that exists with a form on whole or on vertical or horizontal carried out.

For example, following technology is disclosed in the flat 9-5057 communique of TOHKEMY: in the inequality of the light transmission of the shadow mask that color Braun tube is used is checked, to taking and the gradation data of the image that obtains from a side's of shadow mask interarea side irradiates light and from the opposing party's interarea side, carrying out smoothing by medium filtering handles and asks for the smoothing data, the standardized data that calculates based on utilizing the smoothing data that gradation data is carried out division arithmetic, show the image of having emphasized to answer the shadow mask of detected inequality, thereby realize the simplification of visual examination and the raising of checking precision.

In recent years, be not rely on visual and automatically detect inequality from the image of checking object, in order to check the object a little less than the light and shade contrast ratio, carry out emphasizing to handle with respect to the contrast of the image of object.; during the inequality of the resist on being coated on the panel that liquid crystal indicator uses is checked; at the large-scale glass substrate upper surface of a slice during, produce at the edge of the electrode part of the peripheral part of each panel and can not detect uneven wire zone or the uneven wire zone that detects precise decreasing with the multi-disc panel.The high-pass filtering that the main cause that produces this zone is considered to carry out before contrast is emphasized to handle for the concentration change of the low frequency of removing the effect that the overslaugh contrast emphasizes is handled.

Summary of the invention

The objective of the invention is to uneven testing fixture, the precision that near the inequality the edge of raising from the image that object obtains is checked at the gradation unequal of checking object.

Uneven testing fixture of the present invention comprises: maintaining part, and it keeps object; Shoot part, its interarea to object is taken; Image processing part, it carries out Flame Image Process to the object images of being obtained by shoot part and generates the image of finishing dealing with, image processing part has: the edge detecting element, its detect in object images the zone of different mutually concentration the border, be the edge; High-pass filtering portion, it carries out high-pass filtering to object images and handles; Contrast is emphasized portion, its contrast to the object images after being handled by high-pass filtering portion is emphasized, during the pixel value of the concerned pixel of high-pass filtering portion in asking for the image of finishing dealing with, the pixel groups that will be used in computing in object images is restricted to the pixel of paying close attention to the zone under the image in fact.

When high-pass filtering is handled, be restricted to the pixel in the zone under the concerned pixel in fact by the pixel groups that will be used in computing, can improve the precision that near the inequality the edge in the object images is checked.

In a preferred mode of the present invention, the high-pass filtering of high-pass filtering portion is handled and is based on the concerned pixel is that the pixel value of the whole pixel in the window of approximate centre carries out, in the time of near window is present in the edge, window is reduced, thereby window is avoided zone in addition, the affiliated zone of concerned pixel.Thus, characteristic and other the position that near the inequality in edge is checked is approximate.

In another preferred mode of the present invention, the high-pass filtering of high-pass filtering portion is handled, and only carries out based on the pixel value that with the concerned pixel is the pixel in zone in the pixel in the window of fixing size of approximate centre, under the concerned pixel.When this mode, can simply carry out the setting of window.

Preferred uneven testing fixture also has the uneven detecting element of value of the degree of the gradation unequal of asking for the indicated object thing based on finishing dealing with image by calculation process.

The present invention also is applicable to the uneven inspection method of the gradation unequal of checking object in addition, and then, be applicable to that also record makes computing machine carry out the recording medium of the program of this inequality inspection method.

Above-mentioned purpose and other purpose, feature, form and advantage also can be clear according to the detailed description of carrying out below of the present invention with reference to accompanying drawing.

Description of drawings

Fig. 1 is the figure of structure of the uneven testing fixture of expression first embodiment.

Fig. 2 is the planimetric map of expression objective table and shoot part.

Fig. 3 is the uneven figure that detects the flow process of action of expression.

Fig. 4 is the figure that amplifies the part of expression second image.

Fig. 5 is the figure of the flow process of high-pass filtering processing.

Fig. 6 is the figure that amplifies the part of expression second image.

Fig. 7 is the figure of the part of expression photographic images.

Fig. 8 is the finish dealing with figure of a part of image of expression.

Fig. 9 is the figure of a part of the image of finishing dealing with of expression comparative example

Figure 10 is the figure of the flow process handled of the high-pass filtering of the uneven testing fixture of expression second embodiment.

Figure 11 is the figure that amplifies the part of expression second image.

Figure 12 is the finish dealing with figure of a part of image of expression.

Embodiment

Fig. 1 is the figure of structure of the uneven testing fixture 1 of expression first embodiment of the invention.Uneven testing fixture 1 is a device of checking the gradation unequal (following is called " inequality ") that is coated in the resist on the glass substrate (following is called " substrate ") 9, the figure that a plurality of panels of using with liquid crystal indicator at substrate 9 upper surfaces are used.

Uneven testing fixture 1 have the transmitance that keeps substrate 9 objective table 2, to be maintained at laminal substrate 9 on the objective table 2 as shown in Fig. 1 (Z) illumination part 3 of the lower surface irradiates light of the interarea of side, to the conduct of substrate 9 (+Z) shoot part 5 taken of the upper surface of the interarea of side and the computing machine 4 that carries out various calculation process.Computing machine 4 possesses the storage part 43 that the shooting control part 41, CPU and its peripheral circuit that are provided with as the control stand of the special use that is used to control shoot part 5 carry out storing various information such as the operational part 42 of various calculation process and storer, Fixed disk.Read special-purpose program 81 and be stored in the storage part 43 from the recording medium 8 of embodied on computer readable such as disk, CD, photomagneto disk via the illustrated reading device of omission that is arranged in the computing machine 4, carry out calculation process by CPU according to program 81, promptly pass through computing machine 4 executive routines 81, thereby computing machine 4 carries out inequality inspection.In addition, storage part 43 self also can be recording medium a kind of of logging program 81.

Fig. 2 is the planimetric map of expression objective table 2 and shoot part 5.Shoot part 5 possesses: head 52 has the linear sensor 51 of staggered a plurality of (being 10 in the present embodiment) CCD on the Y direction of this head 52 in Fig. 2; And head moving mechanism 53, this head moving mechanism 53 makes head 52 move along the directions X vertical with the Y direction, by utilizing shooting control part 41 (with reference to Fig. 1) control head moving mechanism 53, a plurality of linear sensor 51 whole moving, thus take and obtain image being maintained at substrate 9 on the objective table 2.In uneven testing fixture 1,10 linear sensors 51 that have 5000 capturing elements are respectively arranged on the distance of 2m, by each capturing element the zone in 40 μ m four directions on the substrate 9 are taken as 1 pixel.Substrate 9 is the squares in the four directions of about 2m, obtains the image of 50000 pixels * 50000 pixels by shoot part 5.

As shown in Figure 2, have 8 panel graphs of a correspondence using with liquid crystal indicator at substrate 9 upper surfaces, each figure has middle body (hereinafter referred to as " display part 91 ") that is used to show and the electrode part 92 that is configured to the electrode pattern formation usefulness of edge shape in its periphery.Because display part 91 is different with the light transmission of electrode part 92, thus in the image of substrate 9, with the corresponding zone of display part 91 and different mutually with the corresponding regional deep or light degree (hereinafter referred to as " concentration ") of the electrode part that is adjacent 92.Equally, with the corresponding zone of the periphery 93 in the outside of electrode part 92 and also different mutually with the corresponding regional concentration of electrode part 92.

As shown in Figure 1, operational part 42 has: image processing part 421, it is as carrying out the function that calculation process realize by CPU according to program 81, carries out the Flame Image Process of the image (hereinafter referred to as " photographic images ") that obtained by shoot part 5 is generated the image of finishing dealing with; And uneven detecting element 422, it asks for the value of the uneven degree of expression substrate 9 by calculation process based on finishing dealing with image.Image processing part 421 has: compression of images portion 4211, and it compresses photographic images and generates first image; Low-pass filtering portion 4212, it carries out low-pass filtering treatment to first image and generates second image; Edge detecting element 4213, its detect 2 different zones of concentration border (for example border of display part 91 and electrode part 92), be the edge; High-pass filtering portion 4214, it carries out the high-pass filtering processing to second image and generates the 3rd image; And contrast emphasizes portion 4215, and its contrast to the 3rd image emphasizes to generate the image of finishing dealing with.

Fig. 3 is the figure that the inequality on 1 pair of substrate 9 of the uneven testing fixture of expression detects the flow process of action.In uneven testing fixture 1, at first by illumination part 3 shown in Figure 1, to the lower surface irradiates light (step S11) that is maintained at the substrate 9 on the objective table 2.Then, the upper surface of 51 pairs of substrates 9 of a plurality of linear sensor that utilization is moved along the directions X among Fig. 1 by head moving mechanism 53 is taken, and the data of photographic images are sent to the operational part 42 of computing machine 4 and receive (step S12) by image processing part 421.Preferred photographic images obtains than (signal/noise ratio) with high S/N, as required, also can carry out correction for drift corresponding to the sensitivity of each capturing element of a plurality of linear sensors 51 to photographic images.

When photographic images is received by image processing part 421, compress by 4211 pairs of photographic images of compression of images portion and to generate first image (step S13).Here, (X, the pixel value of pixel Y) is expressed as F when will be positioned at coordinate in photographic images _XYThe time, with photographic images with s _aPixel * s _aThe scope of pixel is that unit compresses in first image that generates, and is positioned at coordinate (x, the pixel value A of concerned pixel y) _XyThrough type (1) is asked for.

$A_{\mathrm{xy}}=\underset{X=s_{a}x}{\overset{s_a(x+1)-1}{\mathrm{\Σ}}}\underset{Y=s_{a}y}{\overset{s_a(y+1)-1}{\mathrm{\Σ}}}{F}_{\mathrm{XY}}/{s_a}^2......\left(1\right)$

In the present embodiment, because s _aBe 4 (pixels), so the pixel count of first image is 12500 pixels * 12500 pixels, the S/N of first pixel is than bringing up to 4 times of photographic images.

When generating first image, the low-pass filtering treatment of being undertaken first image by low-pass filtering portion 4212, and from the influence that first image suppresses high frequency noise generate smoothing second image (step S14).The window of the operating range of decision low-pass filtering treatment is that length on one side is (2s ₁+ 1) square of pixel in second image, is positioned at coordinate (x, the pixel value L of concerned pixel y) _XyUse near the pixel value A (with reference to formula (1)) of first image of each pixel of concerned pixel, through type (2) is asked for.In the present embodiment, s ₁Be 1 (pixel).

$L_{\mathrm{xy}}=\underset{X=x-s_1}{\overset{x+{\mathrm{<figure-callout\; id="1"\; label="s"\; filenames="A20051010977000181.png,A20051010977000242.png"\; state="\{\{state\}\}">s</figure-callout>}}_1}{\mathrm{\&Sigma;}}}\underset{Y=y-s_1}{\overset{y+s_1}{\mathrm{\&Sigma;}}}{A}_{\mathrm{XY}}/{({2\mathrm{<figure-callout\; id="1"\; label="s"\; filenames="A20051010977000181.png,A20051010977000242.png"\; state="\{\{state\}\}">s</figure-callout>}}_1+1)}^2......\left(2\right)$

Fig. 4 is near the figure in a bight that amplifies the expression zone suitable with 1 panel of second image.In the following description, the zone in the image corresponding with display part 91 and electrode part 92 also only shows as " display part 91 " and " electrode part 92 ".In Fig. 4, indicate cross hatch and the pixel 901 represented is as the pixel on the edge 900 on the border of display part 91 and electrode part 92, hereinafter referred to as " edge pixel 901 ".The upper right side at the edge 900 in Fig. 4 indicates parallel oblique line and the pixel 911 represented is the pixels on the display part 91, hereinafter referred to as " display part pixel 911 ".In addition, indicate different parallel oblique lines and the pixel 921 represented is the pixels on the electrode part 92 across edge 900 and at the opposition side of display part 91, hereinafter referred to as " electrode part pixel 921 ".As shown in Figure 4, in the present embodiment, edge 900 has the width of 2 pixel portion.

When generating second image, the edge detecting element 4213 by shown in Figure 1 detects edge shown in Figure 4 900, and the coordinate of each edge pixel 901 is stored in (step S15) in the storage part 43.In addition, if the edge detected before high-pass filtering is handled, then also can carry out before low-pass filtering treatment or before the resolution compression.Then, carry out high-pass filtering by 4214 pairs second images of high-pass filtering portion and handle, generating to have removed from second image becomes the 3rd image (step S16) of concentration change that contrast described later is emphasized the low frequency of the obstacle handled.Be positioned at coordinate (x, the pixel value H of concerned pixel y) here, _Xy ¹Use near the pixel value L (with reference to formula (2)) of second image of each pixel of concerned pixel, ask for formula (3).

$H_{\mathrm{<figure-callout\; id="1"\; label="xy"\; filenames="A20051010977000181.png,A20051010977000242.png"\; state="\{\{state\}\}">xy</figure-callout>}}^1=L_{\mathrm{xy}}/(\underset{X=x-s_2}{\overset{x+s_2}{\mathrm{\&Sigma;}}}\underset{Y=y-s_2}{\overset{y+s_2}{\mathrm{\&Sigma;}}}{L}_{\mathrm{XY}}/{({2s}_2+1)}^2)......\left(3\right)$

The window (hereinafter referred to as " high pass window ") of the operating range that formula (3) expression is handled as the decision high-pass filtering and to use with concerned pixel be (2s as the length on each limit at center ₂+ 1) situation of the foursquare window of pixel, in the present embodiment, s ₂For variable, and be 12 (pixels) to the maximum.

Fig. 5 is the figure of the flow process handled of the high-pass filtering of expression step S16.In high-pass filtering portion 4214 (with reference to Fig. 1), at first, suitably select initial concerned pixel by computing value of asking for of high-pass filtering processing, from the coordinate of concerned pixel and the coordinate that is stored in the edge pixel 901 the storage part 43, ask for the distance of the line direction (horizontal direction among Fig. 4) between concerned pixel and each edge pixel 901 and the distance of column direction (vertical direction among Fig. 4), each edge pixel 901 is asked for a side's big in the distance of above line direction and column direction value (step S161).And the minimum value of this big side's relevant with all edge pixels 901 value is set up related as the minimum value and value d between concerned pixel and the edge 900 with concerned pixel.

In high-pass filtering portion 4214, confirm that whether concerned pixel is present near the edge 900 (step S162), when setting up related minimum value d more than or equal to 13 (pixels) with concerned pixel, is judged as fully away from edge 900, s ₂Be maximal value 12 (pixel) that the high pass window is set at the full-size (i.e. 25 pixel four directions) (step S163) of regulation.

In addition, when the minimum value d of concerned pixel less than 13 pixels, be judged as concerned pixel and be present near the edge 900 s ₂Be (d-1) that the high pass window dwindles (step S164).At this moment, the high pass window and the edge 900 that are dwindled join.If this is because other edge pixel 901 of hypothesis is present in the high pass window, because a big side's value must be littler than minimum value d in the distance of line direction between other the edge pixel 901 of concerned pixel and this and column direction, and then the high pass window dwindles.

Fig. 6 is identical with Fig. 4, is near the figure in bight that amplifies the expression zone suitable with 1 panel of second image, the minimum value d between each display part pixel 911 of the numeric representation in each display part pixel 911 and the edge 900.For example, at the get the bid display part image of drawings attached mark 911a of Fig. 6 is when paying close attention to image, because the value of paying close attention to a side big in the distance of line direction between image 911a and each edge pixel 901 and column direction is 2 pixels or more than it always, so minimum value d is 2 pixels, s ₂Be 1 (pixel).And the length on each limit of the high pass window 4214a that represents with thick line among Fig. 6 is that 3 pixels (are (2s ₂+ 1) pixel when using minimum value d to represent, is (2d-1) pixel).

Like this, in high-pass filtering portion 4214, the length of setting each limit is (2s ₂+ 1) the foursquare high pass window 4214a of pixel, as high pass window 4214a during fully away from edge 900, high pass window 4214a is fixed to the full-size of regulation, in the time of near being present in edge 900, high pass window 4214a dwindles, and avoids the affiliated zone of concerned pixel 911a, is the zone (promptly being edge 900 and electrode part 92 in Fig. 6) beyond the display part 91.

Consequently, whole pixel in the high pass window 4214a becomes the interior pixel in the affiliated zone of concerned pixel 911a, is display part pixel 911, under this state, pixel value based on the whole pixel in the high pass window 4214a carries out the high-pass filtering processing, asks for the pixel value H of the concerned pixel 911a of the 3rd image _Xy ¹(with reference to formula (3)) (step S165).And generate the 3rd image (step S166, S167) by concerned pixel is carried out the high-pass filtering processing successively when next pixel is switched.In addition, when concerned pixel 911a is on the edge pixel 900, does not carry out high-pass filtering and handle H _Xy ¹Be 1.

When the high-pass filtering processing finishes, emphasize by contrast that 4215 pairs the 3rd image degree of comparing of portion are emphasized to handle and generate the image of finishing dealing with that the data storage of the image of finishing dealing with is (Fig. 3: step S17) in storage part 43.In the image of finishing dealing with, be positioned at coordinate (x, the pixel value E of concern image y) _XyUse the pixel value H of the concerned pixel of the 3rd image _Xy ¹, contrast emphasizes to compare r _c, and background value b, ask for formula (4).In the present embodiment, r _cBe 0.01,0.02,0.05 or 0.1.

$E_{\mathrm{xy}}=b((H_{\mathrm{xy}}^1-1)/r_c+1)......\left(4\right)$

Fig. 7 is the figure of the part of expression photographic images.The region representation electrode part 92 in the left side at the edge 900 among Fig. 7, the region representation display part 91 on right side.In addition, the difference at interval that is marked on the parallel oblique line of electrode part 92 and display part 91 is represented the difference of the concentration in two zones, and expression is concentration ratio display part 91 height of narrow electrode part 92 at interval.

Fig. 8 is that expression is carried out above-mentioned Flame Image Process to photographic images shown in Figure 7 and the figure of the part of the image of finishing dealing with that generates.As shown in Figure 8, in the image of finishing dealing with, for example in photographic images, emphasize to confirm the bar shaped inequality 95 of difficult level.Fig. 9 is that expression will be near the edge 900 also fixedly under the situation of the size of high pass window 4214a, and the image of finishing dealing with when carrying out high-pass filtering and handle based on the whole pixel in the high pass window 4214a is used as the figure of comparative example.

In comparative example, because near the pixel the edge 900, based on more different 2 zones of concentration (promptly, display part 91 among Fig. 7 and electrode part 92) both sides' the pixel value of pixel carry out high-pass filtering and handle, so as shown in Figure 9, can confirm the bar shaped inequality 95 of level, but produce the satisfied zone of pixel value (promptly emphasize to handle and pixel value become calculating less than 0 or maximal value or the zone more than it) 94 near the edge attached 900 by contrast.Below, will can not detect uneven zone 94 by the saturated of pixel value and be called " inequality can not detect the zone ".Producing the reason that this inequality can not detect the zone is because using the pixel of 900 both sides, edge to carry out high-pass filtering handles, so emphasized the cause with the change of the pixel value of the directional correlation at crosscut edge 900.

On the other hand, at uneven testing fixture 1, because do not use the pixel value of pixel that is present in the zone of concerned pixel opposition side across edge 900 to carry out high-pass filtering handles by the high pass window 4214a that uses variable size, so as shown in Figure 8, compare with comparative example, can be at edge 900 (because in the image of finishing dealing with shown in Figure 8, can not be identified as clear and definite edge, so represent with double dot dash line) near the inequality of can dwindling can not detect 94 (zones that with dashed lines clips in Fig. 8), zone, near the inequalities that can improve the edge 900 are checked precision.

When contrast emphasizes that processing finishes, data by 422 pairs of images of finishing dealing with of uneven detecting element shown in Figure 1 are carried out calculation process, there are uneven local inequality and the unequal various inequalities of bar shaped that on line direction or column direction, produce the inequality of wire at the whole inequality that exists with point in inequality on the whole base plate 9, in a part of part on the substrate 9, uneven degree (so-called uneven intensity) is carried out quantification, calculate the uneven degree of expression value, be evaluation of estimate (step S18).Below, the acquiring method about the evaluation of estimate of various inequalities is described.

When asking for whole uneven evaluation of estimate, at first carry out FFT (high speed Fourier transform) by 422 pairs of uneven detecting elements foursquare each evaluation region in the image of finishing dealing with the big minispread of regulation, ask for frequency spectrum F as the formula (5) ^k _UvHere, k is that the evaluation region that expression becomes object is the small tenon of k evaluation region in the whole evaluation region of the image of finishing dealing with, (X _K0, y _K0) and (x _K1, y _K1) be the coordinate of pixel on the summit of k evaluation region of regulation.In addition, s _fAs the formula (6), be the length on one side of evaluation region, in the present embodiment, s _fBe 64 (pixels) or 128 (pixels).

$F_{\mathrm{uv}}^k=|\frac{1}{s_f}\underset{X=x_{k0}}{\overset{x_{\mathrm{<figure-callout\; id="1"\; label="k"\; filenames="A20051010977000181.png,A20051010977000242.png"\; state="\{\{state\}\}">k</figure-callout>}1}}{\mathrm{\&Sigma;}}}\underset{{Y=y}_{k0}}{\overset{y_{k1}}{\mathrm{\&Sigma;}}}{E}_{\mathrm{XY}}\&CenterDot;e^{-i\&CenterDot;2(\mathrm{uX}+\mathrm{vY})/s_f}|......\left(5\right)$

s _f＝x _k1-x _k0+1＝y _k1-y _k0+1 ……(6)

And at frequency spectrum F ^k _UvIn, the power of the frequency band of the regulation of low frequency carries out integrating as the formula (7), asks for the whole uneven evaluation of estimate P about k evaluation region ^k _fHere, λ ₁And λ ₂It is respectively the upper and lower bound of the frequency of integrating scope.By above processing, in uneven detecting element 422, each evaluation region of the image of finishing dealing with is asked for evaluation of estimate P successively ^k _f

$P_f^k=\underset{u={\mathrm{\&lambda;}}_1}{\overset{{\mathrm{\&lambda;}}_2}{\mathrm{\&Sigma;}}}\underset{v={\mathrm{\&lambda;}}_1}{\overset{{\mathrm{\&lambda;}}_2}{\mathrm{\&Sigma;}}}{F}_{\mathrm{uv}}^k......\left(7\right)$

In addition, Bu Jun evaluation of estimate as a whole also can be used the standard deviation of pixel value of each evaluation region of the image of finishing dealing with.At this moment, the evaluation of estimate P of k evaluation region ^k _dUtilize formula (8) to ask for by uneven detecting element 422.Here, the size of evaluation region and be shaped as identical with the situation of formula (5) and formula (6), at formula (8) same operation parameter k, (x _K0, y _K0) and (x _K1, y _K1), but the shape of evaluation region is not limited to square, can wait other shape (also suitable these parameters of using in the explanation of the evaluation of estimate of below other) for rectangle or circle.n _kThe pixel count of representing k evaluation region as shown in Figure 9.

$P_d^k=\sqrt{\frac{n_k\underset{X=x_{k0}}{\overset{{\mathrm{<figure-callout\; id="1"\; label="x\; k"\; filenames="A20051010977000181.png,A20051010977000242.png"\; state="\{\{state\}\}">x</figure-callout>}}_k}{\mathrm{\&Sigma;}}}\underset{Y=y_{k0}}{\overset{y_{k1}}{\mathrm{\&Sigma;}}}{E_{\mathrm{XY}}}^2-{\left(\underset{X=x_{k0}}{\overset{{\mathrm{<figure-callout\; id="1"\; label="x\; k"\; filenames="A20051010977000181.png,A20051010977000242.png"\; state="\{\{state\}\}">x</figure-callout>}}_k}{\mathrm{\&Sigma;}}}\underset{Y=y_{k0}}{\overset{y_{k1}}{\mathrm{\&Sigma;}}}{E}_{\mathrm{XY}}\right)}^2}{n_k(n_k-1)}}......\left(8\right)$

n _k＝(x _k1-x _k0+1)(y _k1-y _k0+1) ……(9)

About local uneven, for example by uneven detecting element 422, the finish dealing with computing shown in mean value (hereinafter referred to as " both full-pixel mean value ") the AVE through type (10) of pixel value of both full-pixel of image and asking for, according to from the threshold value of only big (perhaps little) setting of this mean value with both full-pixel 2 values after, the number that calculates the size of regulation or its above join domain (zone that promptly connects a plurality of pixels with identical value) is as local uneven evaluation of estimate.Here, be the finish dealing with line direction of image and the pixel count of column direction (wid+1) and (hei+1), in the present embodiment, as mentioned above, both sides are 12500 (pixels).

$\mathrm{AVE}=\underset{X=0}{\overset{\mathrm{wid}}{\mathrm{\&Sigma;}}}\underset{Y=0}{\overset{\mathrm{hei}}{\mathrm{\&Sigma;}}}{E}_{\mathrm{XY}}/(\mathrm{wid}+1)(\mathrm{hei}+1)......\left(10\right)$

In addition, as local uneven evaluation of estimate, each evaluation region of the square (perhaps rectangle) that the evaluation region when the whole uneven evaluation of estimate of contrast is calculated is little is asked for the mean value of pixel value, also can ask for the standard deviation of the mean value of evaluation region.Perhaps the quantity at the edge in each evaluation region can be made as local uneven evaluation of estimate.

When asking for the evaluation of estimate of bar shaped inequality, at first, carry out low-pass filtering treatment and generate the 4th image by 422 pairs of the uneven detecting elements image of finishing dealing with.The window of low-pass filtering is that an edge lengths is (2s ₃+ 1) square of pixel in the 4th image, is positioned at coordinate (x, the pixel value L of concerned pixel y) ¹ _XyUse near the pixel value E (with reference to formula (4)) of the image of finishing dealing with of each pixel of concerned pixel, through type (11) is asked for.In the present embodiment, s ₃Be 1 (pixel).

${\mathrm{<figure-callout\; id="1"\; label="L\; xy"\; filenames="A20051010977000181.png,A20051010977000242.png"\; state="\{\{state\}\}">L</figure-callout>}}_{\mathrm{xy}}^{}=\underset{X=x-s_3}{\overset{x+s_3}{\mathrm{\&Sigma;}}}\underset{Y=y-s_3}{\overset{y+s_3}{\mathrm{\&Sigma;}}}{E}_{\mathrm{XY}}/{({2s}_3+1)}^2......\left(11\right)$

And, by uneven detecting element 422, to the foursquare evaluation region of calculating identical regulation with whole uneven evaluation of estimate, as the formula (12), calculate respectively in the projection value of the line direction that each row is asked for maximal value and as the evaluation of estimate P of the bar shaped inequality (so-called level inequality) of line direction ^k _H1Here, the small tenon Y of which row of expression evaluation region gets y _K0～y _K1Value.In addition, b and formula (4) are similarly background value.

$P_{h1}^k=\max [\underset{X=x_{k0}}{\overset{x_{k1}}{\mathrm{\&Sigma;}}}{\mathrm{<figure-callout\; id="1"\; label="L\; XY"\; filenames="A20051010977000181.png,A20051010977000242.png"\; state="\{\{state\}\}">L</figure-callout>}}_{\mathrm{XY}}^{}/b],(Y=y_{k0}~y_{k1})......\left(12\right)$

Equally, as the formula (13), calculate respectively in the projection value of the column direction that each row of evaluation region are asked for maximal value and as the evaluation of estimate P of the bar shaped inequality (so-called vertical inequality) of column direction ^k _V1Here, the small tenon X of which row of expression evaluation region gets x _K0～x _K1Value.

$P_{v1}^k=\max [\underset{Y=y_{k0}}{\overset{y_{k1}}{\mathrm{\&Sigma;}}}{\mathrm{<figure-callout\; id="1"\; label="L\; XY"\; filenames="A20051010977000181.png,A20051010977000242.png"\; state="\{\{state\}\}">L</figure-callout>}}_{\mathrm{XY}}^{}/b],(X=x_{k0}~x_{k1})......\left(13\right)$

In addition, as other evaluation of estimate of bar shaped inequality, also can ask for the uneven parallel direction of the bar shaped that should detect on the mean value of the pixel value of the evaluation region of long rectangle, the standard deviation of the mean value in in-service evaluation zone.

In uneven detecting element 422, to compare to the evaluation of estimate of the substrate 9 of the various inequalities that calculate as mentioned above with the various not related threshold values that are stored in advance in the storage part 43, when any one evaluation of estimate surpasses corresponding threshold, detect the inequality (step S19) of the kind corresponding with this evaluation of estimate.

As described above, in uneven testing fixture 1, during the pixel value of the concerned pixel of low-pass filtering portion 4214 in asking for the 3rd image (promptly being set as the pixel that generates the operand when finishing dealing with image), the pixel of pixel groups substantial limitation by will in second image shown in Figure 6, being used in computing in high pass window 4214a, promptly as in the display part pixel 911 in the display part 91 in the zone under the concerned pixel 911a, thereby can improve the precision that near the inequalities the edge 900 are checked.

In addition, reduced by near high pass window 4214a edge 900, avoid concerned pixel 911a affiliated area zone in addition, thereby near edge 900, because concerned pixel 911a is positioned at the almost center that is used in a plurality of pixels that high-pass filtering handles, so also can make characteristic that near the inequalities the edge 900 check and fully check that away from the inequality in the zone at edge 900 characteristic is approximate.And then, by making high pass window 4214a be square, and 2 times only little 1 pixels of the minimum value of a big side's value in the distance of the length that makes each limit than the line direction of 901 of concerned pixel 911a and edge pixels and column direction, can avoid the zone beyond the zone under the concerned pixel 911a easily and set high pass window 4214a.

And then, in uneven testing fixture 1,, can significantly suppress the influence at edge 900, and automatically obtain the evaluation of estimate of the uneven degree of expression substrate 9 by the calculation process of uneven detecting element 422.

Below, describe at the uneven testing fixture of second embodiment of the invention.The apparatus structure of the uneven testing fixture of second embodiment is identical with uneven testing fixture 1 shown in Figure 1, below each is constituted major part indicates identical Reference numeral.In addition, the inequality of second embodiment detects action except the method difference that the high-pass filtering of high-pass filtering portion 4214 is handled, and is identical with first embodiment.

Testing fixture by second embodiment carries out inequality when checking, at first, (Fig. 3: step S11), the data of the photographic images that is obtained by shoot part 5 receive (step S12) by image processing part 421 to substrate 9 irradiates lights by illumination part shown in Figure 13.Then, compress by 4211 pairs of photographic images of compression of images portion and to generate first image (step S13), carry out low-pass filtering treatment by low-pass filtering portion 4212 and generate second image (step S14).Then, detect edge 900, after the coordinate of each edge pixel 901 is stored in storage part 43, undertaken generating the 3rd image (step S15, S16) with the high-pass filtering processing of second image by high-pass filtering portion 4214 by edge detecting element 4213.These processing are identical with the situation of first embodiment.

Figure 10 is the figure of the flow process handled of the high-pass filtering of the uneven testing fixture of expression second embodiment.Figure 11 is identical with Fig. 4, is near the figure in bight that amplifies 1 display part 91 of expression second image.In high-pass filtering portion 4214, at first in order to determine the scope of high-pass filtering processing, setting with concerned pixel 911a is the foursquare high pass window 4214a (representing with thick dashed line in the drawings) (step S161a) at center.The size of high pass window 4214a is fixed, and the maximum sized high pass window 4214a of the length on each limit and first embodiment is similarly 25 pixels, and (length that is about to each limit is made as (2s ₄+ 1) during pixel, s4 is 12 (pixels)), but in Figure 11, conveniently describe as 7 pixels in order to illustrate.

When setting high pass window 4214a, coordinate based on concerned pixel 911a and edge pixel 901, in the pixel in high pass window 4214a, as the zone under the concerned pixel 911a, be the pixel (being display part pixel 911) in the display part 91, zone (zone of representing with heavy line among Figure 11 is hereinafter referred to as " the join domain ") 4214b that the specific pixel groups that is connected with concerned pixel 911a by not hindered by edge 900 forms.And by high-pass filtering portion 4214, as the formula (14), only the pixel value L (with reference to formula (2)) based on the pixel in the join domain 4214b of second image carries out the high-pass filtering processing, asks for the pixel value H of the concerned pixel 911a of the 3rd image ² _Xy(step S162a).Here, m _XyIt is the pixel count in the join domain 4214b.

$H_{\mathrm{xy}}^2=L_{\mathrm{xy}}/(\underset{X=x-s_4}{\overset{x+{\mathrm{<figure-callout\; id="4"\; label="s"\; filenames="A20051010977000181.png"\; state="\{\{state\}\}">s</figure-callout>}}_4}{\mathrm{\&Sigma;}}}\underset{Y=y-s_4}{\overset{y+s_4}{\mathrm{\&Sigma;}}}\left(\begin{array}{ccc}{L}_{\mathrm{XY}}:\mathrm{if}& \mathrm{connected}& L_{\mathrm{XY}}\\ 0:\mathrm{else}& & \end{array}\right)/m_{\mathrm{xy}})......\left(14\right)$

And, each pixel of second pixel is carried out the high-pass filtering processing successively generates the 3rd image (step S163a, S164a).In addition, when concern image 911a is on the edge 900, does not just carry out high-pass filtering and handle H ² _XyBe 1.

When the high-pass filtering processing finishes, emphasize by contrast that 4215 pairs the 3rd image degree of comparing of portion are emphasized to handle and generate the image of finishing dealing with that the data storage of the image of finishing dealing with is at storage part 43 (Fig. 3: step S17).Then, carry out calculation process, as stating, calculate the evaluation of estimate (step S18) of the whole inequality of substrate 9, local inequality and bar shaped inequality by the data of 422 pairs of images of finishing dealing with of uneven detecting element.And, to compare to the evaluation of estimate of the substrate 9 of various inequalities with the various not related threshold values that are kept at storage part 43 in advance, when any one evaluation of estimate surpasses corresponding threshold, detect the inequality (step S19) of the kind corresponding with this evaluation of estimate.

Figure 12 is the part that the uneven testing fixture of expression by second embodiment carries out the image of finishing dealing with that Flame Image Process generates.As shown in figure 12, in the uneven testing fixture of second embodiment, the inequality relevant with bar shaped uneven 95 can not be able to be detected regional 94 (zones that with dashed lines clips in Figure 12) and narrow down to the zone (when the being Figure 11 example zone of 2 pixel wide) corresponding with edge 900.

As described above, in the uneven testing fixture of second embodiment, because to the concerned pixel 911a adjacent, also carry out high-pass filtering and handle, so can improve the precision that near the inequalities the edge 900 are checked more based near a plurality of pixels the concerned pixel 911a with edge 900.In addition, by the fixing size of high pass window 4214a, high-pass filtering is handled oversimplified.

In the uneven testing fixture of second embodiment, identical with first embodiment, during the pixel value of the concerned pixel of high-pass filtering portion 4214 in asking for the 3rd image, by the image sets that will in second image shown in Figure 11, be used in computing be limited in pixel in the join domain 4214b, promptly in fact as the pixel in the display part 91 in the zone under the concerned pixel 911a, can improve the precision that near the inequalities the edge 900 are checked.In addition, by the calculation process of uneven detecting element 422, the evaluation of estimate of the uneven degree of expression substrate 9 is also obtained in the influence that can significantly suppress edge 900 automatically.

More than, be illustrated at embodiments of the present invention, but the present invention is defined in above-mentioned embodiment, can carry out various changes.

For example, in shoot part 5, can replace linear sensor and be provided with 2 the dimension ccd sensors be used as shoot part.In addition, in the above-described embodiment, the irradiation of the light that is undertaken by illumination part 3 is across substrate 9 and carry out from the opposition side of shoot part 5, the light of transmission substrate 9 incides image pickup part 5, also can be that the irradiation of light is carried out and incided the structure of image pickup part 5 at the light of substrate 9 reflection from the same side of shoot part 5.In addition, when there is enough bright other light sources (for example carrying out the common illumination at the uneven scene of checking) in the shooting to substrate 9, can omit illumination part 3.

Pay close attention to image 911a if the almost center of high pass window 4214a, the pixel count on each limit of high pass window 4214a also can be an even number.In addition, high pass window 4214a may not necessarily be defined in square, also can be the shape that rectangle or circle wait other, but can simply carry out being preferably rectangle on the setting this point of high pass window 4214a.

In the 4214a of high-pass filtering portion, even greatly also can keep this angle of processing speed consistently from high pass window 4214a is become, handle as high-pass filtering, preferably look like the full average treatment shown in the formula (3), handle but also can carry out other various processing, the processing of for example weighted mean or medium filtering.In addition, second image carried out Fourier transform etc. and transfer to the frequency space, eliminate the power of low-frequency component after, change view data into by contravariant once more and generate the 3rd image.

In the graphics process of image processing part 421, before high-pass filtering is handled, carry out compression processing and low-pass filtering treatment to photographic images, but the S/N of photographic images when high or the influence of high frequency noise hour, also can omit these processing.On the contrary, the processing that also can between each step, suitably insert other.

The uneven testing fixture of above-mentioned embodiment for example can be used in various uneven inspection of the inequality inspection of light transmission as the shadow mask that color Braun tube is used, that arrange the open-work plates that form a plurality of open-works etc., various other objects.

More than the present invention is carried out detailed describing and illustrating, but the explanation stated is illustration rather than qualification.Therefore, only be appreciated that otherwise just depart from the scope of the present invention a plurality of distortion and form can be arranged.

Claims (14)

1. a uneven testing fixture is a uneven testing fixture of checking the gradation unequal of object, it is characterized in that, comprising:

Maintaining part, it keeps object;

Shoot part, its interarea to above-mentioned object is taken;

Image processing part, it carries out Flame Image Process to the object images of being obtained by above-mentioned shoot part and generates the image of finishing dealing with,

Above-mentioned image processing part has:

The edge detecting element, its detect in above-mentioned object images the zone of different mutually concentration the border, be the edge;

High-pass filtering portion, it carries out high-pass filtering to above-mentioned object images and handles;

Contrast is emphasized portion, and its contrast to the above-mentioned object images after being handled by above-mentioned high-pass filtering portion emphasizes,

During the pixel value of the concerned pixel of above-mentioned high-pass filtering portion in asking for the above-mentioned image of finishing dealing with, the pixel groups that will be used in computing in above-mentioned object images is restricted to the pixel in the zone under the above-mentioned concern image in fact.

2. uneven testing fixture as claimed in claim 1, it is characterized in that, the high-pass filtering of above-mentioned high-pass filtering portion is handled and is based on above-mentioned concerned pixel is that the pixel value of the whole pixel in the window of approximate centre carries out, in the time of near above-mentioned window is present in above-mentioned edge, above-mentioned window is reduced, thereby above-mentioned window is avoided zone in addition, the affiliated zone of above-mentioned concerned pixel.

3. uneven testing fixture as claimed in claim 2 is characterized in that above-mentioned window is a rectangle.

4. uneven testing fixture as claimed in claim 3, it is characterized in that, above-mentioned window is a square, above-mentioned high-pass filtering portion asks for a side's big in the distance of line direction between each pixel in above-mentioned concerned pixel and the above-mentioned edge and column direction value, will be than the value of 2 times of only little 1 pixels of the minimum value of an above-mentioned big side's relevant with both full-pixel in the above-mentioned edge the value length as each limit of above-mentioned window.

5. uneven testing fixture as claimed in claim 1, it is characterized in that the high-pass filtering of above-mentioned high-pass filtering portion is only handled based on the pixel value that with above-mentioned concerned pixel is the pixel in zone in the pixel in the window of fixing size of approximate centre, under the above-mentioned concerned pixel and carried out.

6. uneven testing fixture as claimed in claim 5 is characterized in that above-mentioned window is a rectangle.

7. uneven testing fixture as claimed in claim 1 is characterized in that, also has based on above-mentioned to finish dealing with image and ask for the uneven detecting element of value of degree of the gradation unequal of the above-mentioned object of expression by calculation process.

8. uneven inspection method is based on the irregular inspection method that the object images that obtains from object is checked the gradation unequal of above-mentioned object, it is characterized in that, comprising:

Detect in object images the zone of different mutually concentration the border, be the operation at edge;

Above-mentioned object images is carried out the operation that high-pass filtering is handled;

The contrast of carrying out the above-mentioned object images after above-mentioned high-pass filtering is handled is emphasized and generated the operation of the image of finishing dealing with;

Detect the operation of the gradation unequal of above-mentioned object based on the above-mentioned image of finishing dealing with,

Carry out in the operation that high-pass filtering handles above-mentioned, during the pixel value of a concerned pixel in asking for the above-mentioned image of finishing dealing with, the pixel groups that is used in computing in above-mentioned object images is restricted to the pixel in the zone under the above-mentioned concerned pixel in fact.

9. uneven inspection method as claimed in claim 8, it is characterized in that, above-mentioned high-pass filtering is handled and is based on above-mentioned concerned pixel is that the pixel value of the whole pixel in the window of approximate centre carries out, in the time of near above-mentioned window is present in above-mentioned edge, above-mentioned window is reduced, thereby above-mentioned window is avoided zone in addition, the affiliated zone of above-mentioned concerned pixel.

10. uneven inspection method as claimed in claim 9 is characterized in that above-mentioned window is a rectangle.

11. uneven inspection method as claimed in claim 10, it is characterized in that, above-mentioned window is a square, carry out in the operation that high-pass filtering handles above-mentioned, ask for a side's big in the distance of line direction between each pixel in above-mentioned concerned pixel and the above-mentioned edge and column direction value, will be than the value of 2 times of only little 1 pixels of the minimum value of an above-mentioned big side's relevant value length as each limit of above-mentioned window with both full-pixel in the above-mentioned edge.

12. uneven inspection method as claimed in claim 8, it is characterized in that above-mentioned high-pass filtering is only handled based on the pixel value that with above-mentioned concerned pixel is the pixel in zone in the pixel in the window of fixing size of approximate centre, under the above-mentioned concerned pixel and carried out.

13. uneven inspection method as claimed in claim 12 is characterized in that above-mentioned window is a rectangle.

14. recording medium, be that record makes computing machine check the recording medium of program of the gradation unequal of above-mentioned object based on the object images that obtains from object, it is characterized in that the execution of being undertaken by the aforementioned calculation machine of said procedure is that the aforementioned calculation machine is carried out:

Detect in object images the zone of different mutually concentration the border, be the operation at edge;

Above-mentioned object images is carried out the operation that high-pass filtering is handled;

The contrast of carrying out the above-mentioned object images after above-mentioned high-pass filtering is handled is emphasized and generated the operation of the image of finishing dealing with;

Detect the operation of the gradation unequal of above-mentioned object based on the above-mentioned image of finishing dealing with,

Carry out in the operation that high-pass filtering handles above-mentioned, during the pixel value of a concerned pixel in asking for the above-mentioned image of finishing dealing with, the pixel groups that is used in computing in above-mentioned object images is restricted to the pixel in the zone under the above-mentioned concerned pixel in fact.

Images