CN101210890A - Defect detecting device and method, image sensor device and module - Google Patents

Abstract

The present invention discloses a defect detecting device and method, image sensor device and module. The defect detecting device includes: pixel value correcting part, for correcting pixel value for detecting image of object in a defect area, so as to emphasize the defect area of the detecting image relative to other area of detecting image; a block splitting processing part, for splitting the detecting image with corrected pixel value into a plurality of blocks and calculating block addition value or block average, the addition value is obtained by adding up pixel values of pixels in each block, the average is obtained through the addition value being divided by the number of the pixels in each block; a good/bad judgement part, for judging whether existing outlier of the addition value or the block average by means of the statistical treatment of a statistical processing part, so as to judge whether existing the defect area. By means of emphasizing the defect area of the image of the image sensor, it is able to implement defect detecting device and defect detecting method with high precision for detecting defect area.

Description

Defect detecting device and method, image sensor devices and module

Technical field

The present invention relates to a kind ofly in the detection step of making image sensor devices (Image Sensor Device), detect the recording medium that defect detecting device, defect inspection method, image sensor devices, image sensor module, defects detection program and the computing machine of defective can read accurately.

Background technology

When carrying out the digital picture attribute test, whether normal important to existing defect area to judge between right and wrong, described defect area is meant that pixel value is the zone of inhomogeneous variation.Especially recent years, the portable phones of static type digital camera (Digital Still Camera), built-in camera etc. constantly enlarge the demand of image sensor devices, therewith correspondingly, will be realistic high-qualityization of existing image sensor devices, require in testing procedure the defect area that produced in the digital picture (imageing sensor image) that obtains by image sensor devices, be that spot defective, irregular defective and line defect are implemented to detect.

That is to say that even image sensor devices is certified products, because pixel value with respect to pixel coordinate shade (shading) composition of slowly variation and the influence of noise contribution takes place, it is certain that pixel value can not keep.Particularly, when having spot defective, irregular defective or line defect etc. in the imageing sensor image, pixel value will because of each defective along with complicated variation takes place in pixel coordinate, so, be difficult to detect these defectives, thereby need a kind of picture quality detection method of highly sensitive detection defective.

At this, " point defect " is meant the defective of following state, that is: in the imageing sensor image, one of them pixel value is compared with the pixel value around it and had obvious and bigger the difference and become relatively more outstanding value.In addition, " spot defective " is meant the defective of following state, that is: the difference of the arbitrary pixel value in a plurality of pixel values in certain zone and on every side pixel value when " point defect " difference with on every side pixel value little." irregular defective " is meant the defective of following state, that is: the difference between the pixel value is also littler than the difference between the pixel value of " spot defective ", and the zone at a plurality of pixels place is bigger than " spot defective "." line defect " is meant the defective of following state, that is: column direction, line direction or the pixel value arranged on the tilted direction at any angle compare to pixel value around it in the imageing sensor image, exists significantly poor and becomes relatively more outstanding value.

In addition, " shade " is meant following state, that is: pixel value takes place to change slowly with respect to pixel coordinate, and pixel value diminishes gradually towards upper part, end portion, left end portion and the right end portion of imageing sensor image.The reason that shade takes place is owing to the pixel with respect to image central authorities, and the sensitivity of image end pixel reduces.

Past, by the testing staff with the visual detection of finishing spot defective, irregular defective and the line defect of the flat-panel monitor of imageing sensor image, liquid crystal display etc.But, because this detection relies on testing staff's subjective judgement, so, there is following problems, that is: because the influence of the deviation of testing staff's examination criteria and the testing staff health when detecting will cause the testing result instability.In addition, also there is the problem that is difficult to realize each defective is carried out quantification.To this, developed in recent years in making the manufacturing step of image sensor devices etc. defective quantification and the pick-up unit that detects, and begun to advance the robotization that utilizes this pick-up unit to carry out to detect.Usually, this pick-up unit photographic images and this image is implemented Flame Image Process detect each defective.

For example, (the patented claim Publication Laid-Open 2004-294202 of Japan communique discloses day to patent documentation 1: a kind of like this defect inspection method below on October 21st, 2004) having disclosed.Promptly, at first according to the kind of defective, to generate the downscaled images that a plurality of sizes are dwindled by the image value of the taken image of imaging apparatus (detected image), perhaps after implementing to emphasize the Filtering Processing of defective, each bright, the dark defective in the detected image is carried out the statistical treatment of monochrome information.Then, based on this statistics, decision detects the defective candidate after being used to detect the threshold value of defective candidate, and calculates out the quantitative evaluation value of the defective candidate that is detected.Thus, can judge whether there is the defective candidate in the detected image according to the kind of defective.

Specifically, carry out the spot defects detection of this defect inspection method as described below.That is, at first utilize smoothing filter to handle, perhaps carry out form and handle (smoothing processing).Then, dwindle by the resulting image of smoothing processing and be made into the downscaled images of various sizes.Afterwards, utilizing spatial filter is the Top-Hat wave filter implements to emphasize the spot defective to each downscaled images contrast.Afterwards, lack of proper care to handle and make that dim spot is emphasized.Then, carry out statistical computation according to the brightness value of each pixel in the detected image, and according to coming setting threshold from the resulting statistics of this statistical computation.Afterwards, judge in detected image, whether there is the defective candidate according to the luminance threshold that sets.

But, in patent documentation 1, to spot defective, strip flaw, irregular defective and line defect any one detect when handling, though implement smoothing processing in order to remove noise contribution, but it is only, still bigger by the influence that noise contribution brought by smoothing processing.Therefore, the defect inspection method that utilizes patent documentation 1 to be put down in writing then might be judged there being the zero defect candidate to make mistakenly.

For example, as shown in figure 24, when implementing patent documentation 1 described statistical treatment for the image that has carried out rim detection (with aftermentioned), learn: the defect area A (with reference to Figure 24) that will be detected is as shown in figure 25 covered by noise contribution and is difficult to be detected.Therefore,, will there be such possibility, that is, should makes the judgement of defectiveness candidate, but make the judgement of zero defect candidate mistakenly if utilize the defect inspection method shown in the patent documentation 1.

In addition, in patent documentation 1, use the luminance picture that synthesizes by RGB as detected image.Therefore, the defect inspection method that utilizes patent documentation 1 to be disclosed, the image sensor devices that then might produce defective to the combination owing to RG or RB is made the false judgment that the defective candidate has or not.

For example, image shown in Figure 26 is that luminance picture has been implemented the image after the rim detection.As shown in figure 27, when being implemented statistical treatment by the image after the rim detection, learn: detection sensitivity reduces for the line defect that will be detected, and this should but not emphasized as the part that line defect is detected.Therefore, in this case, utilize the defect inspection method of patent documentation 1 then to be difficult to detect the line defect that should be detected.

Summary of the invention

The present invention develops in view of the above problems, its purpose is to provide a kind of and emphasizes to handle by the defect area enforcement to the imageing sensor image, thereby can detect defect detecting device, defect inspection method, image sensor devices, the image sensor module of defective accurately.

In order to solve above-mentioned problem, defect detecting device of the present invention is a kind of defect detecting device that detects defect area in the digital picture, and wherein, the pixel value that the pixel value of this defect area compares to the zone around this defect area is uneven variation.Defect detecting device of the present invention, have: the pixel value correction unit, proofread and correct the pixel value in the detected image, make that the pixel value in the zone that should detect as above-mentioned defect area is emphasized with respect to the pixel value in the zone outside the above-mentioned defect area, wherein, above-mentioned detected image is the image that will carry out defective area detection, the piece cutting treatmenting part, above-mentioned detected image segmentation is become a plurality of and obtain the piece additive value or piece mean value, wherein, above-mentioned additive value is that the pixel value to the pixel that exists in each piece carries out resulting value after the addition, and above-mentioned mean value is that above-mentioned additive value is divided by the resulting value of the number of the pixel that exists in this piece; Have or not above-mentioned additive value of judging part output or piece mean value to defect area, wherein, this defect area has or not judging part to judge whether to exist the outlier of above-mentioned additive value or piece mean value by statistical treatment, thereby judges whether to exist above-mentioned defect area.

According to said structure, in detected image, exist under the situation of defect area, the pixel value correction unit is proofreaied and correct the pixel value of detected image, makes that the pixel value of this defect area is emphasized.In addition, the piece cutting treatmenting part is a plurality of with detected image segmentation and obtains the piece additive value or piece mean value.That is, in defect detecting device, pixel value correction unit and piece cutting treatmenting part are handled detected image respectively.Piece additive value of trying to achieve at the piece cutting treatmenting part or piece mean value are output to defect area and have or not judging part.Then, defect area has or not judging part to carry out statistical treatment according to this piece additive value or piece mean value, detected image is had the judgement of area free from defect.

Thus, owing to can distinguish defect area and the noise contribution that is present in the detected image really, so can detect defect area accurately.In addition, can suppress false judgment to there being area free from defect to do.

In addition, piece additive value or piece mean value will be output to defect area and have or not judging part, and obtain the timing of the processing of this piece additive value or piece mean value for the piece cutting treatmenting part, and can be before the pixel value correction unit be handled, also can be after the pixel value correction unit be handled.Promptly, can be before the pixel value correction unit be implemented to proofread and correct to the pixel value of detected image, the piece cutting treatmenting part is obtained the piece additive value or the piece mean value of each piece of detected image earlier, also can after implementing to proofread and correct to the pixel value of detected image, the pixel value correction unit obtain piece additive value or piece mean value.

In order to solve above-mentioned problem, defect detecting device of the present invention is a kind of defect detecting device that detects defect area in the digital picture, and wherein, the pixel value that the pixel value of this defect area compares to the zone around this defect area is uneven variation.Defect detecting device of the present invention, has the piece cutting treatmenting part, detected image segmentation is become in advance block-shaped according to the shape set of above-mentioned defect area, and obtain piece additive value or piece mean value, wherein, above-mentioned detected image is the image that will carry out defective area detection, and above-mentioned additive value is that the pixel value to the pixel that exists in each piece carries out resulting value after the addition, and above-mentioned mean value is resulting value after the number of above-mentioned additive value divided by the pixel that exists in this piece; Have or not above-mentioned additive value of judging part output or piece mean value to defect area, wherein, this defect area has or not judging part to judge whether to exist the outlier of above-mentioned additive value or piece mean value by statistical treatment, thereby judges whether to exist above-mentioned defect area.

According to said structure, in the piece cutting treatmenting part, detected image segmentation is become in advance block-shaped according to the shape set of above-mentioned defect area, and obtain the piece additive value or the piece mean value of each piece.By have or not judging part to export this piece additive value or piece mean value to defect area, defect area has or not judging part to carry out statistical treatment according to this piece additive value or piece mean value, detected image is had the judgement of area free from defect.

Thus, corresponding with defect area various block-shaped because the piece cutting treatmenting part utilizes, carry out the detection of each defect area separately, so can carry out specific detection to each defect area accurately.In addition, because defect area has or not judging part to utilize the piece additive value or the piece mean value of the resulting detected image of piece cutting treatmenting part enforcement processing to handle, can suppress false judgment to there being area free from defect to do.

In order to solve above-mentioned problem, defect detecting device of the present invention is a kind of defect detecting device that detects defect area in the digital picture, and wherein, the pixel value that the pixel value of this defect area compares to the zone around this defect area is uneven variation.Defect detecting device of the present invention, have: image production part, generate color difference image according to detected image, wherein, above-mentioned detected image is the image and the piece cutting treatmenting part that will carry out defective area detection, the color difference image that above-mentioned image production part is generated is divided into a plurality of and obtain the piece additive value or piece mean value, wherein, above-mentioned additive value is that the pixel value to the pixel that exists in each piece carries out resulting value after the addition, and above-mentioned mean value is resulting value after the number of above-mentioned additive value divided by the pixel that exists in this piece; Have or not above-mentioned additive value of judging part output or piece mean value to defect area, wherein, this defect area has or not judging part to judge whether to exist the outlier of above-mentioned additive value or piece mean value by statistical treatment, thereby judges whether to exist above-mentioned defect area.

According to said structure, image production part generates the color difference image as detected image.The piece cutting treatmenting part is divided into a plurality of with color difference image, and obtains the piece additive value or the piece mean value of each piece.Then, this piece additive value or piece mean value are output to defect area and have or not judging part, and defect area has or not judging part to carry out statistical treatment according to this piece additive value or piece mean value, and detected image is had the judgement of area free from defect.

Thus, can detect in the detected image owing to look changes the defective produce.In addition, changing owing to look under the situation that the defective cause is present in the vertical direction of detected image or a certain row on the horizontal direction, can detect ordinate defective or horizontal line defective accurately.And then defect area has or not the judging part utilization to handle at the piece additive value or the piece mean value of the handled color difference image of piece cutting treatmenting part, can suppress the false judgment to there being area free from defect to do.

In order to solve above-mentioned problem, defect inspection method of the present invention is a kind of defect inspection method that detects defect area in the digital picture, and wherein, the pixel value that the pixel value of this defect area compares to the zone around this defect area is uneven variation.Defect inspection method of the present invention, comprise: the pixel value aligning step, proofread and correct the pixel value in the detected image, make that the pixel value in the zone that should detect as above-mentioned defect area is emphasized with respect to the pixel value in the zone outside the above-mentioned defect area, wherein, above-mentioned detected image is the image that will carry out defective area detection, and piece dividing processing step, above-mentioned detected image segmentation is become a plurality of and obtain the piece additive value or piece mean value, wherein, above-mentioned additive value is that the pixel value to the pixel that exists in each piece carries out resulting value after the addition, and above-mentioned mean value is resulting value after the number of above-mentioned additive value divided by the pixel that exists in this piece; Have or not above-mentioned additive value of judging part output or piece mean value to defect area, wherein, this defect area has or not judging part to judge whether to exist the outlier of above-mentioned additive value or piece mean value by statistical treatment, thereby judges whether to exist above-mentioned defect area.

Utilize the defect inspection method of said structure, owing in each step, realized the processing same with defect detecting device of the present invention, so can access effect and the effect identical with defect detecting device of the present invention.

In order to solve above-mentioned problem, defect inspection method of the present invention is a kind of defect inspection method that detects defect area in the digital picture, and wherein, the pixel value that the pixel value of this defect area compares to the zone around this defect area is uneven variation.Defect inspection method of the present invention, have: image generates step, generate color difference image according to detected image, wherein, above-mentioned detected image is the image that will carry out defective area detection, and piece dividing processing step, to generate the color difference image that generates in the step at above-mentioned image and be divided into a plurality of, and obtain piece additive value or piece mean value, wherein, above-mentioned additive value is that the pixel value to the pixel that exists in each piece carries out resulting value after the addition, and above-mentioned mean value is resulting value after the number of above-mentioned additive value divided by the pixel that exists in this piece; Have or not above-mentioned additive value of judging part output or piece mean value to defect area, wherein, this defect area has or not judging part to judge whether to exist the outlier of above-mentioned additive value or piece mean value by statistical treatment, thereby judges whether to exist above-mentioned defect area.

Utilize the defect inspection method of said structure, owing in each step, realized the processing same with defect detecting device of the present invention, so can access effect and the effect identical with defect detecting device of the present invention.

It is very clear that other purposes of the present invention, feature and advantage can become in the following description.In addition, come clear and definite advantage of the present invention with reference to accompanying drawing below.

Description of drawings

Fig. 1 is the block diagram of summary structure of the defect detecting device of expression one embodiment of the present invention.

Fig. 2 is the figure of the colour original of employed detected image in the expression defect detecting device shown in Figure 1.

Fig. 3 is the figure of another colour original of employed detected image in the expression defect detecting device shown in Figure 1.

Fig. 4 is the figure of expression according to the luminance picture that colour original generated of detected image shown in Figure 2.

Fig. 5 is the figure of expression according to the luminance picture that colour original generated of detected image shown in Figure 3.

Fig. 6 is the figure of expression according to the color difference image that colour original generated of detected image shown in Figure 3.

Fig. 7 represents a histogrammic example of the distribution of the piece additive value of trying to achieve by the piece cutting treatmenting part by defect detecting device shown in Figure 1.

Fig. 8 is the process flow diagram of good and bad judging part when each piece that detected image generated being carried out good and bad judgement in the expression defect detecting device shown in Figure 1.

Fig. 9 is that the maximal value of the piece additive value of being tried to achieve according to statistical treatment portion of the good and bad judging part in the expression defect detecting device shown in Figure 1 is carried out good and bad process flow diagram when judging.

Figure 10 is the process flow diagram of the processing sequence in the expression defect detecting device shown in Figure 1.

Figure 11 is the figure of an example of employed wave filter in the pixel value correction unit of expression defect detecting device shown in Figure 1.

Figure 12 is the figure that has used resultant image under the situation of wave filter shown in Figure 11 in the pixel value correction unit of expression defect detecting device shown in Figure 1.

Figure 13 is the figure of the image after the statistical treatment portion of expression defect detecting device shown in Figure 1 handles image shown in Figure 12.

Figure 14 (a) is the figure of another example of employed wave filter in the pixel value correction unit of expression defect detecting device shown in Figure 1.

Figure 14 (b) is the figure of other examples of employed wave filter in the pixel value correction unit of expression defect detecting device shown in Figure 1.

Figure 14 (c) is the figure of other example of employed wave filter in the pixel value correction unit of expression defect detecting device shown in Figure 1.

Figure 15 (a) is the figure of the example cut apart of the piece in the piece cutting treatmenting part of expression defect detecting device shown in Figure 1.

Figure 15 (b) is the figure of other examples of cutting apart of the piece in the piece cutting treatmenting part of expression defect detecting device shown in Figure 1.

Figure 15 (c) is the figure of other examples of cutting apart of the piece in the piece cutting treatmenting part of expression defect detecting device shown in Figure 1.

Figure 15 (d) is the figure of other examples of cutting apart of the piece in the piece cutting treatmenting part of expression defect detecting device shown in Figure 1.

Figure 16 is the figure of the luminance picture of employed detected image in the expression defect detecting device shown in Figure 1.

Figure 17 is the rim detection portion of expression defect detecting device shown in Figure 1 carries out the image after the rim detection to luminance picture shown in Figure 16 figure.

Figure 18 is after the piece cutting treatmenting part of expression defect detecting device shown in Figure 1 is divided into the rectangle of vertical length with image block shown in Figure 17, the figure of the image after statistical treatment portion handles.

Figure 19 is after the piece cutting treatmenting part of expression defect detecting device shown in Figure 1 is divided into square with image block shown in Figure 17, the figure of the image after statistical treatment portion handles.

Figure 20 is the rim detection portion of expression defect detecting device shown in Figure 1 carries out the image after the rim detection to color difference image shown in Figure 6 figure.

Figure 21 is the figure of the image after the statistical treatment portion of expression defect detecting device shown in Figure 1 handles image shown in Figure 20.

Figure 22 is the example that the color filters of presentation video sensor image is arranged, and the correlogram between the pixel of the pixel of imageing sensor image and coloured image, luminance picture, color difference image.

Figure 23 is the block diagram of schematic configuration of the image sensor module of the built-in defect detecting device shown in Figure 1 of expression.

Figure 24 is the rim detection portion of expression defect detecting device shown in Figure 1 carries out the image after the rim detection to luminance picture shown in Figure 4 figure.

Figure 25 is that expression utilizes the defect inspection method of prior art image shown in Figure 24 to be carried out the figure of the image under the situation of statistical treatment.

Figure 26 is the rim detection portion of expression defect detecting device shown in Figure 1 carries out the image after the rim detection to luminance picture shown in Figure 5 figure.

Figure 27 is that expression utilizes the defect inspection method of prior art image shown in Figure 26 to be carried out the figure of the image under the situation of statistical treatment.

Embodiment

(1. Zhuan Zhi basic structure)

Below by Fig. 1～Figure 27 one embodiment of the present invention is described.

Fig. 1 is the block diagram of schematic configuration of the defect detecting device 1 of expression present embodiment.Defect detecting device 1 comprises: brightness/color difference image generating unit (image production part) 11, point defect removal portion 12, noise remove portion 13, compression of images portion 14, rim detection portion 15, pixel value correction unit 16, piece cutting treatmenting part 17, statistical treatment portion 18 (defect area has or not judging part) and good and bad judging part 19 (defect area has or not judging part).

Brightness/color difference image generating unit 11 reads becomes the view data of detected object image (detected image), and generates luminance picture or color difference image according to this detected image.Wherein, this detected image is that the imaging apparatus (not shown) of CCD (chargecoupled devices) sensor etc. by having in the image taking sensor device for example is captured.As Fig. 4 or shown in Figure 5, luminance picture is meant and utilizes following formula:

Y＝-0.299×R+0.587×G+0.114×B

With Fig. 2 or shown in Figure 3 (R: red, G: green, B: the pixel value of the colour original of indigo plant) shown detected image is transformed into the image of brightness signal Y with RGB.In addition, utilize the colour original of detected image shown in Figure 2, generate luminance picture shown in Figure 4; Utilize the colour original of detected image shown in Figure 3, generate the luminance picture of Fig. 5.

In addition, color difference image is meant and utilizes following formula as shown in Figure 6:

U＝B-Y

=-0.299 * R-0.587 * G+0.866 * B+ imbalance

V＝R-Y

=0.701 * R-0.587 * G-0.144 * B+ imbalance

The pixel value of colour original with the represented detected image of RGB shown in Figure 3 is transformed into the image of colour difference signal U and V.At this, Y represents above-mentioned luminance signal.In addition, the value of GTG in the middle of the imbalance expression, for example, and in the time of 8, imbalance=128, in the time of 10, imbalance=512.In addition, color difference image shown in Figure 6 is to utilize the equation of U=B-Y to generate on the basis of the colour original of detected image shown in Figure 3, but is not limited thereto, and also can utilize the equation of V=R-Y to generate.

Point defect removal portion 12 utilizes point defect to remove wave filter and removes point defect in the detected image.It is a kind of like this wave filter that point defect is removed wave filter, promptly, for example from being to obtain intermediate value, maximal value and the minimum value of pixel value 9 pixels in 3 * 3 zones at center with the concerned pixel, if the pixel value of concerned pixel is maximal value (minimum value), and 8 pixels around comparing to exist under the situation of poor intentionally (Significance Difference), and the pixel value of concerned pixel is replaced into intermediate value.In addition, the point defect wave filter is not limited in this, also can be that pixel value with concerned pixel is replaced into the mean value of above-mentioned 9 pixels or is the wave filter of the mean value of 8 pixels around the concerned pixel.

Noise in the detected image is removed by using smoothing filter by noise remove portion 13.The described smoothing filter here is meant: can remove the radio-frequency component of image, remove the wave filter of noise.For example, obtain with the concerned pixel be 9 pixels in 3 * 3 zones at center pixel value mean value and with the pixel value of this mean value as concerned pixel, such wave filter is equivalent to smoothing filter.

Compression of images portion 14 comes detected image is compressed processing by longitudinal size and the lateral dimension that dwindles detected image.Thus, will reduce the data volume of detected image, and can remove the noise contribution of failing to remove by noise remove portion 13.

Rim detection portion 15 uses Laplce (Laplacian) wave filter for example to emphasize the edge of spot defective, irregular defective and line defect.Here, described Laplace filter is the wave filter that is used for the edge of the deep or light variation of abstract image.For example, four times of values of the pixel value of concerned pixel deduct that resulting value is as the pixel value of concerned pixel behind the pixel value of pixel (near 4 pixels) up and down, and such wave filter is equivalent to Laplace filter.

Pixel value correction unit 16 utilizes wave filter to come the correction pixels value.Thus, can further remove denoising, and emphasize the defect area in the detected image.In addition, the detailed description of being handled about pixel value correction unit 16 will be in aftermentioned.

Piece cutting treatmenting part 17 is latticed with detected image segmentation, thereby generates a plurality of.In addition, piece cutting treatmenting part 17 is obtained the piece additive value, and this piece additive value is to carry out resulting value after the addition for being divided into whole pixel values in each piece of detected image of a plurality of.Mode), standard deviation etc. what obtain at this is the piece additive value, but is not limited to this, represents the value of this piece to get final product if can obtain, for example the whole mean value of pixel values, mode (mode: in the piece.Under the different situation of the area of divided each piece, the mean value that also can obtain the whole pixel values in each piece is piece mean value, and replaces the piece additive value with piece mean value.In addition, about will setting forth in the back in detail that piece cutting treatmenting part 17 is handled.

Statistical treatment portion 18 implements statistical treatment in order to detect the piece that has spot defective, irregular defective or line defect to piece additive value or piece mean value that piece cutting treatmenting part 17 is obtained.If have the abnormal conditions of spot defective, irregular defective or line defect etc. in piece, piece additive value that then above-mentioned defective is existing or piece mean value just become outlier (outlier) on statistics.Particularly, statistical treatment portion 18 carries out following processing, obtains maximal value, mean value, standard deviation respectively in a plurality of piece additive value or piece mean value that is:.

The maximal value of 19 pairs of piece additive values of good and bad judging part is carried out the outlier judgement, judges that promptly whether this piece additive value or piece mean value are outlier, judge the quality of detected image thus.Outlier is judged according to following formula.That is,

Evaluation of estimate (maximum)=(maximal value-mean value)/standard deviation 〉=judgment threshold

At this, the maximal value in the following formula, mean value and standard deviation are tried to achieve by statistical treatment portion 18.

And then good and bad judging part 19 judges whether contain spot defective, irregular defective or line defect in a plurality of that are cut apart by piece cutting part 17.In this case, above-mentioned judgement is carried out according to following formula.

Evaluation of estimate=(the piece additive value-mean value of each piece)/standard deviation 〉=judgment threshold

At this, mean value in the following formula and standard deviation are tried to achieve by statistical treatment portion 18.In addition, about determining the method for judgment threshold, will set forth in the back.

In addition, in defect detecting device 1, also comprise storer (not shown).Storer is used for being stored in the various processing needed wave filter, parameter and results in.For example, memory stores in noise remove portion 13 employed smoothing filter, in pixel value correction unit 16 mean value etc. of employed wave filter, the piece additive value of trying to achieve by piece cutting treatmenting part 17 or piece mean value, the piece additive value of trying to achieve by statistical treatment portion 18, and the good and bad judged result exported of good and bad judging part 19.

(the 2. processing of good and bad judging part)

Below explanation is by the method for obtaining of good and bad judging part 19 decision judgment thresholds.In addition, recorded and narrated the situation of utilizing the piece additive value in the following description, also can replace the situation of piece additive value for utilizing piece mean value.

Fig. 7 is the histogrammic example that expression piece additive value distributes.In Fig. 7, transverse axis is represented the piece additive value, and the longitudinal axis is represented the number of piece.If detected image is the image that the image sensor devices from certified products obtains, so, the noise contribution of failing to remove in point defect removal portion 12, noise remove portion 13 and compression of images portion 14 is comparatively outstanding, and the distribution of piece additive value or piece mean value is the shape that is similar to standard profile.Thus, utilization is obtained judgment threshold from the mean value and the standard deviation of the piece additive value of the resultant image of imageing sensor of certified products.

At this, judgment threshold can be obtained according to following,, prepares one or more qualitative pictures that is, and obtains the piece additive value or the piece mean value of each qualitative picture,

The standard deviation of judgment threshold=(mean value of the maximal value of piece additive value-piece additive value)/piece additive value.

In addition, judgment threshold is to utilize the mean value of the piece additive value that statistical treatment portion 18 (with reference to Fig. 1) obtained to obtain, but is not limited to this, also can utilize the mode (mode) of piece additive value or median (intermediate value) to replace the piece additive value.

In addition, obtain the method for judgment threshold, also can utilize Vladimir Smirnov to reject method of inspection, determine judgment threshold according to data volume n and rejection region α (=0.01 etc.) except utilizing following formula.Vladimir Smirnov (Smirnov Grubbs) is rejected method of inspection and is meant, checks out the method for statistics outlier from the sampled data of taking from same parent.According to this method, by definite level (be also referred to as " rejection region ", adopt 0.01,0.05 value usually) and sampled data amount intentionally, thereby can determine threshold value singlely, wherein, this threshold value is used to determine whether the data of detected object are outlier.

In addition, the value (evaluation of estimate) that is adopted when carrying out good and bad judgement is to carry out normalized value with standard deviation, and therefore, judgment threshold is not an absolute value, but is set to the multiple with respect to standard deviation.Set judgment threshold as described above, not judged by the quality of the pixel value deviation effects between detected image.

Then, the processing that good and bad judging part 19 is carried out is described.Fig. 8 represents each piece that is generated by detected image is carried out good and bad flow process when judging.

At first, good and bad judging part 19 is selected a piece (S30) that does not carry out good and bad judgment processing as yet in detected image, then, obtains selected evaluation of estimate (S31).In addition, obtain evaluation of estimate by following formula,

Evaluation of estimate=(piece additive value-mean value of selected)/standard deviation.

Then, 19 pairs of evaluations of estimate of good and bad judging part and judgment threshold compare (S32), if evaluation of estimate more than or equal to judgment threshold, just is judged as image inferior with detected image, and with the coordinate and the evaluation of estimate write store (S33) of this piece.

On the other hand, if evaluation of estimate less than judgment threshold, good and bad judging part 19 just is judged as qualitative picture with detected image, the piece (piece is untreated) that judges whether not carry out as yet in addition good and bad judgment processing again (S34).If be untreated piece, then return the processing of S30.

Do not exist when being untreated piece if in S34, judge, just whether detected image is existed the information write-in memory (S35) of defective.In S35, also can judge the quality of image by in S33, whether having write storer, can also carry out following processing, that is: once more the evaluation of estimate that writes among the S33 is judged, obtain the quality grade of this detected image, and with its write store.

Below, specify the processing of the quality grade of obtaining detected image.In order to obtain quality grade, can be according to evaluation of estimate and judgment threshold poor, set a plurality of grades.For example, can be according to the extent of evaluation of estimate and judgment threshold, be set for " greatly ", " in ", the benchmark of the Three Estate of " little ", and, also the quality grade of detected image can be set at the benchmark of the Three Estate of " the degree of inferiority is higher ", " the degree of inferiority is general ", " the degree of inferiority is lower ".In addition, also can estimate the quality grade of detected image by setting a plurality of judgment thresholds.

In addition, to being undertaken the good and bad judgment processing, can also implement the processing of good and bad judging part 19 except that as described above according to the maximal value of piece additive value by each piece that detected image generated.Treatment scheme about under this situation describes by Fig. 9.

At first, good and bad judging part 19 is asked for evaluation of estimate (maximum) (S40) according to following formula, that is,

Evaluation of estimate (maximum)=(mean value of the maximal value of piece additive value-piece additive value)/standard deviation.

Then, 19 pairs of evaluations of estimate of trying to achieve in S40 of good and bad judging part (maximum) and judgment threshold carry out size relatively (S41), and in S42 or S43 with the judged result write store.

(the 3. summary of treatment scheme)

Figure 10 is the process flow diagram of the processing sequence of expression defect detecting device 1 pair of detected image when carrying out defects detection.

Brightness/color difference image generating unit 11 reads detected image, and generates luminance picture or color difference image (S1).For luminance picture or color difference image by brightness/color difference image generating unit 11 is generated, point defect removal portion 12 for example utilizes defective to remove wave filter and removes point defect (S2), and noise remove portion 13 for example utilizes smoothing filter to remove noise contribution (S3).Afterwards, for the data volume of subduing detected image is also further removed noise contribution, by compression of images portion 14 compression detected images (S4).Then, the edge (S5) of spot defective, irregular defective or the line defect etc. that exist in the detected image of having emphasized to be compressed by rim detection portion 15.

Emphasized that by 16 pairs of quilts of pixel value correction unit the detected image at defective edge carries out pixel value and proofreaies and correct, various defectives will be emphasized (S6).Then, emphasized that by 17 pairs of quilts of piece cutting treatmenting part the detected image of defective carries out piece and cuts apart, and obtained the piece additive value or the piece mean value (S7) of each piece.Then, utilize this piece additive value or piece mean value, carry out statistical treatment, that is, obtain maximal value, mean value and standard deviation (S8) in piece additive value or the piece mean value by statistical treatment portion 18.By will comparing, thereby judge the good and bad of detected image or in each piece, whether have defective (S9) from evaluation of estimate and the above-mentioned judgment threshold that these values are obtained.

According to said structure, the defect detecting device 1 of present embodiment is by being provided with pixel value correction unit 16, and therefore definite defect area and the noise region that is present in the detected image of distinguishing, can detect defect area accurately.In addition, can also suppress false judgment to there being area free from defect to do.

In addition, defect detecting device 1 is owing to be provided with piece cutting treatmenting part 17, thereby decides the shape of piece according to the kind of defect areas such as spot defective, irregular defective and line defect.Thus, can carry out specific detection to each defect area accurately.And then 17 utilizations of piece cutting treatmenting part set in advance, and each corresponding with the defective of spot defective, line defect and oblique generation is block-shaped to be detected.Thus, even under situation about can not handle, also can carry out specific detection to each defect area accurately by pixel value correction unit 16.

And then defect detecting device 1 generates color difference image owing to have brightness/color difference image generating unit 11, thereby also can detect in the detected image owing to look changes the defective that produces.In addition, changing owing to look under the situation of a certain row that the defective produce is positioned at a certain row of vertical direction of detected image or horizontal direction, can detect ordinate defective or horizontal line defective accurately.

In addition, in the processing of defect detecting device shown in Figure 10 1, S6 and S7 are illustrated in processing that pixel value correction unit 16 carried out to finish the back and begun the processing carried out by piece cutting treatmenting part 17.But the present invention is not limited to this processing sequence, also can be to handle the back at piece cutting treatmenting part 17 to begin the processing carried out by pixel value correction unit 16.

Specifically, the detected image that detects the edge by rim detection portion 15 among 17 couples of S5 shown in Figure 10 of piece cutting treatmenting part carries out piece to be cut apart, and obtains the piece additive value or the piece mean value of each piece.Then, the detected image after 16 pairs of pixel value correction units have been handled by this piece cutting treatmenting part 17 carries out pixel value and proofreaies and correct.That is to say that 16 pairs of piece additive value or piece mean values of being obtained by piece cutting treatmenting part 17 of pixel value correction unit are proofreaied and correct.

In addition, can also be: the detected image that 17 pairs of piece cutting treatmenting parts detect the edge by rim detection portion 15 only carries out piece earlier to be cut apart, and afterwards, is handled by pixel value correction unit 16.In this case, cut apart by piece and the detected image that has been corrected pixel value again after the processing of being undertaken by pixel value correction unit 16, be output to piece cutting treatmenting part 17.Then, at piece cutting treatmenting part 17, obtain the piece additive value or the piece mean value of each piece of this detected image.

And then in S3～S5 of Figure 10, the processing sequence of noise remove portion 13, compression of images portion 14 and rim detection portion 15 is not limited to above-mentioned, also can handle according to any order.

In addition, in the defect detecting device 1 of present embodiment, be provided with statistical treatment portion 18 and good and bad judging part 19, but be not limited to this, statistical treatment portion 18 and good and bad judging part 19 also can be set in the external device (ED) such as tester etc.

Below describe the processing that pixel value correction unit 16 and piece cutting treatmenting part 17 carry out in detail.In addition, be described in detail in defects detection under the situation of utilizing color difference image.

(the 4. processing of pixel value correction unit)

Pixel value correction unit 16 utilizations wave filter is as shown in figure 11 proofreaied and correct the pixel value of detected image.In wave filter shown in Figure 11, the pixel value before transverse axis is represented to proofread and correct, the pixel value after the longitudinal axis is represented to proofread and correct.As shown in figure 11, by detected image is used this wave filter, learn: compare to the variable quantity a1 of pixel value than lower part, the variable quantity a2 of the higher part of pixel value is bigger.Therefore, by utilizing this wave filter, behind high more regional calibrated of the pixel value of detected image, its pixel value changes more greatly, so in the detected image of proofreading and correct by pixel value correction unit 16, compare to darker zone relatively, brighter zone will be emphasized.

Specifically, in brightness/color difference image generating unit 11, utilize the colour original generation luminance picture as shown in Figure 4 of detected image shown in Figure 2.After this luminance picture has carried out the processing of above-mentioned S2～S4 (with reference to Figure 10), as shown in figure 24, detect the edge by rim detection portion 15.And, by wave filter the image that this is detected the edge is carried out filtering in pixel value correction unit 16, obtain image shown in Figure 12.

In the defect detecting device of prior art, because in the image after rim detection as shown in figure 24, be difficult to distinguish the defect area A and the noise contribution that will be detected, so as shown in figure 25, in the image after having carried out statistical treatment, defect area A is covered by noise contribution, thereby has the possibility of ignoring this defect area A.But the defect detecting device of present embodiment 1 is owing to be provided with pixel value correction unit 16, so, as shown in figure 12, can distinguish noise contribution and the zone brighter than noise contribution is defect area A, and this defect area A is emphasized.In addition, to image shown in Figure 12, the processing by piece cutting treatmenting part 17 and statistical treatment portion 18 are carried out can positively detect the defect area A that will be detected as shown in Figure 13.

As mentioned above, defect detecting device 1 is distinguished defect area and the noise contribution that is present in the detected image, so can detect defect area accurately really by having pixel value correction unit 16.Thus, defect detecting device 1 can suppress the false judgment to there being area free from defect to do.

In addition, pixel value correction unit 16 is carried out the correction of the pixel value of detected image by using wave filter as shown in figure 11, but is not limited to this, and for example also can using, the wave filter shown in Figure 14 (a)～(c) carries out the correction of pixel value.

Identical with Figure 11, the pixel value before the transverse axis of Figure 14 (a)～(c) is represented to proofread and correct, the pixel value after the longitudinal axis is represented to proofread and correct.Figure 14 (a) is that expression is used to emphasize the wave filter of pixel value than lower part.By utilize this wave filter in pixel value correction unit 16, the contrast of the spot defects composition that pixel value is higher reduces, and averages out by piece cutting treatmenting part 17.Thus, because can weaken the spot defects composition,, detect the big and spot defective less of area with the luminance difference of surrounding pixel so can not be subjected to the influence of point defect and spot defects.Therefore, the processing that can be undertaken by piece cutting treatmenting part 17 and statistical treatment portion 18 positively detects the big and spot defective less with the luminance difference of surrounding pixel of area the detected image after pixel value correction unit 16 has carried out handling.In addition, so-called spot defects is meant, and is bigger than point defect, but the defective littler than common spot defective.

In addition, Figure 14 (b) and (c) be the wave filter that to emphasize the optional position.Particularly by utilizing the wave filter shown in Figure 14 (c), owing to can improve the contrast of specific pixel value level, so can emphasize defective, this defective has the value than pixel value higher level.Thus, can further detect spot defective and irregular defective effectively.In addition, what Figure 14 (b) represented is the wave filter that can only emphasize specific pixel value level, so can emphasize to have a large amount of spot defectives that are similar to the value of these pixel value levels.

(the 5. processing of a cutting treatmenting part)

The piece dividing processing that following illustrated block cutting treatmenting part 17 is carried out.In defect detecting device 1, piece cutting treatmenting part 17 is that spot defective, irregular defective or line defect decide after the shape of piece and cut apart according to the defect area of being transferred by the pixel value correction unit Final 16.

Shown in Figure 15 (a), piece cutting treatmenting part 17 becomes to comprise the square at spot defective edge or is similar to foursquare rectangle detected image segmentation.In addition, according to the size of piece,, will appear on the limit of detected image and have the pixel that is not comprised by piece if begin to cut apart successively from the upper left side of detected image.In this case, carrying out piece according to the limit of image cuts apart and gets final product.In addition, piece is cut apart also and can not carried out in order from the upper left side, as long as can comprise whole pixels of detected image when cutting apart.In addition, by use a plurality of shown in Figure 15 (a) piece of shape, the edge of above-mentioned spot defective will be comprised in the piece.

In addition, be under the situation of line defect at above-mentioned defect area, particularly this line defect is under the situation of ordinate defective, piece cutting treatmenting part 17 becomes to comprise vertical grow elongated square at line defect edge with detected image segmentation shown in Figure 15 (b).Particularly when imaging apparatus is ccd sensor or cmos sensor, be easy to generate the ordinate defective by in the image of making a video recording.Therefore, for this ordinate defective, detected image segmentation is become shown in Figure 15 (b) is vertically elongated square comparatively effective than what grow.

And then, be under the situation of horizontal line defective at above-mentioned defect area, piece cutting treatmenting part 17 becomes to comprise the laterally elongated square than what grow of line defect edge like that with detected image segmentation shown in Figure 15 (c).Particularly on the flat-panel monitor of liquid crystal display or PDP display etc., be easy to generate the horizontal line defective in the shown digital picture.Therefore, shown in Figure 15 (c), detected image segmentation is become laterally long elongated square comparatively effective for this horizontal line defective.

In addition, be under the situation of oblique defective at above-mentioned defect area, piece cutting treatmenting part 17 becomes to comprise the parallelogram at oblique defective edge like that with detected image segmentation shown in Figure 15 (d).In addition, in this was cut apart, the parallelogram that can utilize the horizontal direction of detected relatively image for example to be 45 ° or-45 ° (interior angle is 45 ° or 135 °) also can use the parallelogram that is other angle.

As mentioned above, defect detecting device 1 decides the shape of piece according to the kind of the defect area that occurs in the detected image, so can carry out specific detection to each defect area accurately owing to have piece cutting treatmenting part 17.

In addition, it is block-shaped that piece cutting treatmenting part 17 utilization has transferred the detected image of defect area to decide by the pixel value correction unit Final 16, also can utilize the detected image that detects the edge by rim detection portion 15 among the S5 shown in Figure 10 to decide block-shaped.

In this case, piece cutting treatmenting part 17 is predetermined the block-shaped of Figure 15 (a)～(d) according to spot defective, line defect (ordinate defective or horizontal line defective) and oblique defective, block-shaped the detected image that detects the edge is detected by each.

Specifically, at first, the square that piece cutting treatmenting part 17 carries out shown in Figure 15 (a) is cut apart.Afterwards, obtain piece additive value or piece mean value, carry out the processing of statistical treatment portion 18, and by whether existing the spot defective to judge in 19 pairs of detected images of good and bad judging part.Then, carrying out the vertically long rectangle shown in Figure 15 (b) cuts apart.Then, same as described abovely by the processing of statistical treatment portion 18 and good and bad judging part 19, judge whether there is the ordinate defective in the detected image.Particularly under the situation of the flat-panel monitor that has used liquid crystal display etc., in order to judge whether there is the horizontal line defective in the detected image, cut apart by the laterally long rectangle that piece cutting treatmenting part 17 carries out shown in Figure 15 (c) by statistical treatment portion 18 and good and bad judging part 19.Then, after the cutting apart of parallelogram of having carried out shown in Figure 15 (d),, judge in detected image, whether there is oblique defective by the processing of statistical treatment portion 18 and good and bad judging part 19.In addition, also can utilize each block-shaped detected image to be detected according to any order.In addition, utilizing each block-shaped when detected image is detected, can come the parameter in the processing of variation point defective removal portion 12, noise remove portion 13, compression of images portion 14, rim detection portion 15 and pixel value correction unit 16 or the kind of wave filter according to the shape of each piece.

At this, the processing that piece cutting treatmenting part 17 is carried out under the situation that contains the ordinate defective in detected image is described with reference to Figure 16～Figure 19.

Figure 16 is the luminance picture that is utilized as detected image, contains the ordinate defective that should be detected in this luminance picture.When this luminance picture is implemented the processing of S2～S5 shown in Figure 10, can access the luminance picture that is detected the edge as shown in figure 17.Then, 17 pairs of these luminance pictures of piece cutting treatmenting part are handled.

Figure 18 is that statistical treatment portion 18 carries out the result images after the statistical treatment after being illustrated in piece cutting treatmenting part 17 and having carried out cutting apart with the vertically long rectangle shown in Figure 15 (b).In addition, Figure 19 is illustrated in piece cutting treatmenting part 17 to have carried out after square shown in Figure 15 (a) cuts apart, and statistical treatment portion 18 has carried out the result images after the statistical treatment.

As shown in figure 19, under piece cutting treatmenting part 17 carried out situation that square cuts apart, for the ordinate defective that should be detected, sensitivity was lower.Therefore, in good and bad judging part 19, the ordinate defective that is difficult to be detected is judged as defect area.To this, as shown in figure 18, when having carried out rectangular cutting apart in piece cutting part 17, for the ordinate defective that should be detected, sensitivity is the highest.Therefore, good and bad judging part 19 can detect the ordinate defective that should be detected accurately.

As mentioned above, utilize under the situation of the detected image that has detected the edge by rim detection portion 15 at defect detecting device 1, even detected image is not implemented treatment for correcting in pixel value correction unit 16, can utilize block-shaped detection the shown in Figure 15 (a)～(d) yet.That is to say, utilize each block-shaped detect corresponding that is set in advance in the piece cutting treatmenting part 17 with spot defective, line defect and oblique defective.Thus, can carry out specific detection to each defect area accurately.

(processing when 6. utilizing color difference image)

The defect detecting device 1 of present embodiment generates luminance picture or color difference image in brightness/color difference image generating unit 11.Below explanation detects the situation of ordinate defective from detected image.In addition, about the detection of this ordinate defective, do not limit at this and whether to handle by 16 pairs of detected images of pixel value correction unit.In addition, be located in the piece cutting treatmenting part 17, detected image is divided into the rectangle that the lengthwise shown in Figure 15 (b) is grown by piece.

At first, brightness/color difference image generating unit 11 generates luminance picture shown in Figure 5 and color difference image shown in Figure 6 according to the colour original of detected image shown in Figure 3.The luminance picture and the color difference image that generate are handled by S2～S5 shown in Figure 10 respectively.Figure 20 is illustrated in the color difference image that detects the edge among the S5 by rim detection portion 15, and Figure 26 represents similarly to detect with Figure 20 the luminance picture at edge.

After by piece cutting treatmenting part 17 this luminance picture and color difference image being divided into vertically long rectangle, obtain piece additive value or piece mean value.Figure 21 and Figure 27 be expression according to this piece additive value or piece mean value, carry out result images after the statistical treatment by statistical treatment portion 18.

At this, as shown in figure 27, when detected image was luminance picture, the defect area that should be detected as the ordinate defective was not detected as defect area; And as shown in figure 21, when detected image was color difference image, then having detected should be as the defect area of ordinate defective.

Defect detecting device 1 has brightness/color difference image generating unit 11, generates color difference image.Thus, exist under the situation of defective owing to look changes, can detect this defective at detected image.Particularly, can detect this ordinate defective because look changes under the situation that the defective that is produced is vertical row.About its reason, will describe with reference to Figure 22 following.

Figure 22 is the example that the color filters of imageing sensor image is arranged, and has represented the correlogram between the pixel of the pixel of imageing sensor image and coloured image, luminance picture, color difference image.As shown in figure 22, each pixel of imageing sensor image is expressed as R, G or B signal respectively, and 4 pixels of imageing sensor image are equivalent to 1 pixel of coloured image, luminance picture or color difference image.In addition, the color filters of imageing sensor image is arranged regularly and is made file and the represented file alternate repetition of GBGBGB... (that is GB group) that RGRGRG... (that is RG group) is represented arrange.

At this, when in a certain file ordinate defective being arranged, this file pixel value is different from left side the 2nd row of this file or the pixel value of the homochromy wave filter arrangement that right side the 2nd is listed as, and the variation of this file pixel value is regarded as look and changes on color digital image.Therefore, be under the situation of color difference image adopting look information, when producing the defective of file, can detect the ordinate defective accurately owing to the look variation.

In addition, as mentioned above, in piece cutting treatmenting part 17, color difference image is carried out piece cut apart, form vertically long rectangle, but be not limited to this.Piece cutting treatmenting part 17 also can be divided into the color difference image piece the laterally long rectangle shown in Figure 15 (c).In this case, when in a certain line, producing the horizontal line defective, with the situation that produces the ordinate defective in the same manner, the variation of this line pixel value is regarded as look and changes on color digital image.Therefore, when producing the defective of line, can detect the horizontal line defective accurately owing to the look variation.

In addition, color difference image shown in Figure 6 is detected image transformation to be become the image of colour difference signal U, but is not limited to this, utilizes the image that is transformed into colour difference signal V also can access same result.

In addition, the ordinate defective that luminance picture comprised shown in Figure 5 more is difficult to be detected than the ordinate defective that luminance picture comprised shown in Figure 16.That is, the ordinate defective that former figure comprised that the ordinate defective that colour original comprised of detected image shown in Figure 3 compares to luminance picture shown in Figure 16 more is difficult to be detected, and wherein, this former figure is the colour original of detected image.

As mentioned above, in defect detecting device 1, brightness/color difference image generating unit 11 generates color difference image as illustrated in fig. 6.

So, detected image block is divided under the situation of the ordinate defective that vertically long rectangle is difficult to detect even only in detected image, contain by piece cutting treatmenting part 17, defect detecting device 1 can detect above-mentioned ordinate defective accurately by utilizing above-mentioned color difference image.In addition, under being divided into detected image block laterally than long rectangular situation, piece cutting treatmenting part 17 also can obtain same effect.

According to said structure, the defect detecting device 1 of present embodiment is distinguished defect area and the noise contribution that is present in the detected image really, thereby can be detected defect area accurately owing to have pixel value correction unit 16.Thus, can suppress false judgment to there being area free from defect to do.

In addition, defect detecting device 1 is owing to have piece cutting treatmenting part 17, and its kind according to defect areas such as spot defective, irregular defective and line defects decides the shape of piece.So, can carry out specific detection to each defect area accurately.And then 17 utilizations of piece cutting treatmenting part set in advance, and each corresponding with the defective of spot defective, line defect and oblique generation is block-shaped to be detected.Thus, even under situation about can not handle, also can carry out specific detection to each defect area accurately by pixel value correction unit 16.

And then defect detecting device 1 is owing to have brightness/color difference image generating unit 11, and generates color difference image, changes the defective that cause because of look thereby can detect in detected image.Particularly, when the defective that causes because of the look variation is a longitudinal row or horizontal row, can detect ordinate defective or horizontal line defective accurately.

(example 7. is installed)

Figure 23 is a block diagram of representing the schematic configuration of image sensor module 2 built-in in the defect detecting device 1 of present embodiment.Image sensor module 2 possesses image sensor devices 3 and digital signal processor (digital signal processor is designated hereinafter simply as DSP) 4.In addition, DSP4 possesses the defect detecting device 1 of present embodiment.

After the photodiode of image sensor devices 3 by its inner all each pixel reads light signal, and this converting optical signals become electric signal.This electric signal is carried out digital conversion, and be transformed into the image C that in the DSP4 of back one-level, to handle, and export to DSP4.

DSP4 is the specific microprocessor that is used for digital signal processing, can realize processing at a high speed.Image C is input in the defect detecting device 1 set in the DSP4 and carries out above-mentioned processing (with reference to Figure 10), afterwards the quality of image C is judged.According to judged result, for example,, the quality of image sensor devices 3 or image sensor module 2 is judged by being set at the good and bad judging part (not shown) of device in the defect detecting device 1 in the DSP4 to this image C.In addition, the good and bad judged result of image sensor devices 2 is output to the display part (not shown) such as the testing fixture that is used for check image sensor assembly 2.

As mentioned above because image sensor module 2 has defect detecting device 1, thereby the quality of the image that can be accurately image sensor devices 3 be read judge, so, also can detect the quality of judging image sensor devices 3 accurately.

In addition, as shown in figure 23, defect detecting device 1 is set among the DSP4 in the image sensing module 2, still, as long as defect detecting device 1 is set in the image sensor module 2, for example, also can be set at the inside of image sensor devices 2.In addition, except the inside of image sensor module 2, defect detecting device 1 also can be installed in the external device (ED), for example can be installed in have the RGB separated part, CPU (central processing unit) portion and having in the image processing apparatus (not shown) of a plurality of storeies etc., perhaps also can be installed in the tester (not shown) that is used for detecting quality of digital images.In this case, defect detecting device 1 can possess above-mentioned repertoire, for example externally is provided with in the device: in the piece treatment step as shown in Figure 1, and the statistical treatment portion 18 that begins from the centre, good and bad judging part 19 etc.In addition, the defect test device 1 of present embodiment, after rim detection portion 15 had emphasized the edge of spot defective, irregular defective, line defect etc., pixel value correction unit 16 was handled again.But being not limited to this, also can be such structure, after promptly pixel value correction unit 16 has carried out earlier handling, and the structure that rim detection portion 15 handles again.

(8. replenishing)

The defect inspection method of being carried out by the defect detecting device of the present embodiment also program of can be used as is recorded in the recording medium of embodied on computer readable, wherein, this recording medium recording the program of carrying out by computing machine.Its result can provide a kind of recording medium that can freely carry, and it has write down the program of carrying out the defect inspection method of present embodiment.

Recording medium also can be to be used for the storer (not shown) handled at microcomputer, for example, it can be program medium such as ROM (read only memory), also can be the following program medium that can read, thereby that is: in the program reading device that is provided with, insert the program medium that recording medium reads, wherein, this program reading device (not shown) is as external memory.

No matter be which kind of situation, can and carry out the program of being preserved by microprocessor access, also can adopt following manner, that is: read routine is downloaded to the program of being read in the program storage area (not shown) of microcomputer, carries out this program.In this case, the program of download usefulness is stored in the body apparatus in advance.

The said procedure medium is the recording medium that can separate with body, it also can be the following medium of fixedly maintenance program, that is: the band system of tape, cartridge tape etc., the disc system of CD that comprises disk, CD-ROM, MO, MD, the DVD etc. of floppy disk, hard disk etc., the card system of IC-card (comprising memory card), light-card etc., perhaps, the semiconducter memory system of mask model ROM, EPROM (Erasable ProgrammableRead Only Memory), EEPROM (Electrically Erasable Programmable ReadOnly Memory), flash rom etc.

In addition, under said circumstances, because it constitutes: the system that can be connected with the communication network that comprises the internet, so, also can resemble from the downloaded program and fluidly keep program.In addition,, can in advance above-mentioned download be stored in the receiver with program, perhaps, from other recording medium, download under the situation of downloaded program.

The present invention is not limited to above-mentioned embodiment, can carry out various changes in the scope shown in the claim.That is, also be comprised in the technical scope of the present invention by the resulting embodiment of technological means that is combined in appropriateness change in the scope shown in the claim.

As mentioned above, defect detecting device of the present invention is a kind of defect detecting device that detects defect area in the digital picture, and wherein, the pixel value that the pixel value of this defect area compares to the zone around this defect area is uneven variation.Defect detecting device of the present invention, have: the pixel value correction unit, proofread and correct the pixel value in the detected image, make that the pixel value in the zone that should detect as above-mentioned defect area is emphasized with respect to the pixel value in the zone outside the above-mentioned defect area, wherein, above-mentioned detected image is the image that will carry out defective area detection, and piece cutting treatmenting part, above-mentioned detected image segmentation is become a plurality of and obtain the piece additive value or piece mean value, wherein, above-mentioned additive value is that the pixel value to the pixel that exists in each piece carries out resulting value after the addition, and above-mentioned mean value is that above-mentioned additive value is divided by the resulting value of the number of the pixel that exists in this piece; Have or not above-mentioned additive value of judging part output or piece mean value to defect area, wherein, this defect area has or not judging part to judge whether to exist the outlier of above-mentioned additive value or piece mean value by statistical treatment, thereby judges whether to exist above-mentioned defect area.

According to said structure, in detected image, exist under the situation of defect area, pixel value correction unit correction pixels value makes that the pixel value of this defect area is emphasized.In addition, the piece cutting treatmenting part is a plurality of with detected image segmentation and obtains the piece additive value or piece mean value.That is, in defect detecting device, pixel value correction unit and piece cutting treatmenting part are handled detected image separately.Piece additive value of obtaining at the piece cutting treatmenting part or piece mean value are output to defect area and have or not judging part.Then, defect area has or not judging part, carries out statistical treatment according to this piece additive value or piece mean value, detected image is had the judgement of area free from defect.

Thus, owing to can distinguish defect area and the noise contribution that is present in the detected image really, so can detect defect area accurately.In addition, can suppress false judgment to there being area free from defect to do.

In addition, piece additive value or piece mean value will be output to defect area and have or not judging part, and obtain the timing of the processing of this piece additive value or piece mean value for the piece cutting treatmenting part, and can be before the pixel value correction unit be handled, also can be after the pixel value correction unit be handled.Promptly, can be before the pixel value correction unit be implemented to proofread and correct to the pixel value of detected image, the piece cutting treatmenting part is obtained the piece additive value or the piece mean value of each piece of detected image earlier, also can after implementing to proofread and correct to the pixel value of detected image, the pixel value correction unit obtain piece additive value or piece mean value.

And then above-mentioned cutting treatmenting part preferably cut apart having implemented the detected image that pixel value proofreaies and correct by above-mentioned pixel value correction unit.In addition, above-mentioned pixel value correction unit is preferably proofreaied and correct the pixel value that above-mentioned cutting treatmenting part cut apart the piece of above-mentioned detected image gained, and above-mentioned cutting treatmenting part preferably obtained piece additive value or the piece mean value of having been implemented each piece of pixel value correction by above-mentioned pixel value correction unit.Above-mentioned pixel value correction unit is preferably proofreaied and correct being divided into a plurality of and the pixel value of having obtained the above-mentioned detected image of above-mentioned additive value or piece mean value by above-mentioned cutting treatmenting part.

According to said structure, pixel value correction unit and piece cutting treatmenting part are handled detected image.Thus, owing to can distinguish defect area and the noise contribution that is present in the detected image really, so can detect defect area accurately.In addition, can suppress false judgment to there being area free from defect to do.

And then above-mentioned pixel value correction unit is preferably by utilizing wave filter that above-mentioned detected image is proofreaied and correct, and makes the high more zone of pixel value of detected image, and the variation of its pixel value is big more.

According to said structure, the pixel value correction unit is come the correction pixels value by the wave filter that high more its pixel value of zone of pixel value that utilizes detected image changes more greatly, so, make that the pixel value of defect area is emphasized.Thus, owing to can distinguish defect area and the noise contribution that is present in the detected image really, so can detect defect area accurately.

Defect detecting device of the present invention is a kind of defect detecting device that detects defect area in the digital picture, and wherein, the pixel value that the pixel value of this defect area compares to the zone around this defect area is uneven variation.Defect detecting device of the present invention has the piece cutting treatmenting part, detected image segmentation is become in advance block-shaped according to the shape set of above-mentioned defect area, and obtain piece additive value or piece mean value, wherein, above-mentioned detected image is the image that will carry out defective area detection, above-mentioned additive value is that the pixel value to the pixel that exists in each piece carries out resulting value after the addition, and above-mentioned mean value is resulting value after the number of above-mentioned additive value divided by the pixel that exists in this piece; Have or not above-mentioned additive value of judging part output or piece mean value to defect area, wherein, this defect area has or not judging part to judge whether to exist the outlier of above-mentioned additive value or piece mean value by statistical treatment, thereby judges whether to exist above-mentioned defect area.

According to said structure, the piece cutting treatmenting part becomes in advance block-shaped according to the shape set of above-mentioned defect area with detected image segmentation, and obtain the piece additive value or the piece mean value of each piece, afterwards, export this piece additive value or piece mean value to defect area and have or not judging part.Defect area has or not judging part to carry out statistical treatment according to this piece additive value or piece mean value, and detected image is had the judgement of area free from defect.

Thus, because the piece cutting treatmenting part utilizes each corresponding with defect area block-shaped, carry out the detection of each defect area separately, so can carry out specific detection to each defect area accurately.In addition, because defect area has or not the piece additive value or the piece mean value of judging part utilization handled detected image in the piece cutting treatmenting part to handle, so can suppress false judgment to there being area free from defect to do.

Above-mentioned block-shaped preferred vertically long rectangle.For example, when taking detected image, be easy to generate the ordinate defective that connects vertical direction in the detected image by the imaging apparatus of ccd sensor or cmos sensor etc.According to said structure, because the block-shaped rectangle of in the piece cutting treatmenting part, using of length that comprised vertically, so can positively detect the ordinate defective.

Above-mentioned block-shaped preferred laterally long rectangle.For example, when the digital picture on the flat-panel monitor that is displayed on liquid crystal display or PDP display etc. becomes detected image, may produce the horizontal line defective of the horizontal direction that connects detected image.According to said structure, because the block-shaped rectangle of in the piece cutting treatmenting part, using of length that comprised laterally, so can positively detect the horizontal line defective.

Defect detecting device of the present invention is a kind of defect detecting device that detects defect area in the digital picture, and wherein, the pixel value that the pixel value of this defect area compares to the zone around this defect area is uneven variation.Defect detecting device of the present invention, have: image production part, generate color difference image according to detected image, wherein, above-mentioned detected image is the image and the piece cutting treatmenting part that will carry out defective area detection, the color difference image that above-mentioned image production part is generated is divided into a plurality of and obtain the piece additive value or piece mean value, wherein, above-mentioned additive value is that the pixel value to the pixel that exists in each piece carries out resulting value after the addition, and above-mentioned mean value is resulting value after the number of above-mentioned additive value divided by the pixel that exists in this piece; Have or not above-mentioned additive value of judging part output or piece mean value to defect area, wherein, this defect area has or not judging part to judge whether to exist the outlier of above-mentioned additive value or piece mean value by statistical treatment, thereby judges whether to exist above-mentioned defect area.

According to said structure, image production part generates the color difference image as detected image.The piece cutting treatmenting part is divided into a plurality of with color difference image, and obtains the piece additive value or the piece mean value of each piece.Then, this piece additive value or piece mean value are output to defect area and have or not judging part, and defect area has or not judging part to carry out statistical treatment according to this piece additive value or piece mean value, and detected image is had the judgement of area free from defect.

Thus, can detect in the detected image owing to look changes the defective that produces.In addition, changing owing to look under the situation that the defective cause is present in the vertical direction of detected image or a certain row on the horizontal direction, can detect ordinate defective or horizontal line defective accurately.And then defect area has or not judging part to utilize piece additive value or the piece mean value of the color difference image that the piece cutting treatmenting part handled to handle, and can suppress the false judgment to there being area free from defect to do.

Above-mentioned cutting part preferably becomes a plurality of with above-mentioned detected image segmentation, makes adjacent piece have superimposed each other part.

If make adjacent piece each other zero lap partly detected image segmentation is a plurality of, defect area will be crossed over adjacent piece sometimes.In this case, defect area may be disperseed, thereby has influence on different two piece additive value or piece mean value, therefore, the situation of defect area might occur judging whether rightly to exist.

According to said structure, the piece cutting treatmenting part has adjacent piece superimposedly partly detected image segmentation to be a plurality of, therefore, can defect area to be suppressed in the piece reliably each other.Thus, can prevent that defect area from having influence on different two piece additive value or mean value, so, can detect defect area accurately.

And then defect detecting device of the present invention preferably also has: point defect removal portion, from above-mentioned detected image, remove point defect, and in this point defect, the difference of the pixel value of certain pixel and the pixel value around it is bigger; Noise remove portion utilizes wave filter to remove the noise contribution of detected image, and wherein, above-mentioned detected image is to have been removed the image of point defect by above-mentioned point defect removal portion; And compression of images portion, the detected image of having removed noise contribution by above-mentioned noise remove portion is compressed processing.

According to said structure,, remove the noise contribution of detected image in noise remove portion in the point defect that detected image is removed by point defect removal portion.Thus, because before noise processed portion handles, from detected image, remove as the point defect outside the detected object, so, can prevent that point defect is as the defect area of spot defective of detected object etc. and be detected.In addition, because compression of images portion compresses detected image, thus can reduce the data volume of detected image, and remove the noise contribution of failing to remove by noise processed portion.

In addition, the defect detecting device of said structure also can be set in image sensor devices or the image sensor module.

Defect inspection method of the present invention is a kind of defect inspection method that detects defect area in the digital picture, and wherein, the pixel value that the pixel value of this defect area compares to this defect area peripheral region is uneven variation.Defect inspection method of the present invention, comprise: the pixel value aligning step, proofread and correct the pixel value in the detected image, make that the pixel value in the zone that should detect as above-mentioned defect area is emphasized with respect to the pixel value in the zone outside the above-mentioned defect area, wherein, above-mentioned detected image is the image that will carry out defective area detection, and piece dividing processing step, above-mentioned detected image segmentation is become a plurality of and obtain the piece additive value or piece mean value, wherein, above-mentioned additive value is that the pixel value to the pixel that exists in each piece carries out resulting value after the addition, and above-mentioned mean value is that above-mentioned additive value is divided by the resulting value of the number of the pixel that exists in this piece; Have or not above-mentioned additive value of judging part output or piece mean value to defect area, wherein, this defect area has or not judging part to judge whether to exist the outlier of above-mentioned additive value or piece mean value by statistical treatment, thereby judges whether to exist above-mentioned defect area.

According to the defect inspection method of said structure, owing in each step, realized the processing same with defect detecting device of the present invention, so can access effect and the effect identical with defect detecting device of the present invention.

Defect inspection method of the present invention is a kind of defect inspection method that detects defect area in the digital picture, and wherein, the pixel value that the pixel value of this defect area compares to the zone around this defect area is uneven variation.Defect inspection method of the present invention, comprise piece dividing processing step, detected image segmentation is become in advance block-shaped according to the shape set of above-mentioned defect area, and obtain piece additive value or piece mean value, wherein, above-mentioned detected image is the image that will carry out defective area detection, and above-mentioned additive value is that the pixel value to the pixel that exists in each piece carries out resulting value after the addition, and above-mentioned mean value is resulting value after the number of above-mentioned additive value divided by the pixel that exists in this piece; Have or not above-mentioned additive value of judging part output or piece mean value to defect area, wherein, this defect area has or not judging part to judge whether to exist the outlier of above-mentioned additive value or piece mean value by statistical treatment, thereby judges whether to exist above-mentioned defect area.

According to the defect inspection method of said structure, owing in each step, realized the processing same with defect detecting device of the present invention, so can access effect and the effect identical with defect detecting device of the present invention.

Defect inspection method of the present invention is a kind of defect inspection method that detects defect area in the digital picture, and wherein, the pixel value that the pixel value of this defect area compares to the zone around this defect area is uneven variation.Defect inspection method of the present invention, comprise: image generates step, generate color difference image according to detected image, wherein, above-mentioned detected image is the image that will carry out defective area detection, and piece dividing processing step, to generate the color difference image that generates in the step at above-mentioned image and be divided into a plurality of, and obtain piece additive value or piece mean value, wherein, above-mentioned additive value is that the pixel value to the pixel that exists in each piece carries out resulting value after the addition, and above-mentioned mean value is resulting value after the number of above-mentioned additive value divided by the pixel that exists in this piece; Have or not above-mentioned additive value of judging part output or piece mean value to defect area, wherein, this defect area has or not judging part to judge whether to exist the outlier of above-mentioned additive value or piece mean value by statistical treatment, thereby judges whether to exist above-mentioned defect area.

According to the defect inspection method of said structure, owing in each step, realized the processing same with defect detecting device of the present invention, so can access effect and the effect identical with defect detecting device of the present invention.

In addition, by means of making computing machine carry out the defects detection program of above-mentioned defect inspection method, utilize computing machine also can obtain effect and the effect same with defect inspection method of the present invention.And, by with above-mentioned defects detection procedure stores in the recording medium of embodied on computer readable, can on computing machine arbitrarily, carry out above-mentioned defects detection program.

Promptly, as mentioned above, defect detecting device of the present invention, have: the pixel value correction unit, proofread and correct the pixel value in the detected image, make that the pixel value in the zone that should detect as above-mentioned defect area is emphasized with respect to the pixel value in the zone outside the above-mentioned defect area, wherein, above-mentioned detected image is the image that will carry out defective area detection, and piece cutting treatmenting part, above-mentioned detected image segmentation is become a plurality of and obtain the piece additive value or piece mean value, wherein, above-mentioned additive value is that the pixel value to the pixel that exists in each piece carries out resulting value after the addition, and above-mentioned mean value is that above-mentioned additive value is divided by the resulting value of the number of the pixel that exists in this piece; Have or not above-mentioned additive value of judging part output or piece mean value to defect area, wherein, this defect area has or not judging part to judge whether to exist the outlier of above-mentioned additive value or piece mean value by statistical treatment, thereby judges whether to exist above-mentioned defect area.

Thus, owing to can distinguish defect area and the noise contribution that is present in the detected image really, so can detect defect area accurately.In addition, can suppress false judgment to there being area free from defect to do.

In addition, defect detecting device of the present invention, has the piece cutting treatmenting part, detected image segmentation is become in advance block-shaped according to the shape set of above-mentioned defect area, and obtain piece additive value or piece mean value, wherein, above-mentioned detected image is the image that will carry out defective area detection, above-mentioned additive value is that the pixel value to the pixel that exists in each piece carries out resulting value after the addition, and above-mentioned mean value is resulting value after the number of above-mentioned additive value divided by the pixel that exists in this piece; Have or not above-mentioned additive value of judging part output or piece mean value to defect area, wherein, this defect area has or not judging part to judge whether to exist the outlier of above-mentioned additive value or piece mean value by statistical treatment, thereby judges whether to exist above-mentioned defect area.

Thus, because the piece cutting treatmenting part utilizes each corresponding with defect area block-shaped, carry out the detection of each defect area separately, so can specific to each defect area accurately detection.In addition, because defect area has or not the piece additive value or the piece mean value of the detected image that the judging part utilization handled in the piece cutting treatmenting part to handle, so can suppress false judgment to there being area free from defect to do.

In addition, as mentioned above, defect detecting device of the present invention, have: image production part, generate color difference image according to detected image, wherein, above-mentioned detected image is the image that will carry out defective area detection, and piece cutting treatmenting part, the color difference image that above-mentioned image production part is generated is divided into a plurality of and obtain the piece additive value or piece mean value, wherein, above-mentioned additive value is that the pixel value to the pixel that exists in each piece carries out resulting value after the addition, and above-mentioned mean value is resulting value after the number of above-mentioned additive value divided by the pixel that exists in this piece; Have or not above-mentioned additive value of judging part output or piece mean value to defect area, wherein, this defect area has or not judging part to judge whether to exist the outlier of above-mentioned additive value or piece mean value by statistical treatment, thereby judges whether to exist above-mentioned defect area.

Thus, can detect in detected image owing to look changes the defective produce.In addition, changing owing to look under the situation that the defective cause is present in the vertical direction of detected image or a certain row on the horizontal direction, can detect ordinate defective or horizontal line defective accurately.And then defect area has or not piece additive value or the piece mean value of the color difference image that the judging part utilization handled at the piece cutting treatmenting part to handle, and can suppress the false judgment to there being area free from defect to do.

In addition, by defect detecting device of the present invention, can detect the existing defective of image sensor devices accurately.Defect detecting device of the present invention is specially adapted to require the Quality Detection of high quality images sensor component, also can be applied to the detection of the shown digital picture of flat-panel monitor (for example: LCD or plasma display).

More than, the present invention is had been described in detail, above-mentioned embodiment or embodiment only are the examples that discloses technology contents of the present invention, the present invention is not limited to above-mentioned concrete example, should not carry out the explanation of narrow sense, can in the scope of spirit of the present invention and claim, carry out various changes and implement it the present invention.

Claims (16)

1. a defect detecting device detects defect area in digital picture, and wherein, the pixel value that the pixel value of this defect area compares to the zone around this defect area is uneven variation, it is characterized in that having:

The pixel value correction unit, proofread and correct the pixel value in the detected image, make the pixel value in the zone that should detect as above-mentioned defect area be emphasized that with respect to the pixel value in the zone outside the above-mentioned defect area wherein, above-mentioned detected image is the image that will carry out defective area detection; And

The piece cutting treatmenting part, above-mentioned detected image segmentation is become a plurality of and obtain the piece additive value or piece mean value, wherein, above-mentioned additive value is that the pixel value to the pixel that exists in each piece carries out resulting value after the addition, above-mentioned mean value is that above-mentioned additive value is divided by the resulting value of the number of the pixel that exists in this piece

Have or not above-mentioned additive value of judging part output or piece mean value to defect area, wherein, this defect area has or not judging part to judge whether to exist the outlier of above-mentioned additive value or piece mean value by statistical treatment, thereby judges whether to exist above-mentioned defect area.

2. defect detecting device according to claim 1 is characterized in that:

Above-mentioned cutting treatmenting part cut apart having implemented the detected image that pixel value proofreaies and correct by above-mentioned pixel value correction unit.

3. defect detecting device according to claim 1 is characterized in that:

The pixel value that above-mentioned pixel value correction unit is cut apart the piece of above-mentioned detected image gained to above-mentioned cutting treatmenting part is proofreaied and correct;

Above-mentioned cutting treatmenting part obtained piece additive value or the piece mean value of having been implemented each piece of pixel value correction by above-mentioned pixel value correction unit.

4. defect detecting device according to claim 1 is characterized in that:

Above-mentioned pixel value correction unit is proofreaied and correct being divided into a plurality of and the pixel value of having obtained the above-mentioned detected image of above-mentioned additive value or piece mean value by above-mentioned cutting treatmenting part.

5. defect detecting device according to claim 1 is characterized in that:

Above-mentioned pixel value correction unit utilizes wave filter that above-mentioned detected image is proofreaied and correct, and makes that the variation of its pixel value of zone that pixel value is high more in above-mentioned detected image is big more.

6. a defect detecting device detects defect area in digital picture, and wherein, the pixel value that the pixel value of this defect area compares to the zone around this defect area is uneven variation, it is characterized in that:

Has the piece cutting treatmenting part, detected image segmentation is become in advance block-shaped according to the shape set of above-mentioned defect area, and obtain piece additive value or piece mean value, wherein, above-mentioned detected image is the image that will carry out defective area detection, above-mentioned additive value is that the pixel value to the pixel that exists in each piece carries out resulting value after the addition, and above-mentioned mean value is resulting value after the number of above-mentioned additive value divided by the pixel that exists in this piece

Have or not above-mentioned additive value of judging part output or piece mean value to defect area, wherein, this defect area has or not judging part to judge whether to exist the outlier of above-mentioned additive value or piece mean value by statistical treatment, thereby judges whether to exist above-mentioned defect area.

7. defect detecting device according to claim 6 is characterized in that:

The above-mentioned block-shaped rectangle that is vertical than length.

8. defect detecting device according to claim 6 is characterized in that:

The above-mentioned block-shaped rectangle that is horizontal than length.

9. a defect detecting device detects defect area in digital picture, and wherein, the pixel value that the pixel value of this defect area compares to the zone around this defect area is uneven variation, it is characterized in that having:

Image production part generates color difference image according to detected image, and wherein, above-mentioned detected image is the image that will carry out defective area detection; And

The piece cutting treatmenting part, the color difference image that above-mentioned image production part is generated is divided into a plurality of and obtain the piece additive value or piece mean value, wherein, above-mentioned additive value is that the pixel value to the pixel that exists in each piece carries out resulting value after the addition, above-mentioned mean value is resulting value after the number of above-mentioned additive value divided by the pixel that exists in this piece

Have or not above-mentioned additive value of judging part output or piece mean value to defect area, wherein, this defect area has or not judging part to judge whether to exist the outlier of above-mentioned additive value or piece mean value by statistical treatment, thereby judges whether to exist above-mentioned defect area.

10. according to any described defect detecting device in the claim 1,6,9, it is characterized in that:

Above-mentioned cutting part becomes a plurality of with above-mentioned detected image segmentation, makes adjacent piece have superimposed each other part.

11. any described defect detecting device according in the claim 1,6,9 is characterized in that also having:

Point defect is removed by point defect removal portion from above-mentioned detected image, in this point defect, the difference of the pixel value of certain pixel and the pixel value around it is bigger;

Noise remove portion utilizes wave filter to remove the noise contribution of detected image, and wherein, above-mentioned detected image is to have been removed the image of point defect by above-mentioned point defect removal portion; And

Compression of images portion compresses processing to the detected image of having removed noise contribution by above-mentioned noise remove portion.

12. an image sensor devices is characterized in that:

Has any described defect detecting device in the claim 1,6,9.

13. an image sensor module is characterized in that:

Has any described defect detecting device in the claim 1,6,9.

14. a defect inspection method detects defect area in digital picture, wherein, the pixel value that the pixel value of this defect area compares to the zone around this defect area is uneven variation, it is characterized in that, comprising:

The pixel value aligning step, proofread and correct the pixel value in the detected image, make the pixel value in the zone that should detect as above-mentioned defect area be emphasized that with respect to the pixel value in the zone outside the above-mentioned defect area wherein, above-mentioned detected image is the image that will carry out defective area detection; And

Piece dividing processing step, above-mentioned detected image segmentation is become a plurality of and obtain the piece additive value or piece mean value, wherein, above-mentioned additive value is that the pixel value to the pixel that exists in each piece carries out resulting value after the addition, above-mentioned mean value is resulting value after the number of above-mentioned additive value divided by the pixel that exists in this piece

Have or not above-mentioned additive value of judging part output or piece mean value to defect area, wherein, this defect area has or not judging part to judge whether to exist the outlier of above-mentioned additive value or piece mean value by statistical treatment, thereby judges whether to exist above-mentioned defect area.

15. a defect inspection method detects defect area in digital picture, wherein, the pixel value that the pixel value of this defect area compares to this defect area peripheral region is uneven variation, it is characterized in that:

Comprise piece dividing processing step, detected image segmentation is become in advance block-shaped according to the shape set of above-mentioned defect area, and obtain piece additive value or piece mean value, wherein, above-mentioned detected image is the image that will carry out defective area detection, above-mentioned additive value is that the pixel value to the pixel that exists in each piece carries out resulting value after the addition, and above-mentioned mean value is resulting value after the number of above-mentioned additive value divided by the pixel that exists in this piece;

Have or not above-mentioned additive value of judging part output or piece mean value to defect area, wherein, this defect area has or not judging part to judge whether to exist the outlier of above-mentioned additive value or piece mean value by statistical treatment, thereby judges whether to exist above-mentioned defect area.

16. a defect inspection method detects defect area in digital picture, wherein, the pixel value that the pixel value of this defect area compares to the zone around this defect area is uneven variation, it is characterized in that, comprising:

Image generates step, generates color difference image according to detected image, and wherein, above-mentioned detected image is the image that will carry out defective area detection; And

Piece dividing processing step, to generate the color difference image that generates in the step at above-mentioned image and be divided into a plurality of, and obtain piece additive value or piece mean value, wherein, above-mentioned additive value is that the pixel value to the pixel that exists in each piece carries out resulting value after the addition, above-mentioned mean value is resulting value after the number of above-mentioned additive value divided by the pixel that exists in this piece

Have or not above-mentioned additive value of judging part output or piece mean value to defect area, wherein, this defect area has or not judging part to judge whether to exist the outlier of above-mentioned additive value or piece mean value by statistical treatment, thereby judges whether to exist above-mentioned defect area.

Images