WO1997036260A1 - Method, apparatus and system for verification of patterns - Google Patents

Method, apparatus and system for verification of patterns

Info

Publication number
WO1997036260A1
WO1997036260A1 PCT/CA1997/000196 CA9700196W WO1997036260A1 WO 1997036260 A1 WO1997036260 A1 WO 1997036260A1 CA 9700196 W CA9700196 W CA 9700196W WO 1997036260 A1 WO1997036260 A1 WO 1997036260A1
Authority
WO
Grant status
Application
Patent type
Prior art keywords
image
template
comparison
figure
acquired
Prior art date
Application number
PCT/CA1997/000196
Other languages
French (fr)
Inventor
Eric Clifford Pearson
Bradley John Mccloy
Joseph Daniel Burjoski
Original Assignee
Focus Automation Systems Inc.
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date

Links

Classifications

    • GPHYSICS
    • G06COMPUTING; CALCULATING; COUNTING
    • G06TIMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00Image analysis
    • G06T7/0002Inspection of images, e.g. flaw detection
    • G06T7/0004Industrial image inspection
    • G06T7/001Industrial image inspection using an image reference approach
    • BPERFORMING OPERATIONS; TRANSPORTING
    • B41PRINTING; LINING MACHINES; TYPEWRITERS; STAMPS
    • B41FPRINTING MACHINES OR PRESSES
    • B41F33/00Indicating, counting, warning, control, or safety devices
    • B41F33/0036Devices for scanning or checking the printed matter for quality control
    • GPHYSICS
    • G06COMPUTING; CALCULATING; COUNTING
    • G06TIMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00Indexing scheme for image analysis or image enhancement
    • G06T2207/30Subject of image; Context of image processing
    • G06T2207/30108Industrial image inspection
    • G06T2207/30141Printed circuit board [PCB]
    • GPHYSICS
    • G06COMPUTING; CALCULATING; COUNTING
    • G06TIMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00Indexing scheme for image analysis or image enhancement
    • G06T2207/30Subject of image; Context of image processing
    • G06T2207/30108Industrial image inspection
    • G06T2207/30148Semiconductor; IC; Wafer

Abstract

A method, apparatus and system for verifying the establishment of a pattern includes means for storing a template of the pattern, means for acquiring an image of the established pattern, a first image processing means to alter at least one of said template and acquired images to produce at least two resultant images, comparison means to compare the two resultant images with the other of said template and acquired image and means to evaluate the results of each comparison to determine if the established pattern includes a defect. Both gray scale and binary processing and comparisons are disclosed for use as required.

Description

METHOD. APPARATUS AND SYSTEM FOR VERIFICATION OF PATTERNS Field Of The Invention

The present invention relates to a method, apparatus and system for the automatic verification of patterns established on a substrate or other surface. More specifically, the present invention relates to a system and method for determining if an established pattern contains one or more defects. Background Of The Invention

In many fields, such a printing, it is desired to determine that a pattern has been properly formed on the substrate or other target surface. For example, when printing labels for pharmaceuticals, it is desirable to insure that dosage information has been properly printed on the label. Specifically, a defect such as one which resulted in one or more labels having a missing decimal point which would otherwise change 1.0 mg to 10 mg, with perhaps serious and/or fatal results, is unacceptable. Similarly, for any high quality printing such as postage stamps or security printing for lottery tickets, etc., it is important that the printed pattern be correct within preselected bounds. This is also true for other patterns such as printed circuit boards, semiconductor device features, flexible printed circuits, etc.

Prior art verification of printed patterns has been performed either post production (i.e. - after the pattern establishment process has been completed) or in real-time (i..e. - while the pattern establishment process is being performed). Post production verification has disadvantages in that a great deal of wastage can occur as there is no immediate feedback to the pattern establishing process, to correct the process in view of detected defects. Thus, if for example a process error occurs in the middle of a production half of the production run of finished product may include defects and have to be discarded. Further, post production verification requires an additional, separate step and therefore results in increased labor costs and increases the time required to complete a production run. Real-time verification systems are therefore often preferred over post production systems.

However, prior art real-time verification systems suffer from their own problems in that, when high confidence is required in determining errors, many false indications of defects are produced as no means is available to distinguish between actual defects and acceptable variations due to normal pattern establishment process variations. In some circumstances, when high confidence is not required, the occurrence of false indications of defects in prior art systems has been lowered by reducing the sensitivity of the pattern verification system, but this necessarily results in reduced quality control of the final product.

Further, due to limitations in the speed with which prior art systems could perform verification, such systems often required the limitation of the production speed of the pattern establishment process to less than the available maximum processing speed.

It is desired therefore, to have a pattern verification system which can verify the establishment of patterns to a desired degree of accuracy, accommodate normal process variations without indicating false defects and which can operate with relatively high speed pattern establishment processes. Summary Of The Invention

It is an object of the present invention to provide a novel method, apparatus and system for determining if an established pattern contains, a defect which obviates or mitigates at least one of the disadvantages of the prior art.

According to a first aspect of the present invention, there is provided a method of determining if an established pattern contains one or more defects, comprising the steps of: (i) defining a digital template image representing a desired pattern;

(ii) acquiring a digital image of an established pattern;

(iii) image processing at least one of said defined digital template and said acquired digital image to obtain at least first and second resultant images;

(iv) performing at least one pair of comparisons comprising: a first comparison between the other of said defined template and said acquired digital image and said first resultant image; and a second comparison between the other of said defined template and said acquired digital image and said second resultant image;

(v) evaluating the results of at least said at least one pair of comparisons to determine if said established pattern contains one or more defects. According to another aspect of the invention, there is provided a system to determine if an established pattern contains oneor more defects, comprising: means to store a defined digital template for said pattern; means to acquire a digital image of said established pattern; image processing means to process at least one of said defined digital template and said acquired digital image to obtain at least first and second resultant images; means to perform at least one pair of comparisons comprising a comparison between the other of said defined digital template and said acquired digital image and said first resultant image and a comparison between the other of said defined digital template and said acquired digital image and said second resultant image; and means to evaluate the results of said comparisons to determine if said established pattern contains one or more defects. Brief Description Of The Drawings

Preferred embodiments of the present invention will now be described, by way of example only, with reference to the attached Figures, wherein: Figure 1 shows a block diagram of a prior art pattern verification system;

Figure 2 shows a representation of a template image;

Figure 3 shows a representation of an image of a pattern to be verified;

Figure 4 shows a representation of a prior art subtraction technique to identify defects in a pattern; Figure 5 shows a representation of images of four possible configurations of an example pattern;

Figure 6 shows a representation of a prior art subtraction technique to identify defects;

Figure 7 shows an array of digital values representing an image of a pattern;

Figure 8 shows a representation of a median rank filter being applied to one of the digital values of Figure 7;

Figure 9 shows a representation of the result of the processing the array of Figure 7 with a median rank filter;

Figures 10a through lOd represent possible outcomes of a dark feature missing comparison;

Figures 1 la through lid represent possible outcomes of a light feature missing comparison;

Figures 12a through 12c represent possible outcomes of an extra dark feature comparison; Figures 13a through 13c represent possible outcomes of an extra light feature comparison;

Figure 14 shows a block diagram of a preferred configuration of a gray scale system in accordance with the present invention which employs four comparisons;

Figure 15 shows the system of Figure 14 as implemented in commercially available Cyberscan® modules; Figure 16a shows an example Template image for use in performing a pattern verification with binary comparisons;

Figure 16b shows an example pattern image to be verified with binary comparisons;

Figure 17 shows a graphical representation of the results of a first binary comparison;

Figure 18 shows a graphical representation of the results of a second binary comparison; Figure 19 shows a graphical representation of the results of a third binary comparison;

Figure 20 shows a graphical representation of the results of a fourth binary comparison;

Figure 21 shows a graphical representation of the results of a fifth binary comparison;

Figure 22 shows a graphical representation of the results of a sixth binary comparison;

Figure 23 shows a graphical representation of the results of a seventh binary comparison; Figure 24 shows a graphical representation of the results of an eighth binary comparison;

Figure 25 shows a graphical representation of the results of a ninth binary comparison;

Figure 26 shows a graphical representation of the results of a tenth binary comparison;

Figure 27 shows a block diagram of a preferred configuration of a binary system in accordance with the present invention; and Figure 28 shows the system of Figure 27 as implemented in commercially available

Cyberscan® modules. Detailed Description Of The Invention

For clarity, a prior art pattern verification system will first be described before detailed discussion of the present invention. A pattern verification system is indicated generally at 10 in Figure 1. System 10 comprises a suitable camera 14 which acquires an image of a pattern 18 to be verified. As used herein, the term pattern is intended to comprise any text, schematic, graphics, physical configuration or other features which can be imaged by camera 14. Also, as used herein the spraying, pressing, embossing, etching, lithographing, depositing, mounting or otherwise causing a pattern to be formed is referred to as establishing a pattern. The image from camera 14 is digitized and stored in image digitizer 22 in the form of an array of 8 bit pixels, the value of each pixel representing the degree of darkness in the acquired digitized image with 0 being black and with 255 being white.

In system 10, the desired pattern has been digitized and stored in image memory 26. This stored image of the desired pattern is commonly referred to as a template or 'golden template' as it represents the ideal pattern image against which the acquired image will be compared. A suitable processor 30 then compares, as described below, the template image in image memory 26 to the acquired image stored in digitizer 22 and produces an output to output means 34, which may be a video display, an error annunciator or any other suitable output device as will occur to those of skill in the art.

Specifically, the comparison process proceeds as follows. Figure 2 shows a representation of the template image stored in image memory 26 and Figure 3 shows a representation of an image which has been acquired by camera 14 and digitized and stored in digitizer 22. It will be understood by those of skill in the art that, while these images are shown in the Figures with the pixels being represented with filled blocks, in fact the digitized images are stored as an array of digital values which may be operated upon mathematically by processor 30. A reference fiducial 50 is included in the template image and a corresponding reference fiducial 54 is included in the pattern whose image was acquired by camera 14. These reference fiducials are used to align the images so that they are spatially coincident. In example of Figures 3 and 4, assuming that the upper left hand pixel of each Figure is indexed as (0,0), the center of fiducial 54 is located at (4,2) whereas the center of the fiducial 50 in the template image is located at (3,1). Accordingly, to make the images spatially coincident, the acquired image is shifted left one pixel and upwards one pixel by processor 30.

Once the template and acquired images are spatially coincident, a simple subtraction operation is performed to identify defects. As shown in Figure 4, in an ideal situation any non-zero valued pixels in the result of this operation represent defects in the acquired image. Specifically, the two non-zero pixels 58 in the result are the two defects in the acquired image. These two pixels can be output to output means 34 or used in any other manner to report the detection of defects in the pattern under consideration.

Unfortunately, the prior art system described above has serious drawbacks when it is applied to many typical pattern establishment processes. One of the main problems with such prior art systems is that most processes for establishing a pattern are not perfect and experience some variation in the established pattern which is not considered to be a defect. For example, in printing processes the positioning of various features will often drift, relative to an intended position and/or to each other, due to normal process variations. For example, Figure 5 shows four images including three images 60a, 60b and 60c wherein the positioning of the decimal point has drifted relative to the 1 and 0 features and one image 60d wherein the decimal point has not been printed.

If the above-described prior art system 10 is employed to verify the patterns of images 60b, 60c and 60d, using image 60a as a template, analysis of each of images 60b, 60c and 60d will result in defects being reported to output means 34, even though only image 60d actually contains a defect, i.e. - the missing decimal point. For example, the outcome of the subtraction operation-type analysis of image 60b using image 60a as a template is shown in Figure 6 wherein the resultant 64 indicates multiple false defects 68.

The present invention will now be described with reference to Figures 7 through 27. The present invention is operable to detect defects in the establishment of a pattern while reducing or even eliminating the number of false defects which would be produced due to normal process variations in the establishment of a pattern. In the immediately following discussion, a printing process will be discussed as a representative example of a method of establishing patterns. It will be appreciated however, by those of skill in the art, that the present invention is not limited to use with printed patterns and can in fact be used with a wide range of pattern establishment processes including, but not limited to, lithography, etching, gravure printing, sputtering of layers, mechanical placement of components, etc. In the present invention, features of a pattern are allowed to drift, due to normal process variations, within predefined bounds. However, if the features drift outside of these predefined bounds, or if the features are malformed or missing, or if spurious features are present, a defect will be deemed to have occurred.

In one preferred embodiment which operates with gray scale images, the present invention utilizes four separate comparison operations, described below. When processing speed is of interest, such as in the real-time monitoring of a high speed printing process, these comparison operations may be conducted in parallel. While, as is discussed below, the outcome of a properly selected pair of comparison operations can provide adequate defect detection, it is believed that the outcome of two properly selected pairs of comparison operations provides an additional advantage in detecting defects.

The present invention makes use of known image processing techniques. Specifically, in the presently preferred gray scale embodiment of the instant invention, discussed below with reference to Figures 7 through 15, a form of image processing which is referred to as gray scale moφhology is employed. Further, in the presently preferred embodiment, the gray scale moφhology operations are implemented with rank value filters. As will be apparent to those of skill in the art, the present invention is not limited to rank value filter-based gray scale moφhology, and any other suitable image processing technique and/or implementation may be employed.

In the presently preferred binary embodiment of the instant invention, discussed below with reference to Figures 16 through 28, the preferable image processing technique which is employed is a form of correlation filtering, specifically binary correlation, although any other suitable binary image processing technique may also be employed.

Rank value filters, such as those used in the gray scale moφhology embodiment, operate on each pixel in an image by ranking the values of the pixels in a neighborhood (kernel) centered on the pixel being processed, referred to herein as the target pixel. The size of the neighborhood, also referred to as the size of the kernel, can be defined as R rows by C columns and the resulting value of the target pixel is calculated by taking all of the unprocessed pixel values in the R by C kernel centered on the target pixel and sorting them. The resulting value of the target pixel, in a filter of rank N, is set to the value of the Nth ranked pixel. Depending upon the value selected for the rank N, various filter effects may be obtained, including a minimum, maximum and/or median filter. Figure 7 shows an eight by eight pixel representation of an image which is mostly dark (low

"Valued pixels) and which has a light ' + ' feature centered at (4,3), assuming that the upper left corner is indexed as (0,0), and which also has a few randomly positioned light spots. Figure 8 demonstrates the application of a rank value filter with a three by three kernel and N = five, when applied to a target pixel located at (4,3) in Figure 7. With a three by three kernel, setting N = 5 results in a median rank filter function as the ranked fifth value of the nine values is the median value. As shown, the pixels in the three by three kernel centered on (4,3) range in value from 211 to 0 with the fifth ranked value being 175. Thus, the resulting value for the target pixel is 175.

The complete result of applying the median rank value filter to the image of Figure 7 is shown in Figure 9 wherein it can be seen that the random light spots have been eliminated and the size and lightness (i.e. - the value of the pixels) of the v + ' feature have been reduced, although its shape and positioning have not changed.

With N set to the highest rank, a maximum rank filter is obtained. When applied to a gray scale image, a maximum rank filter results in the erosion (i.e. - thinning out) of dark features and the dilation (i.e. - fattening up) of light features and is referred to herein as a "dilation" or "dilate" operation. Conversely, with N set to the lowest rank, a minimum rank filter is obtained. When applied to a gray scale image, a minimum rank filter results in the dilation of dark features and the erosion of light features and is referred to herein as an "erosion" or "erode" operation.

It will be understood by those of skill in the art that the descriptions above refer to a gray scale value system wherein dark image pixels are low valued and light image pixels are high valued. In the reverse system, wherein dark image pixels are high valued and light image pixels are low valued, the above-described effects of the maximum and minimum rank filters are reversed, i.e. - a minimum rank filter will dilate light features and erode dark features and a maximum rank filter will erode light features and dilate dark features. However, as used herein, the term "erode" or "erosion" refers to the erosion of light features and the term "dilate" or "dilation" refers to the dilation of light features and the opposite effects on dark features. It will also be understood by those of skill in the art that gray scale image pixels need not be 8-bit values and other larger or smaller values, as appropriate to the particular application of the present invention, may be employed as desired.

In the presently preferred gray scale embodiment, the instant invention employs four comparison operations which are presently believed to be advantageous for use in identifying defects. These comparison operations are intended to detect: 'extra dark feature' defects; 'missing dark feature' defects; 'extra light feature' defects; and 'missing light feature' defects. Each comparison will be described in turn below.

The comparison to identify a missing dark feature defect can be represented by LiveER0DE ≤ Template^ + Δs where iveER0DE is the acquired image which has been processed with an appropriate moφhological filter (for example, a minimum rank in the case of light valued pixels having greater values than dark valued pixels), TemplateBAW is the non-processed template image and Δ, is an image intensity threshold which is proportional to the intensity value in the template image. Δi is described in more detail below.

In this comparison, as in all of the other comparisons described below, TemplateRAW can be a template which is predefined and stored in a memory or it can be an image which has been previously acquired. For example, in some circumstances TemplateRA can be the unprocessed live image (LiveRAw) acquired during the last verification operation. As will be apparent to those of skill in the art, in this missing dark feature comparison and all of the other comparisons described below, the size of the kernel defines the bounds about the template position within which a feature must be found or else a defect is detected. For example, a three by three kernel, such as that referred to above with respect to Figure 8, allows a feature to drift up to one pixel in any direction from its position in the template. Therefore, defining the size of the kernel used in the rank filtering will define the desired limits for variance in the established pattern. Of course, it is possible to define different kernel sizes for different regions of the pattern to be verified. For example, in the case of the pharmaceutical label mentioned above, a three by three kernel can be employed when image processing a region of the label image which contains dosage information and a five by five kernel can be employed when image processing the remainder of the label. Further, processing with the three by three kernel may proceed in parallel with processing with the five by five kernel or in seriatim, depending upon the particular hardware implementation.

As will be apparent to those of skill in the art, the kernel need not be square and thus the limits for variance can be defined separately for horizontal and vertical directions. For example, a rectangular kernel can be employed if feature drift is acceptable to a larger extent in one direction (e.g. - 'x' or horizontal) than in the other direction (e.g. - 'y' or vertical). Also, non-rectilinear kernel shapes may be employed. For example, a generally circular kernel can be employed when feature drift in a radial manner is acceptable.

In the representations shown in Figures 10a through 13c, and described below, a spatially corresponding line of pixels in each of the acquired image and the template image, at least one of which has been image processed by a selected moφhological filter, is being compared. In these one dimensional representations of the processing of the two dimensional images, the heights of the traces reflect the value of the respective pixels, with higher traces (values) indicating lighter pixels.

The datums for the two traces are spaced by the value Δ, which allows for a predefined amount of variation in the intensity of the pattern. In a preferred embodiment of the present invention, Δ, varies for pixels within the template. Specifically, a value for Δ, is predetermined for each possible pixel value using a logarithmic function. This simulates the operation of the human eye in that Δj is larger for brighter pixels and smaller for darker pixels. In such a case, a larger change in intensity in bright areas is required than the complementary change in dark areas. Thus, the index 'i' for Δj is merely the corresponding pixel value for which the respective Δ, has been predetermined. As will be apparent to those of skill in the art, other techniques of implementing Δ, can also be favorably employed, including setting Δ, to a constant value, calculating Δ, based upon the values and locations of surrounding pixels, etc.

In these comparisons, an acquired image which exactly corresponds to the template image would be represented by an identical pair of traces, vertically spaced by the value Δ,. A non- identical but defect free pattern is represented by any pair of non-intersecting traces and a pattern with one of more defects is represented by a pair of traces with at least one intersection point. It will Be understood by those of skill in the art, that the representations in these Figures are only intended as a convenient means to convey the operation of the comparisons described herein and that the comparisons are in fact performed mathematically and need not be performed on lines of pixels, but instead can be performed in any sequence provided that pairs of spatially coincident pixels of interest are compared.

Figure 10a shows a representation of the missing dark feature comparison wherein the dark feature is present and no drift has occurred from the template position and no defect is thus detected. This is apparent as there is no intersection of the traces and the lowered area of the LiveER0DE trace, which represents the pixels of the dark feature in the acquired image, is centered about the lowered area of the Template trace, which represents the pixels of the dark feature in the Template. As will be apparent to those of skill in the art, the widening of the lowered area of the iveER0DE trace in the Figure is the result of the erosion of the light features (and the corresponding dilation of the dark feature) by the moφhological filter.

Figure 10b shows a representation of the same comparison wherein the dark feature in the acquired image has drifted from the template position, but by an amount less than half the kernel size and thus no defect is detected.

Figure 10c shows a representation of the same comparison wherein the dark feature in the acquired image has drifted from the template position by an amount greater than half of the kernel size and thus a defect is detected, as indicated by the shaded, overlapped, portion.

Finally, Figure lOd shows a representation of the same comparison wherein the dark feature is missing in the acquired image (no lowered area in the LiveER0DE trace) and thus a defect is detected, as indicated by the shaded, overlapped, portion.

The missing light feature comparison can be represented by

LiveD1LAτE ≥ wherein LiveDiLATE is the acquired image which has been processed by a moφhological filter which dilates the light features, TemplateRAW is the unprocessed template image and Δ, is a delta value, as described above. Figure 11a shows a representation of this comparison wherein the light feature is present in the acquired image and has not drifted and thus no defect is reported. Figure 1 lb shows a representation of this comparison wherein the light feature is present in the acquired image and has drifted from the template position by an amount less than one half the kernel size and thus no defect is reported. Figure 1 lc shows a representation of this comparison wherein the light feature is present in the acquired image and has drifted from the template position by an amount greater than one half the kernel size and thus a defect is reported. Finally, Figure 1 Id shows a representation of this comparison wherein the light feature is missing from the acquired image and thus a defect is reported. The extra dark feature comparison can be represented by

LiveMW ≥ TemplateER0DE - Δ: where LiveRAW is the unprocessed acquired image, TemplateER0DE is the template image to which an appropriate moφhological filter has been applied to erode the light features and Δs is a delta value, as described above. Figure 12a shows a representation of this comparison wherein an expected dark feature is found in the acquired image at the template position and thus no defect is reported. Figure I2b shows a representation of this comparison wherein an extra dark feature is detected in the acquired image and thus a defect is reported. Figure 12c shows a representation of this comparison wherein a dark feature is detected in the acquired image at the template location, but it is too dark (i.e. - black rather than gray) and thus an error is detected. The extra light feature comparison can be represented by

Livel , ≤ Template DllATE + Δ, where LiveRAW is the unprocessed acquired image, TemplateD1LΛTE is the template image to which an appropriate moφhological filter has been applied to dilate the light features and Δf is a delta value, as described above. Figure 13a shows a representation of this comparison wherein an expected light feature is found in the acquired image at the template position and thus no defect is reported. Figure 13b shows a representation of this comparison wherein an extra light feature is detected in the acquired image and thus a defect is reported. Figure 13c shows a representation of this comparison wherein a light feature is detected in the acquired image at the template location, but it is too light (i.e. - white rather than gray) and thus an error is detected.

It is contemplated that other comparisons can also be employed with the present invention to identify categories or types of defects prevalent in other particular pattern establishment processes. For example, the comparisons ^ERODE ≤ TemplateDILλπ + Δj '^DILATE ≥ Template ER0DE - Δ, are similar to the extra dark feature and extra light feature comparisons given above, but allow for a greater degree of drift of the pattern. The selection of other suitable comparisons for use with gray scale embodiments of the present invention will be apparent to those of skill in the art and will not be discussed further herein.

The present inventors have determined that by performing at least one appropriate corresponding pair of comparisons, the instant invention can provide good pattern verification / defect detection performance. For example, as will be apparent to those of skill in the art, for dark patterns established on a light substrate one appropriate corresponding pair is the extra dark feature comparison and the missing dark feature comparison. Once the comparisons have been performed, their results can be combined through a logical OR to identify defects.

Similarly, for light patterns established on a dark background, an example of an appropriate corresponding pair of comparisons is the extra light feature comparison and the missing light feature comparison. Again, once the two comparisons have been performed, their results can be combined through a logical OR to identify defects.

While an appropriate corresponding pair of comparisons provides good pattern verification performance, in a preferred gray scale embodiment of the present invention, shown in Figure 14, four comparisons (two pairs) are employed. Specifically, the extra light feature comparison 100, missing light feature comparison 104, extra dark feature comparison 108 and missing dark feature comparison 112 are all performed. An output 116 indicating Extra Feature Defects can be provided by combining the outputs of extra dark feature comparison 108 and extra light feature comparison 100 through a logical OR 120, as can an output 124 indicating Missing Feature Defects by combining the outputs of missing light feature comparison 104 and missing dark feature comparison 112 through a logical OR 128.

While the above-described combination of four comparisons is presently preferred, it will be apparent to those of skill in the art that alternative configurations can be employed. For example, the four comparisons shown in Figure 14 can be arranged to combine the outputs of extra light feature comparison 100 and missing light feature comparison 104 through a logical OR operation and to combine the output of extra dark feature comparison 108 and missing dark feature comparison 112 through a logical OR to obtain outputs indicating Light Feature Defects and Dark Feature Defects. In some embodiments of the present invention, the Template data will not change during a pattern verification process. Accordingly, the processed Template data (i.e. - TemplateER0DE or TemplateDILATE) may either be derived as required, as described above with reference to Figure 14, or pre-computed once and stored for use as needed in a suitable memory. Also, as will be apparent to those of skill in the art, for clarity the buffers or other associated functions required to process the signals have been omitted from Figure 14. For example, if required the LiveRAW signal can be stored in a FIFO buffer before being transmitted to Extra Dark Feature Comparison 108 so that the TemplateER0 E data and the LiveRA data arrive at Extra Dark Feature Comparison 108 simultaneously.

The implementation of the presently preferred gray scale embodiment of the present invention described above is not particularly limited and implementations in dedicated hardware, general puφose computers and/or digital signal processors are contemplated. The present inventors have implemented the gray scale embodiment of the present invention with various Cyberscan® modules, as shown in Figure 15, which are commercially available from Focus Automation Systems Inc., 3-554 Parkside Drive, Waterloo, Ontario, Canada, N2L 5Z4 and which are described in U.S. Patent 5,434,704. In the Figure, eight Cyberscan® modules are employed, and specifically the system comprises: a CAM-DDC3-00 camera module 300; three BUF- 1000-00 buffer modules 304, 308, 312; a GMF-2000-00 gray scale moφhology module 316; a BMF-8000-00 binary moφhology module 320; a LUT- 1000-00 look up table module 324; and a PCD- 1000-00 display module 328. In this implementation, the LiveRAW image, is acquired by a suitable camera 332 and any defects detected by the system are displayed to an operator on a conventional monitor 336.

It should be noted that one of the perceived advantages of the present invention is its susceptibility to implementation in parallel hardware to provide relatively high speed pattern verification / defect detection. For example, in the embodiment shown in Figure 14, each moφhological filter can be implemented by a separate processor, as can each feature comparison. If additional speed is required, two or more processors can perform subsets of each moφhological filter function or comparison and the depth of the data pipeline can also be increased as desired.

As will be apparent to those of skill in the art, the embodiment of the present invention described above is not limited to gray scale patterns. Color patterns can be verified by performing the above-described process on each of three acquired images, each image having been acquired by a separate camera equipped with a respective appropriate color filter (for example - Red, Blue and Green). Alternatively, a color camera can be employed to acquire a color digitized image in an appropriate format, such as an interleaved bitmap, etc. and an appropriate digital colorspace conversion operation (Hue, Saturation and Brightness (HSV) or Luminance, Chrominance r and Xhrominance b (YCrCb), etc.) can be performed to provide three LiveRAW image data streams.

In some pattern verification applications, it is possible to determine the presence or absence of a feature using a binary template and an acquired binary image. For example, in a printed circuit board verification process, it is possible to arrange the lighting, camera, and optics of the image acquiring system such that, in the acquired image, the conductive (copper) traces appear pure white and the circuit board substrate appears pure black. In such a case, the image can be represented by an array of bits, wherein a bit which is equal to 1 represents a white, foreground, pixel (i.e. - a copper trace) and a bit which is equal to 0 represents a black, background, pixel (i.e. - the substrate). Such images are referred to herein as binary images and are a special sub-class of the more general class of digital images which may have pixels represented by a range of values greater than 0 and 1. In a presently preferred binary embodiment of the present invention, in order to verify an established pattern represented by a binary image, binary correlation is employed as the image processing operation. In a binary correlation process, a kernel of preselected size and shape is centered over the target pixel and the value of all of the image pixels within the kernel are summed. This sum is then compared to a predefined threshold value to determine the new value for the target pixel with the target pixel being set to 1 if the threshold is exceeded and 0 if it is not. If, for example, the threshold is set to zero, a binary dilation operation is performed (e.g. - TemplateD1LATE or LiveDiLATE 's obtained). If, on the other hand, the threshold is set to the maximum possible value for the selected kernel (for example a 7x7 kernel has a maximum value of 49), a binary erosion operation is performed (e.g.. - TemplateER0DE or LiveER0DE is obtained). Of course, other threshold values (i.e. - near max, near min, etc.) may be employed as will occur to those of skill in the art and, in the case wherein dark features are represented by 1 's and light features by 0's, the results of the thresholds is reversed. Also, other binary correlation operations exist and any suitable binary correlation operation may be employed as will occur to those of skill in the art.

One of the advantages of binary correlation over the above-described gray scale embodiment of the present invention is that, generally, binary correlation requires less computation than gray scale moφhology and the processors required to implement it are therefore generally less expensive than those required for gray scale moφhology.

In the presently preferred binary embodiment of the present invention, for pattern verification / defect detection puφoses five binary comparisons can be performed, as described below, and each comparison may be performed on the white (foreground) features and on the black (background) features for a total of ten possible comparisons. A description of each comparison operation follows.

In addition to the LiveRAW, LiveD,LΛTE, LiveER0DE, TemplateRAW, TemplateER0DE and TemplateD1LATE data which are similar to those described above in reference to the gray scale embodiment, the binary comparisons described herein can also employ Live0PEN and LiveCL0SE data. The Live0PEN data is obtained by performing an erode correlation operation on the acquired image and then performing a dilate correlation operation on that result. The LiveCL0SE data is obtained by performing a dilate correlation operation on the acquired image and then performing an erode operation on that result. As will be apparent to those of skill in the art, the Live0PEN processing operates to open or 'break apart' two objects joined by an isthmus or the like and the LiveCL0SE processing operates to close or 'fill in' holes in solid objects.

In the presently preferred binary embodiment of the present invention, the LiveERODE and TemplateER0DE images are obtained by performing a maximum (or near maximum, as desired) threshold binary correlation operation on the corresponding RAW signals and the LiveD!LΛTE and TemplateDILATE images are obtained by performing a minimum (or near minimum, as desired) threshold binary correlation operation on the corresponding RAW signals. It will be apparent to those of skill in the art however, that any other suitable binary erode and dilate operations can be employed as will occur to those of skill in the. art.

The ten binary comparisons which can be employed in the presently preferred binary embodiment of the present invention are listed in Appendix A, as are the types of defects for which they are perceived to be useful in detecting. It is presently contemplated that the kernel sizes selected for each comparison would be the same, although it will be apparent to those of skill in the art that this need not be the case and different operations can employ kernels of different sizes as desired and/or appropriate. As will also be apparent to those of skill in the art, in cases wherein the foreground is merely the inverse of the background of an acquired image, some of the ten comparisons will be redundant. In the more general case wherein the foreground and background images may have been acquired with different binarization thresholds, there may be no redundancy of the comparisons.

Verification of patterns / detection of defects with binary comparisons is accomplished in the same manner as the above-described verification with gray scale comparisons, with the comparison data sets being spatially aligned and then processed by appropriate correlation or other filters as required and the selected comparisons being performed. The results of the comparisons are then evaluated to determine if defects are present.

Figure 16a shows an example of a template image for pattern verification with binary comparisons and Figure 16b shows an example of an acquired Live image of an established pattern which is to be verified. As is apparent, the pattern shown in Figure 16b contains numerous defects. Figures 17 through 26 show the defects which will and will not be detected by each of the respective ten comparisons listed in Appendix A.

For example, Figure 17 demonstrates that the "If TemplateER0DE= 1 then Live must = 1 " comparison (labeled (1) in Appendix A) will identify the two horizontal breaks in the right hand vertical copper trace and will detect the pinhole adjacent the bottom of the same trace. However, as is also shown, the other defects, including "neck-down", "extra copper" and "small feature missing" defects are not detected by this comparison. Figures 18 through 26 respectively show examples of the defects detected by the comparisons labeled (2) through (10) in Appendix A.

It will be apparent to those of skill in the art that not all ten binary comparisons need be performed, nor do the comparisons need to be performed in a particular order. For example, as mentioned above, some of the binary comparisons are redundant when the background of the acquired image is merely the inverse of the foreground. Also, some pattern establishment processes are not subject to all types of defects. Accordingly, it is contemplated that the selection of appropriate binary comparisons and the order in which they are conducted will be readily ascertained by those employing the present invention. Figure 27 shows a block diagram of an implementation of the presently preferred binary embodiment of the instant invention. As shown in the Figure, the LiveRAW image 350 from an appropriate camera 354 is transmitted to each of two binary correlation filters 358 and 362 and directly to the processor 366. Binary correlation filter 358 produces the LiveER0DE data which is transmitted to Processor 366 and to binary correlation filter 370 which performs a dilate operation to produce the Live0PEN data which is also transmitted to Processor 366. Binary correlation filter 362 produces the LiveD1LATE data which is transmitted to Processor 366 and to binary correlation filter 374 which performs an erode operation to produce the LiveCL0SE data which is also transmitted to Processor 366.

The TemplateRAw data 378, which can either be stored in memory or obtained from a previous verification operation, is transmitted to each of two binary correlation filters 382 and 386 5 and directly to Processor 366. Binary correlation filter 382 produces the TemplategRODE data and binary correlation filter 386 produces the TemplateDILATE data, each of which are also transmitted to Processor 366. Processor 366 performs the selected comparisons, combines the results of those comparisons and generates an appropriate defect detection output 390.

As will be apparent, for clarity Figure 27 omits the buffers and other related functions which

10 may be required to completely implement the system. However, Figure 28 shows an embodiment of the system of Figure 27 which has been constructed from commercially available Cyberscan® modules. In this Figure, eight Cyberscan® modules are employed, and specifically the system comprises: a CAM-DDC3-00 camera module 400; two BUF-1000-00 buffer modules 404, 408; a BMF-8000-00 binary moφhology module 412; a LUT-1000-00 look up table module 416; a BLB-

15 1000-00 blob module 420; and a PCD-1000-00 display module 424. In this implementation, the LiveRAW image, is acquired by a suitable camera 428 and any defects detected by the system are displayed to an operator on a conventional monitor 432. Output 436 may be used to provide detected defect results to a host computer via, for example, an RS-232 serial interface.

The above-described embodiments are intended to be examples of the present invention and 0 alterations and modifications may be effected thereto, by those of skill in the art, without departing from the scope of the invention which is defined solely by the claims appended hereto.

APPENDIX A

5 (1) If TemplateER0DE= 1 , then LiveRAW must = 1. Used for detecting open defects in foreground, i.e. - pinholes or breaks in copper. Illustrated in Figure 17.

(2) If TemplateDILATE= 0, then LiveRAW must = 0. Used for detecting open defects in background, i.e. - extra copper in a void, a short circuit. Illustrated in Figure 18.

(3) If LiveRAW= 1 , then TemplateDILATE must = 1. Used for detecting open defects in 0 background, i.e. - extra copper in a void, a short circuit. Illustrated in Figure 19.

(4) If LiveRAW= 0, then TemplateER0DE must = 0. Used for detecting open defects in foreground, i.e. - pinholes in copper. Illustrated in Figure 20.

(5) If LiveεRQDE 111 1. tnen TemplateRAW must = 1. Used for detecting relatively large open defects in background, i.e. - large areas of extra copper, big short circuits. Illustrated in Figure 21. 5 (6) If LiveDn.ATE= 0, then TemplateRAW must = 0. Used for detecting relatively large open ^defects in foreground and small missing foreground features, i.e. - large copper voids or missing small paths. Illustrated in Figure 22.

(7) If TemplateRAW = 1 , then Live_ATE must = 1. Used for detecting small missing foreground features, i.e. - copper pad missing. Illustrated in Figure 23. 0 (8) If TemplateRAW= 0, then LiveER0DE must = 0. Used for detecting small missing background features, i.e. - a drill hole in copper is missing. Illustrated in Figure 24. (9) If TemplateER0DE= 1 , then Live0PEN must = 1. Used for detecting narrowing of foreground features, i.e. - copper trace narrows, but is not broken. Often referred to as ' neck-down' . Illustrated in Figure 25.

(10) If TemplateD|LATE= 0, then LiveCL0SE must = 0. Used for detecting narrowing of background features, i.e. - copper trace widens, but is not shorted. Often referred to as 'projection' Illustrated in Figure 26.

Claims

We claim:
1. A method of determining if an established pattern contains one or more defects, comprising the steps of:
(i) defining a digital template image representing a desired pattern;
(ii) acquiring a digital image of an established pattern;
(iii) image processing at least one of said defined digital template and said acquired digital image to obtain at least first and second resultant images;
(iv) performing at least one pair of comparisons comprising: a first comparison between the other of said defined template and said acquired digital image and said first resultant image; and a second comparison between the other of said defined template and said acquired digital image and said second resultant image;
(v) evaluating the results of at least said at least one pair of comparisons to determine if said established pattern contains one or more defects.
2. The method of claim 1 wherein said image processing in step (iii) comprises moφhological image filter operations.
3. The method of claim 2 wherein said processing in step (iii) said first resultant image is produced by an erode filter operation and said second resultant image is produced by a dilate filter operation.
4. The method of claim 3 wherein said erode filter operation and said dilate filter operation are each implemented as rank value filters.
5. The method of claim 1 wherein at least step (iv) is performed on a pixel by pixel basis.
6. The method of claim 1 wherein each of said pair of comparisons in step (iv) are performed substantially in parallel.
7. The method of claim 3 wherein: step (iii) comprises processing said digital template with each of an erode filter operation and a dilate filter operation to produce a resultant eroded template image and a resultant dilated template image and processing said acquired digital image with each of an erode filter operation and a dilate filter operation to produce a resultant eroded acquired image and a resultant dilated acquired image; step (iv) comprises performing a comparison between said acquired digital image and said resultant eroded template image, a comparison between said acquired digital image and said resultant dilated template image, a comparison between said defined template image and said resultant dilated acquired image and a comparison between said defined template image and said resultant eroded acquired image; and step (v) comprises evaluating the results of the four comparisons performed in step (iv) to determine if said established pattern contains one or more defects.
*8. The method of claim 7 wherein step (v) further comprises combining the results of the comparison between said acquired digital image and said resultant eroded template image and the results of the comparison between said acquired digital image said resultant dilated template image with a logical OR operation to determine if the established pattern contains at least one of an extra light feature or an extra dark feature.
9. The method of claim 7 wherein step (v) further comprises combining the results of the comparison between said defined template image and said resultant dilated acquired image and between said defined template image and said resultant eroded acquired image with a logical OR operation to determine if the established pattern is missing at least one of a light feature or a dark feature.
10. The method of claim 8 wherein step (v) further comprises combining the results of the comparison between said defined template image and said resultant dilated acquired image and between said defined template image and said resultant eroded acquired image with a logical OR operation to determine if the established pattern is missing at least one of a light feature or a dark feature.
11. The method of claim 1 wherein said at least first and second resultant images are stored in a memory means.
12. The method of claim 7 wherein said resultant dilated template image and said resultant eroded template image are stored in a memory means.
13. The method of claim 1 wherein said digital template and said acquired digital image are binary images.
14. The method of claim 13 wherein said image processing in step (iii) comprises binary correlation image filter operations.
15. The method of claim 14 wherein step (iii) comprises image processing at least one of said defined digital template and said acquired image with at least one of a dilate filter and an erode filter.
16. The method of claim 15 wherein step (iii) further comprises image processing said acquired image with at least one of a series combination of a dilate filter and an erode filter and a series combination of an erode filter and a dilate filter.
17. The method of claim 15 wherein step (iii) comprises each of said defined digital template and said acquired image being image processed with at least one of a dilate filter and an erode filter to obtain first and second resultant acquired images and first and second resultant template images and wherein step (iv) comprises performing at least one comparison between each of said first and second resultant acquired images and said defined template and performing at least one comparison between each of said first and second resultant template images and said acquired image.
18. The method of claim 1 wherein said pair of comparisons of step (iv) are performed substantially simultaneously.
19. The method of claim 7 wherein said comparisons of step (iv) are performed substantially simultaneously.
20. The method of claim 17 wherein said comparisons of step (iv) are performed substantially simultaneously.
21. The method of claim 1 wherein said template image is defined by acquiring a suitable image with a camera.
22. A system to determine if an established pattern contains one or more defects, comprising: means to store a defined digital template for said pattern; means to acquire a digital image of said established pattern; image processing means to process at least one of said defined digital template and said acquired digital image to obtain at least first and second resultant images; means to perform at least one pair of comparisons comprising a comparison between the other of said defined digital template and said acquired digital image and said first resultant image and a comparison between the other of said defined digital template and said acquired digital image and said second resultant image; and means to evaluate the results of said comparisons to determine if said established pattern contains one or more defects.
23. The system of claim 22 wherein said image processing means comprises at least two moφhological image filters.
24. The system of claim 23 wherein one of said at least two moφhological image filters is an erode filter and wherein another of said at least two moφhological filters is a dilate filter.
25. The system of claim 24 wherein each of said erode filter and said dilate filter are implemented as rank value filters.
26. The system of claim 25 wherein said image processing means comprises at least two binary correlation filters.
27. The system of claim 26 wherein said image processing means comprises at least two binary correlation filters.
28. The system of claim 27 wherein at least one of said at least two binary correlation filters is a dilate filter and wherein another of said at least two binary correlation filters is an erode filter.
29. The system of claim 28 wherein said image processing means further comprises at least one of a series combination of an erode and a dilate filter and a series combination of a dilate and an erode filter.
30. The system of claim 22 wherein said image processing means processes said defined digital template and said first and second resultant images are stored in a memory.
31. The system of claim 22 wherein said defined digital template is a previously acquired digital image.
32. The system of claim 22 wherein said image processing means processes said at least one of said defined digital template and said acquired digital image in parallel to obtain substantially simultaneously said at least first and second resultant images.
PCT/CA1997/000196 1996-03-25 1997-03-25 Method, apparatus and system for verification of patterns WO1997036260A1 (en)

Priority Applications (2)

Application Number Priority Date Filing Date Title
US08/622,193 1996-03-25
US08622193 US5848189A (en) 1996-03-25 1996-03-25 Method, apparatus and system for verification of patterns

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CA 2249265 CA2249265C (en) 1996-03-25 1997-03-25 Method, apparatus and system for verification of patterns

Publications (1)

Publication Number Publication Date
WO1997036260A1 true true WO1997036260A1 (en) 1997-10-02

Family

ID=24493253

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CA1997/000196 WO1997036260A1 (en) 1996-03-25 1997-03-25 Method, apparatus and system for verification of patterns

Country Status (3)

Country Link
US (1) US5848189A (en)
CA (1) CA2249265C (en)
WO (1) WO1997036260A1 (en)

Cited By (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
EP1014294A2 (en) * 1998-11-30 2000-06-28 Canon Kabushiki Kaisha Method and apparatus for mark detection
US6895108B2 (en) 1998-11-05 2005-05-17 Samsung Electronics Co., Ltd. Method for inspecting defects in the shape of object
WO2014029532A3 (en) * 2012-08-24 2014-04-24 Koenig & Bauer Aktiengesellschaft Method for inspecting at least one copy of a printed product

Families Citing this family (43)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US6259827B1 (en) 1996-03-21 2001-07-10 Cognex Corporation Machine vision methods for enhancing the contrast between an object and its background using multiple on-axis images
US6075881A (en) 1997-03-18 2000-06-13 Cognex Corporation Machine vision methods for identifying collinear sets of points from an image
US6608647B1 (en) 1997-06-24 2003-08-19 Cognex Corporation Methods and apparatus for charge coupled device image acquisition with independent integration and readout
US6282328B1 (en) * 1998-01-28 2001-08-28 Cognex Corporation Machine vision systems and methods for morphological transformation of an image with non-uniform offsets
US6236769B1 (en) * 1998-01-28 2001-05-22 Cognex Corporation Machine vision systems and methods for morphological transformation of an image with zero or other uniform offsets
US6381375B1 (en) 1998-02-20 2002-04-30 Cognex Corporation Methods and apparatus for generating a projection of an image
US6628824B1 (en) 1998-03-20 2003-09-30 Ken Belanger Method and apparatus for image identification and comparison
US7016539B1 (en) 1998-07-13 2006-03-21 Cognex Corporation Method for fast, robust, multi-dimensional pattern recognition
US6630998B1 (en) 1998-08-13 2003-10-07 Acushnet Company Apparatus and method for automated game ball inspection
US6687402B1 (en) 1998-12-18 2004-02-03 Cognex Corporation Machine vision methods and systems for boundary feature comparison of patterns and images
US6381366B1 (en) 1998-12-18 2002-04-30 Cognex Corporation Machine vision methods and system for boundary point-based comparison of patterns and images
US6684402B1 (en) 1999-12-01 2004-01-27 Cognex Technology And Investment Corporation Control methods and apparatus for coupling multiple image acquisition devices to a digital data processor
CA2296143A1 (en) 2000-01-18 2001-07-18 9071 9410 Quebec Inc. Optical inspection system
US6748104B1 (en) 2000-03-24 2004-06-08 Cognex Corporation Methods and apparatus for machine vision inspection using single and multiple templates or patterns
US7167583B1 (en) * 2000-06-28 2007-01-23 Landrex Technologies Co., Ltd. Image processing system for use with inspection systems
US6462812B1 (en) 2000-11-21 2002-10-08 Callaway Golf Company Golf ball indicia verification system
US6603882B2 (en) * 2001-04-12 2003-08-05 Seho Oh Automatic template generation and searching method
US6959112B1 (en) 2001-06-29 2005-10-25 Cognex Technology And Investment Corporation Method for finding a pattern which may fall partially outside an image
US20030021437A1 (en) * 2001-07-11 2003-01-30 Hersch Roger David Images and security documents protected by micro-structures
US7298866B2 (en) * 2001-10-15 2007-11-20 Lockheed Martin Corporation Two dimensional autonomous isotropic detection technique
JP3669698B2 (en) * 2002-09-20 2005-07-13 日東電工株式会社 Inspection method and inspection apparatus of a printed material
US7463765B2 (en) * 2003-02-25 2008-12-09 Lamda-Lite Enterprises Incorporated System and method for detecting and reporting fabrication defects using a multi-variant image analysis
US7017492B2 (en) * 2003-03-10 2006-03-28 Quad/Tech, Inc. Coordinating the functioning of a color control system and a defect detection system for a printing press
US7190834B2 (en) 2003-07-22 2007-03-13 Cognex Technology And Investment Corporation Methods for finding and characterizing a deformed pattern in an image
US8081820B2 (en) 2003-07-22 2011-12-20 Cognex Technology And Investment Corporation Method for partitioning a pattern into optimized sub-patterns
DE10339651B3 (en) * 2003-08-28 2005-03-24 Man Roland Druckmaschinen Ag Determining raster color values by direct recording of different print technology substrates involves determining percentage ratio of number of covered elements to total number of image elements
US7153378B2 (en) * 2003-11-21 2006-12-26 Joe & Samia Management Inc. Product labelling
US20050226466A1 (en) * 2004-04-06 2005-10-13 Quad/Tech, Inc. Image acquisition assembly
JP4518835B2 (en) * 2004-05-13 2010-08-04 大日本スクリーン製造株式会社 Defect detecting device, the wiring area extraction apparatus, a defect detection method and a wiring region extraction method
US7423280B2 (en) 2004-08-09 2008-09-09 Quad/Tech, Inc. Web inspection module including contact image sensors
KR100694597B1 (en) * 2005-07-28 2007-03-13 삼성전자주식회사 Method for inspecting a defect of pattern in semiconductor device
WO2007026360A3 (en) * 2005-09-01 2009-05-22 Camtek Ltd A method and a system for creating a reference image using unknown quality patterns
US8111904B2 (en) 2005-10-07 2012-02-07 Cognex Technology And Investment Corp. Methods and apparatus for practical 3D vision system
US8162584B2 (en) 2006-08-23 2012-04-24 Cognex Corporation Method and apparatus for semiconductor wafer alignment
US7996681B1 (en) * 2007-02-26 2011-08-09 Trend Micro Incorporated Methods for implementing color change frequency as image signature
US8270752B2 (en) * 2009-03-17 2012-09-18 Mitsubishi Electric Research Laboratories, Inc. Depth reconstruction filter for depth coding videos
US8189943B2 (en) * 2009-03-17 2012-05-29 Mitsubishi Electric Research Laboratories, Inc. Method for up-sampling depth images
US9202208B1 (en) 2009-05-15 2015-12-01 Michael Redman Music integration for use with video editing systems and method for automatically licensing the same
US8588506B2 (en) * 2011-04-27 2013-11-19 Eastman Kodak Company Image algorithms to reject undesired image features
US9769430B1 (en) * 2011-06-23 2017-09-19 Gentex Corporation Imager system with median filter and method thereof
JP6049358B2 (en) * 2012-08-31 2016-12-21 キヤノン株式会社 Image processing apparatus, image processing method, and program
US9679224B2 (en) 2013-06-28 2017-06-13 Cognex Corporation Semi-supervised method for training multiple pattern recognition and registration tool models
US9008391B1 (en) * 2013-10-22 2015-04-14 Eyenuk, Inc. Systems and methods for processing retinal images for screening of diseases or abnormalities

Citations (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
EP0054598A1 (en) * 1980-12-18 1982-06-30 International Business Machines Corporation Method for the inspection and automatic sorting of objects with configurations of fixed dimensional tolerances, and device for carrying out the method

Family Cites Families (13)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US4395697A (en) * 1980-08-15 1983-07-26 Environmental Research Institute Of Michigan Off-image detection circuit for an image analyzer
US4395698A (en) * 1980-08-15 1983-07-26 Environmental Research Institute Of Michigan Neighborhood transformation logic circuitry for an image analyzer system
US4484346A (en) * 1980-08-15 1984-11-20 Sternberg Stanley R Neighborhood transformation logic circuitry for an image analyzer system
US4398176A (en) * 1980-08-15 1983-08-09 Environmental Research Institute Of Michigan Image analyzer with common data/instruction bus
US4395700A (en) * 1980-08-15 1983-07-26 Environmental Research Institute Of Michigan Image analyzer with variable line storage
US4546444A (en) * 1983-03-15 1985-10-08 E. I. Du Pont De Nemours And Company Data compression interface having parallel memory architecture
US4641356A (en) * 1984-08-24 1987-02-03 Machine Vision International Corporation Apparatus and method for implementing dilation and erosion transformations in grayscale image processing
US4685144A (en) * 1984-10-29 1987-08-04 Environmental Research Institute Of Michigan Image processing system with transformation detection
DE3703047C1 (en) * 1987-02-03 1988-06-23 Mtu Friedrichshafen Gmbh Schmieroelkanal
US4949172A (en) * 1988-09-26 1990-08-14 Picker International, Inc. Dual-mode TDI/raster-scan television camera system
US4922337B1 (en) * 1988-04-26 1994-05-03 Picker Int Inc Time delay and integration of images using a frame transfer ccd sensor
US5452368A (en) * 1993-08-02 1995-09-19 Motorola, Inc. Method of detecting defects in semiconductor package leads
US5434629A (en) * 1993-12-20 1995-07-18 Focus Automation Systems Inc. Real-time line scan processor

Patent Citations (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
EP0054598A1 (en) * 1980-12-18 1982-06-30 International Business Machines Corporation Method for the inspection and automatic sorting of objects with configurations of fixed dimensional tolerances, and device for carrying out the method

Cited By (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US6895108B2 (en) 1998-11-05 2005-05-17 Samsung Electronics Co., Ltd. Method for inspecting defects in the shape of object
EP1014294A2 (en) * 1998-11-30 2000-06-28 Canon Kabushiki Kaisha Method and apparatus for mark detection
EP1014294A3 (en) * 1998-11-30 2005-02-09 Canon Kabushiki Kaisha Method and apparatus for mark detection
US6952484B1 (en) 1998-11-30 2005-10-04 Canon Kabushiki Kaisha Method and apparatus for mark detection
WO2014029532A3 (en) * 2012-08-24 2014-04-24 Koenig & Bauer Aktiengesellschaft Method for inspecting at least one copy of a printed product
US9256931B2 (en) 2012-08-24 2016-02-09 Koenig & Bauer Ag Method for inspecting at least one copy of a printed product

Also Published As

Publication number Publication date Type
CA2249265A1 (en) 1997-10-02 application
CA2249265C (en) 2000-12-26 grant
US5848189A (en) 1998-12-08 grant

Similar Documents

Publication Publication Date Title
US5774573A (en) Automatic visual inspection system
US5911003A (en) Color pattern evaluation system for randomly oriented articles
US4791586A (en) Method of and apparatus for checking geometry of multi-layer patterns for IC structures
US5027417A (en) Method of and apparatus for inspecting conductive pattern on printed board
US5226095A (en) Method of detecting the position of an object pattern in an image
US6198529B1 (en) Automated inspection system for metallic surfaces
US6549683B1 (en) Method and apparatus for evaluating a scale factor and a rotation angle in image processing
US5440650A (en) Image processing system for inspection of overprinted matter
US4758782A (en) Method and apparatus for inspecting printed circuit board
US6909798B1 (en) Method of erasing repeated patterns and pattern defect inspection device
US4642813A (en) Electro-optical quality control inspection of elements on a product
US20030228050A1 (en) Method for pattern inspection
US6320977B1 (en) Method and apparatus for positional detection using pattern matching process
JPH08320294A (en) Device and method for inspecting defect of pattern to be inspected
EP0443062A1 (en) Device for inspecting quality of printed matter and method thereof
JP2005331929A (en) Image analysis method, image analysis program, and pixel evaluation system therewith
WO2001041068A1 (en) Defect inspecting system
JP2004177139A (en) Support program for preparation of inspection condition data, inspection device, and method of preparing inspection condition data
US20070047801A1 (en) Defect detecting method and defect detecting device
US7031511B2 (en) Hole inspection apparatus and method
US5150423A (en) Method of and device for inspecting pattern of printed circuit board
US5892846A (en) Process and apparatus for reading a dot matrix code marking on an article
US20050232476A1 (en) Image analysis method, image analysis program and pixel evaluation system having the sames
US5568564A (en) Image processing apparatus and method for inspecting defects of enclosures of semiconductor devices
JPH07190739A (en) Appearance inspection method for semi-conductor chip and its device

Legal Events

Date Code Title Description
AK Designated states

Kind code of ref document: A1

Designated state(s): AL AM AT AU AZ BA BB BG BR BY CA CH CN CU CZ DE DK EE ES FI GB GE HU IL IS JP KE KG KP KR KZ LC LK LR LS LT LU LV MD MG MK MN MW MX NO NZ PL PT RO RU SD SE SG SI SK TJ TM TR TT UA UG UZ VN AM AZ BY KG KZ MD RU TJ TM

AL Designated countries for regional patents

Kind code of ref document: A1

Designated state(s): GH KE LS MW SD SZ UG AT BE CH DE DK ES FI FR GB GR IE IT LU MC NL PT SE BF BJ CF CG

DFPE Request for preliminary examination filed prior to expiration of 19th month from priority date (pct application filed before 20040101)
121 Ep: the epo has been informed by wipo that ep was designated in this application
ENP Entry into the national phase in:

Ref country code: CA

Ref document number: 2249265

Kind code of ref document: A

Format of ref document f/p: F

Ref document number: 2249265

Country of ref document: CA

NENP Non-entry into the national phase in:

Ref country code: JP

Ref document number: 97533889

Format of ref document f/p: F

REG Reference to national code

Ref country code: DE

Ref legal event code: 8642

122 Ep: pct application non-entry in european phase