Classifications

• • G—PHYSICS
  • G09—EDUCATION; CRYPTOGRAPHY; DISPLAY; ADVERTISING; SEALS
  • G09G—ARRANGEMENTS OR CIRCUITS FOR CONTROL OF INDICATING DEVICES USING STATIC MEANS TO PRESENT VARIABLE INFORMATION
  • G09G3/00—Control arrangements or circuits, of interest only in connection with visual indicators other than cathode-ray tubes
  • G09G3/006—Electronic inspection or testing of displays and display drivers, e.g. of LED or LCD displays
• • G—PHYSICS
  • G02—OPTICS
  • G02F—OPTICAL DEVICES OR ARRANGEMENTS FOR THE CONTROL OF LIGHT BY MODIFICATION OF THE OPTICAL PROPERTIES OF THE MEDIA OF THE ELEMENTS INVOLVED THEREIN; NON-LINEAR OPTICS; FREQUENCY-CHANGING OF LIGHT; OPTICAL LOGIC ELEMENTS; OPTICAL ANALOGUE/DIGITAL CONVERTERS
  • G02F1/00—Devices or arrangements for the control of the intensity, colour, phase, polarisation or direction of light arriving from an independent light source, e.g. switching, gating or modulating; Non-linear optics
  • G02F1/01—Devices or arrangements for the control of the intensity, colour, phase, polarisation or direction of light arriving from an independent light source, e.g. switching, gating or modulating; Non-linear optics for the control of the intensity, phase, polarisation or colour 
  • G02F1/13—Devices or arrangements for the control of the intensity, colour, phase, polarisation or direction of light arriving from an independent light source, e.g. switching, gating or modulating; Non-linear optics for the control of the intensity, phase, polarisation or colour  based on liquid crystals, e.g. single liquid crystal display cells
  • G02F1/1303—Apparatus specially adapted to the manufacture of LCDs
• • G—PHYSICS
  • G02—OPTICS
  • G02F—OPTICAL DEVICES OR ARRANGEMENTS FOR THE CONTROL OF LIGHT BY MODIFICATION OF THE OPTICAL PROPERTIES OF THE MEDIA OF THE ELEMENTS INVOLVED THEREIN; NON-LINEAR OPTICS; FREQUENCY-CHANGING OF LIGHT; OPTICAL LOGIC ELEMENTS; OPTICAL ANALOGUE/DIGITAL CONVERTERS
  • G02F1/00—Devices or arrangements for the control of the intensity, colour, phase, polarisation or direction of light arriving from an independent light source, e.g. switching, gating or modulating; Non-linear optics
  • G02F1/01—Devices or arrangements for the control of the intensity, colour, phase, polarisation or direction of light arriving from an independent light source, e.g. switching, gating or modulating; Non-linear optics for the control of the intensity, phase, polarisation or colour 
  • G02F1/13—Devices or arrangements for the control of the intensity, colour, phase, polarisation or direction of light arriving from an independent light source, e.g. switching, gating or modulating; Non-linear optics for the control of the intensity, phase, polarisation or colour  based on liquid crystals, e.g. single liquid crystal display cells
  • G02F1/1306—Details
  • G02F1/1309—Repairing; Testing
• • G—PHYSICS
  • G02—OPTICS
  • G02F—OPTICAL DEVICES OR ARRANGEMENTS FOR THE CONTROL OF LIGHT BY MODIFICATION OF THE OPTICAL PROPERTIES OF THE MEDIA OF THE ELEMENTS INVOLVED THEREIN; NON-LINEAR OPTICS; FREQUENCY-CHANGING OF LIGHT; OPTICAL LOGIC ELEMENTS; OPTICAL ANALOGUE/DIGITAL CONVERTERS
  • G02F1/00—Devices or arrangements for the control of the intensity, colour, phase, polarisation or direction of light arriving from an independent light source, e.g. switching, gating or modulating; Non-linear optics
  • G02F1/01—Devices or arrangements for the control of the intensity, colour, phase, polarisation or direction of light arriving from an independent light source, e.g. switching, gating or modulating; Non-linear optics for the control of the intensity, phase, polarisation or colour 
  • G02F1/13—Devices or arrangements for the control of the intensity, colour, phase, polarisation or direction of light arriving from an independent light source, e.g. switching, gating or modulating; Non-linear optics for the control of the intensity, phase, polarisation or colour  based on liquid crystals, e.g. single liquid crystal display cells
  • G02F1/133—Constructional arrangements; Operation of liquid crystal cells; Circuit arrangements
  • G02F1/136—Liquid crystal cells structurally associated with a semi-conducting layer or substrate, e.g. cells forming part of an integrated circuit
  • G02F1/1362—Active matrix addressed cells
  • G02F1/136254—Checking; Testing
• • G—PHYSICS
  • G02—OPTICS
  • G02F—OPTICAL DEVICES OR ARRANGEMENTS FOR THE CONTROL OF LIGHT BY MODIFICATION OF THE OPTICAL PROPERTIES OF THE MEDIA OF THE ELEMENTS INVOLVED THEREIN; NON-LINEAR OPTICS; FREQUENCY-CHANGING OF LIGHT; OPTICAL LOGIC ELEMENTS; OPTICAL ANALOGUE/DIGITAL CONVERTERS
  • G02F1/00—Devices or arrangements for the control of the intensity, colour, phase, polarisation or direction of light arriving from an independent light source, e.g. switching, gating or modulating; Non-linear optics
  • G02F1/01—Devices or arrangements for the control of the intensity, colour, phase, polarisation or direction of light arriving from an independent light source, e.g. switching, gating or modulating; Non-linear optics for the control of the intensity, phase, polarisation or colour 
  • G02F1/13—Devices or arrangements for the control of the intensity, colour, phase, polarisation or direction of light arriving from an independent light source, e.g. switching, gating or modulating; Non-linear optics for the control of the intensity, phase, polarisation or colour  based on liquid crystals, e.g. single liquid crystal display cells
  • G02F1/133—Constructional arrangements; Operation of liquid crystal cells; Circuit arrangements
  • G02F1/136—Liquid crystal cells structurally associated with a semi-conducting layer or substrate, e.g. cells forming part of an integrated circuit
  • G02F1/1362—Active matrix addressed cells
  • G02F1/136259—Repairing; Defects

Definitions

• the present invention relates to methods and apparatus for automatic test inspection.
• the invention is illustrated as an automatic inspection method and apparatus for classification of defects, more specifically for detection of "Mura "-type defects in substrates.
• Substrates include liquid crystal flat panel displays, active matrix displays and the like.
• LCD liquid crystal flat panel displays
• Consumer items such as portable video recorders, pocket televisions, notebook computers, engineering work-stations, high-definition televisions (HDTV), and the like incorporate such displays.
• HDTV high-definition televisions
• test operator performs a variety of visual inspections of each display for defects and accepts or rejects the display based upon the operator's perceptions.
• quality and completeness of the inspection is dependent on the individual test operator, who has been trained using limited samples of displays that are characterized as passing or failing. Accordingly, the inspection results are highly subjective, prone to error, and cannot be used consistently and efficiently to monitor, control, and improve the quality of the various manufacturing processes.
• subjective testing criteria results in a lack of industry wide quality standards.
• Fig. 1 shows the role of an automatic inspection apparatus during the final testing stages of the LCFPD manufacturing processes 10.
• the LCFPD undergoes completion, step 12, first inspection via flat panel inspection apparatus, step 14, module assembly, step 16, second inspection via flat panel inspection apparatus, step 18, shipment, step 20, and incoming inspection via flat panel inspection apparatus, step 22.
• a communication network 24 provides an interface between each of the inspections at the flat panel inspection system and a process control work-station 26.
• LCFPD defects encountered at the final inspection are often pixel defects or wide-area pixel defects (also known as Mura defects).
• Problems in the manufacturing process of the LCFPD often cause Mura defects. Because certain manufacturing problems cause certain types of Mura defects, thus identification and elimination of the manufacturing problems often leads to the reduction of Mura defects during subsequent processing runs.
• the Mura defects are often too difficult for the test operator to identify easily, and to categorize efficiently and cost effectively. Accordingly, no easy way of identifying defect types and categorizing such defects for analysis presently exists.
• a method and apparatus for detecting and classifying a defect such as a Mura defect and the like of a substrate are provided.
• the present method includes a sequence of steps which enhance Mura defects for the purposes of defect analysis.
• a method for detecting Mura defects on a substrate for a flat panel display includes the steps of acquiring an image of at least a portion of the substrate, the portion including a second plurality of pixels, the image including values of pixels in the second plurality of pixels, and enhancing differences in the values of pixels in the second plurality of pixels to form an enhanced image, the enhanced image including enhanced values of pixels in the second plurality of pixels.
• the steps of thresholding the enhanced values of pixels in the second plurality of pixels to form a thresholded image, the thresholded image including thresholded values of pixels in the second plurality of pixels, and identifying a third plurality of pixels forming at least one blob within the portion of the substrate in response to the thresholded image are also included.
• the method also includes comparing values of the third plurality of pixels to values of pixels corresponding to an annular region around the at least one blob, and determining a Mura defect in response to the comparison step.
• a computer system for detecting Mura defects on a substrate for a flat panel display comprises an image acquisition device for acquiring an image of at least a portion of the substrate, the portion including a second plurality of pixels, the image including values of pixels in the second plurality of pixels, and an enhancer for enhancing differences in the values of pixels in the second plurality of pixels to form an enhanced image, the enhanced image including enhanced values of pixels in the second plurality of pixels.
• a thresholder for thresholding the enhanced values of pixels in the second plurality of pixels to form a thresholded image, the thresholded image including thresholded values of pixels in the second plurality of pixels, and an identifier for identifying a third plurality of pixels forming at least one blob within the portion of the substrate in response to the thresholded image are also provided.
• the computer system also comprises a comparator for comparing values of the third plurality of pixels to values of pixels corresponding to an annular region around the at least one blob, and a determiner coupled to the comparator for determining a mura defect.
• a method for detecting Mura defects in a substrate for a liquid crystal display includes the steps of acquiring an image of the substrate, and creating a plurality of subsampled images from the image. The method also includes the steps of determining locations of potentially defective pixels in the substrate in response to locations of pixels in each of the plurality of subsampled images having anomalous values, and determining Mura defects in the substrate in response to the locations of potentially defective pixels determined in the substrate.
• a computer system including a computer program for detecting Mura defects in a substrate for a liquid crystal display, the substrate having a plurality of pixels, comprises a computer-readable memory including code that directs an image acquisition device to acquire an image of the substrate, and code that creates a plurality of subsampled images from the image.
• the computer-readable memory also includes code that determines locations of potentially defective pixels in the substrate in response to locations of pixels in each of the plurality of subsampled images having anomalous values, and code that determines Mura defects in the substrate in response to the locations of potentially defective pixels determined in the substrate.
• FIG. 1 is simplified block diagram of uses for a conventional automatic inspection machine
• Fig. 2 illustrates examples of line Mura defects in a conventional flat panel display
• FIG. 3 illustrates examples of area Mura defects in a conventional flat panel display
• Fig. 4 A is a simplified illustration of an embodiment of an inspection apparatus according to the present invention
• Fig. 4B is a block diagram of a system according to an embodiment of the present invention.
• Fig. 5 is a simplified flow diagram of a Mura detection method according to the present invention.
• Fig. 6 illustrates a simplified flow diagram of a method for detecting line Mura according to a preferred embodiment
• Fig. 6 A illustrates a flow diagram for a preferred embodiment for the step of performing defect specific filtering by the image processor
• Fig. 6B illustrates a flow diagram for a preferred embodiment for the step of thresholding each of the images to create blob(s);
• Fig. 6C illustrates a more detailed flow diagram for the step of creating and analyzing blobs from the thresholded images;
• Fig. 6D illustrates a preferred embodiment of a flow diagram for step defect analysis
• Fig. 7 is a simplified flow diagram of a spot Mura detection method according to the present invention.
• Figs. 8 A - 8C illustrate the process of .defining an annular region
• Fig. 9 illustrates a flow diagram of one embodiment of a post processing. method
• Fig. 10 illustrates an example of an embodiment of post processing on an image.
• Mura defects are defined as areas of illumination (pixels on the substrate) which are different, or anomalous, from the neighborhood surrounding the defect, also termed Patterned Brightness Non-Uniformity (BNU).
• BNU Patterned Brightness Non-Uniformity
• Regions of a substrate often include pixels that appear either brighter or darker than pixels surrounding the region, and are classified as Mura defects when specific contrast threshold limits, BNUs, are reached or exceeded. It is important to note that the boundaries for Mura defects are not always well defined and further, within a Mura defect, the BNU may not be homogenous.
• Pixel defects, in contrast are defined as points of illumination that exceed or are lower than the neighboring pixels of the respective pixels. Pixel defects may include individual pixels, clustered pixels, or line segments of pixels that are obvious when visually inspected.
• the present invention is illustrated by way of example with two types of Mura classes, line Mura defects and area Mura defects, as illustrated by Figs. 2 and 3, respectfully.
• Fig. 2 illustrates typical line Mura defects.
• a line Mura defect is defined as a narrow straight or curved strip of illumination which is different from its neighborhood.
• pixels making-up a line Mura defect have anomalous pixels values, compared to values of pixels surrounding the line Mura defect.
• the defect may originate and terminate anywhere within the substrate and can run across the entire length of the substrate.
• the line Mura defect is classified by the length and width of the strip and the angle of occurrence. Typically, such defects have a length-to- width ration exceeding twenty.
• the two rubbing line Mura types may occur either singly or in groups and are related to the alignment layer rubbing process due to imperfections in the surface of the mechanical rollers used in the rubbing process. b) Irregular Line Mura
• the block boundary line Mura is often caused by imperfect seamless joints of the device blocks.
• Fig. 3 illustrates typical area Mura defects.
• An area Mura defect is defined as groups of illumination (pixels on the substrate) which are different from their neighborhood. In other words, pixels making-up an area Mura defect have anomalous pixel values, compared to values of pixels surrounding the area Mura defect.
• the area Mura defect range in size from approximately six pixels in diameter for spot shaped Mura to approximately 25% of the panel display area.
• the elliptical shaped spot Mura is often caused by cell gap variation or clustering of spacer balls, ii) Cluster type spot, circular in shape.
• the cluster type spot Mura is often caused by electro-static charge built-up on spacer balls.
• the panel edge Mura is often caused by polarizer variations or local bleeding of uncured epoxy board material.
• Irregular shape Mura i) Wavy arched shape ranges from small thick shape to more regular L shape
• the irregular shape Mura is often caused by fiber contamination trapped between the polarizer and the glass cleaning process residue or alignment layer rubbing process.
• Fig. 4 A is a simplified illustration of an embodiment of an inspection apparatus 400 according to the present invention.
• the present invention is preferably embodied as a FIS-250 or FIS-300 machine available from Photon Dynamics, Inc.
• the inspection apparatus includes a flat panel display, such as an LCD panel 402 and the like.
• the LCD panel 404 positions on the slidable table 406, and a hinged frame 408, which is brought down to secure the display panel in place.
• the slidable table 406 allows for easy positioning of the LCD panel in an x- y plane under a camera 412 such as a CCD-type camera and the like.
• the slidable table also allows for the LCD panel to be shifted relative to the camera.
• the camera is mounted onto an x-y plane to shift the camera relative to the LCD panel.
• a flexible ribbon type wire 414 supplies drive signals from the pixel drive circuitry in the test system to conductors on the hinged frame 408.
• the camera 412 is preferably a high resolution camera, and is encased with an upper body 426 of the inspection apparatus.
• a monitor 428, a computer 432, and a keyboard 434 are also shown.
• the inspection apparatus includes a plurality of color filters 436, among other features.
• An example of such an inspection apparatus is in U.S. Application Serial No. 08/394,668 (Attorney Docket No. 14116-35-2), which is hereby incorporated by reference for all purposes.
• the flat panel display includes regularly patterned light emitting areas surrounded by light blocking borders.
• the light emitting areas are electrically addressed and are often referred to as pixels.
• the pixels are spaced equally from each other with opaque borders to form a two dimensional periodic pattern.
• the CCD camera may have a construction similar to the flat panel display.
• Each of the pixels in the camera responds to light by converting an electrical signal (with a voltage) which is proportional to the amount of light that strikes the camera pixel.
• the camera pixel includes a border that does not respond to light.
• Each of the pixels are spaced equally from each other, and also form a two dimensional periodic pattern.
• the pattern of pixels forms discrete sampling points of light intensity that define the image impinging on the CCD camera.
• the interference pattern is a periodic modulation of the image voltage signal created by the CCD camera.
• the period of modulation is a function of the period of the pattern of the CCD pixels and the flat panel pixels.
• the periodic modulation of the image often impedes the ability of an inspection system to detect and characterize real defects that may be present on the flat panel display.
• the real defects also modulate the signal but tend not be periodic in nature. Accordingly, methods for reducing or even eliminating the periodic modulation are often used to ensure accurate detection of real defects.
• the aforementioned Application Serial No. 08/394,668 illustrates selected techniques to reduce and even eliminate the periodic modulation. Fig.
• System 200 includes a monitor 210, a computer 220, a keyboard 230, a mouse, an image sensor 240, and a positioning device 250.
• Computer 220 includes familiar computer components such as a processor 260, and memory storage devices, such as a random access memory (RAM) 270, a disk drive 280, and a system bus 290 interconnecting the above components.
• a network interface device (not shown) can be coupled to system bus 290 to provide system 200 with network access.
• RAM 270 and disk drive 280 are examples of tangible media for storage of computer programs, other types of tangible media include floppy disks, removable hard disks, network servers, optical storage media such as CD-ROMS and bar codes, semiconductor memories such as flash memories, read-only-memories (ROMS), ASICs, and battery-backed volatile memories, and the like.
• the system bus may be a PCI bus, NME bus, or the like.
• Positioning device 250 enables the user to position image sensor 240 relative to a substrate, as was previously described.
• An x-y stepper station is but one example of a well known positioning device.
• System 200 includes a Sun SparcStationTM5, running SolarisTM4.1 operating system from Sun Microsystems, Inc. and proprietary hardware and software available from Photon Dynamics, Incorporated.
• Fig. 4B is representative of but one type of system for embodying the present invention. It will be readily apparent to one of ordinary skill in the art that many system types and configurations are suitable for use in conjunction with the present invention.
• the present invention provides for methods and apparatus for identification and classification of Mura defects from a substrate. It is preferred that identification and classification techniques are based upon the contrast of pixels having "anomalous" values to pixels in a background.
• the contrast is termed the brightness non-uniformity value B ⁇ U.
• a relative brightness non-uniformity for pixels is rated from about 1 to about 5, where 5 represents a higher contrast and 1 represents the lowest contrast.
• a specific embodiment correlates each B ⁇ U value with a difference in percentage of gray scale. (As is well known, the gray scale represents the total number of brightness levels available for example, between an inactive pixel and a completely active pixel.) An example of the correlation is shown by Table 1.
• the BNU values may correlate to the gray scale by way of a different correlation.
• the BNU values may also range from 1 to 10, or 1 to 20, or. another. As the range of BNU values increases, the correlation between the BNU value and the gray scale increases or decreases accordingly.
• the BNU value is a linear relationship with respect to the % of gray scale. Of course, the exact BNU values used and their relationship to the gray scale percentage depends upon the particular application.
• the present identification techniques allow for enhanced identification of defects in substrates in a efficient manner. Based upon this identification, the determination of the manufacturing process step that caused the defect is enhanced. Subsequently, the manufacturing step can be modified to inhibit such defects in future production runs.
• the present invention provides classification of Mura defects based in part, to attributes of the defect.
• Paragraphs A-E below describe examples of types of Mura defects and their specific attributes.
• the specific attributes include Mura defect orientation (paragraph a.), defect location (paragraph b.), defect width (paragraph c), defect length (paragraph d.), BNU value (paragraph e.), and others.
• Center line defects A type of defect often caused by circuit patterning misalignment where an excessive gap exists in a center line panel. a. Horizontal orientation b. Centered in vertical direction of panel c. Width-approximately 50 microns d. Length varies e. BNU: 3.0-4.5
• Alignment layer material Film thickness non-uniformities can cause horizontal and vertical (less common) lines. a. Orientation horizontal or vertical or angled b. Position varies c. Wide with poorly defined edges d. Length varies e. BNU: 1.5-3.0
• Alignment layer rubbing defects Defects are often caused by particle contamination between the rubbing roller and the plate. a. Angled orientation (angle would be known by the user) b. Position varies c. 1-3 mm wide d. Length varies e. BNU: 2.0-3.0
• Alignment layer cleaning defects Defects are usually multiple short wavy lines caused by residue of alignment layer cleaning process. a. Angled orientation (generally follows the alignment layer angle) b. Position varies, however clustering may be an important feature c. 1-3 mm wide d. Short length (5-10mm) e. BNU: 1.5-3.0
• Fiber contamination defects Fiber contamination is usually trapped between the polarizer and the glass. a. No specific orientation b. Position varies c. (50-500) microns wide d. Generally arched shaped e. BNU: 1.5-2.5 B. Spot Mura Patterns 1.
• Elliptical shaped spot The elliptical shaped spot is often caused by cell gap variation or clustering of spacer balls. a. Elliptical shape where ellipticity ratio varies from circle to almost a line b. Varied position on panel c. Major diameter: 3-50mm d. Length defined by diameter e. BNU: 2.0 - 3.5 for cell gap
• Cluster type spot Mura Cluster type spot Mura is often caused by electro-static charge build-up on spacer balls. The result is a high density of circular brightness non-uniformities.
• a. Circular shape b. Covers up to 25% of panel with a high density c.
• Diameter 1-3 mm
• BNU 1.0-2.5 (high density of circular brightness non ⁇ uniformities)
• Line type brightness non-uniformity Line type brightness non- uniformity Muras are multiple lines that arch away from the fill port, caused by contamination of the liquid crystal material. a. Arched shaped lines b. Positioned at the fill port c. Line width: 1-5 mm d. Length: 25 mm e. BNU: 2.0-3.0
• a spot type fill port Mura is an elliptical shaped brightness non-uniformity positioned at the fill port. a. Elliptical shape b. Positioned at the fill port c. Major diameter: 5-15 mm d. Length defined by diameter e. BNU: 3.0-4.5 3.
• Arched area An arched area defects are solid arched areas on both sides of the fill port, a. Area type BNU with an arched shape b. Positioned at the fill port c. Approximately 25% of the panel area d. Length defined by occupied panel area e. BNU: 2.0-3.0
• a panel edge Mura is a brightness non-uniformity Mura located around the entire perimeter of the panel active area, typically caused by polarizer variations.
• a block Mura is a large rectangular area brightness non- uniformity, caused by a faulty row or column driver.
• the BNU can be either solid (often caused by a complete driver failure) or
• Width defined by a segment d.
• Length defined by a segment e. BNU: Solid: 4.5-5.0
• Fig. 5 is a simplified flow diagram 500 of a Mura detection method according to the present invention.
• Flow diagram 500 includes steps 520-550.
• an image (frame of data) of a substrate is acquired by an image acquisition device or retrieved from computer memory, step 520.
• Typical image acquisition devices include CCD cameras, line scan camera, frame store cameras, and the like. Examples of preferred image acquisition techniques may be found in Application Serial No. 08/394,668 (Attorney Docket No. 14116-35-2), which is hereby incorporated by reference for all purposes, and assigned to the present assignee.
• an image of the substrate previously acquired may be retrieved from computer memory.
• the present invention provides methods for detection of pixels in the image of the substrate having brightness non-uniformities, step 530.
• a step of the analysis for characterization of the Mura defect is then performed, step 540.
• identification of particular steps in the manufacturing process, which cause the Mura detects is enhanced, step 550. Details of the line Mura detection method are illustrated in Figs. 6 to 6D, and a spot Mura detection method is illustrated in Fig. 7.
• Fig. 6 illustrates a simplified flow diagram 600 of a method for detecting line Mura according to a preferred embodiment.
• Flow diagram 600 includes steps 610-670.
• An outline of embodiments of the present method for Mura detection is as follows.
• the sampling frequency is dependent on the defect size.
• the determining factor is often the line width
• the combination rule preserves a maximum value at each convolved image location and provides a completely enhanced image.
• Each OOI processing region covers a substantially unique range of image pixel values. 3. Threshold of each OOI independently based on its mean and standard deviation
• Each OOI is now a binary representation relative to its background after thresholding.
• Blob Analysis Label original binary blob(s) (typically by way of scanning the display from left to right, and then top to bottom, and combinations thereof)
• a first step in line Mura detection includes acquiring an image of the substrate or retrieving the image from memory, step 610.
• the image(s) are acquired (or captured) by one or more of selected image acquisition devices and techniques known in the art.
• the particular technique used depends upon the application.
• the present invention also does not limit the type of image acquisition technique to this image acquisition technique or others.
• an image processor After image acquisition, an image processor produces a plurality of sub ⁇ sampled images, step 620.
• the subsampled images may be directly subsampled from the original image or be subsampled from a previous subsampled image.
• the selection of the number and ratio for sub-sampling the original image is based upon defect spatial frequency considerations. For example, if a two- dimension image is sub-sampled four square-pixels to one new pixel, effectively the spatial frequency of the new sub-sampled image is one half the original frequency.
• the sub-sampling frequency is generally dependent on the size of the defect or the width of a defect line the user wishes to detect.
• steps 630-670 operate upon the original image, then operate upon a first sub-sampled image, then operate upon a second sub-sampled image, etc.
• step 670 data from each of the images operated upon are combined, as will be discussed.
• Fig. 6 A illustrates a flow diagram for a preferred embodiment for the step of performing defect specific filtering by the image processor.
• Fig. 6A includes steps 632-636.
• frequency filtering of an image can be performed in the time domain by convolving the image with convolution kernels, or alternatively in the frequency domain by multiplying the fourier transform of the image with an image of a filter.
• convolutions in the time domain are preferred.
• the step of performing defect specific filtering begins by defining directional-specific Laplacian kernels, step 632.
• the Laplacian kernels are directionally oriented beginning at about 0 degrees and increment approximately every 15 degrees to about 165 degrees. In this case twelve kernels are defined. A 0 degree and approximately a 75 degree 4x4 Laplacian kernel are illustrated below.
• kernel sizes and increment degrees are envisioned.
• the smallest angle increment often depends upon the processing capabilities and kernel size of the particular system.
• smaller angle increments such as single degree increments are envisioned as technology progresses.
• each image is individually convolved with each of the respective directional kernels, step 636, thus forming a plurality of filtered images. Because each of the Laplacian filters above are directional, each filtered image enhances edges (differences in values of adjacent pixels) oriented in that specific direction within the original image. In the example above, twelve such filtered images are produced.
• step 636 the filtered images are combined to produce an enhanced image, step 636, by way of the combination rule.
• This sequence of steps highlights the edges (differences in intensity values) of features in each image.
• Fig. 6B illustrates a flow diagram for a preferred embodiment for the step of thresholding each of the images to create blob(s).
• Fig. 6B includes steps 642- 646.
• image thresholding begins with the step of forming a histogram representative of the enhanced images, step 642, typically after step 630.
• a histogram represents the brightness of pixels in an image, typically in the form of a gray scale versus number of pixels. The histogram often defines an x-axis representing brightness, and a y- axis representing the number of pixels.
• OI object-of-interest
• Each OOI typically includes areas of the image including clusters of pixels having anomalous intensities.
• Each OOI processing region often represents particular image features such as line Mura defects, spot mura defects, background illuminations, and others.
• a substantially unique range of pixel values in the histogram preferably identifies each OOI.
• each OOI in the image is binarized, based upon a pixel threshold, step 646, to form thresholded images.
• the pixel threshold is preferably determined by the mean and standard deviation of each respective OOI.
• a threshold value for each enhanced image is determined in response to the mean and standard deviation for the entire, respective enhanced image.
• the threshold is set to be from two to three times the standard deviation away from the mean of the image. Threshold values can be set differently to locate pixels that are brighter than other pixels on the image, and to locate pixels that are darker than other pixels on the image.
• Fig. 6C illustrates a more detailed flow diagram for the step of creating and analyzing blobs from the thresholded images.
• Fig. 6C includes steps 651-659.
• blobs are determined and labeled from each threshold image, step 651.
• Blobs are represented as binary representations of the OOI processing region relative to background values.
• a blob may be defined as a completely black image against white background illumination.
• the blob may be represented by a completely white image against a black background illumination. Blobs are determined typically from each OOI processing region.
• the labelling step 651 assigns a label(s) (typically in numerical form) to each of the binary blob(s).
• the label is preferably a unique label such as a number, a letter, or any other character in increasing order or the like.
• An image processor often scans the blob images from left-to-right, and from top-to-bottom, and the like, or any combinations thereof to identify and label each of the binary blob(s).
• the labelling step identifies (or earmarks) and stores each blob for further analysis. In some embodiments, the labelling step may not be necessary but is preferable.
• the embodiment provides for calculating physical statistics about each blob detected, step 652.
• a labelled binary blob acts as a process mask (because of similar size, shape, but of constant gray level) to measure selected statistics of the portion of the original image corresponding to the blob area.
• selected statistics such as range, mean, and standard deviation for the original image within the blob area are easily obtained by calculations.
• the selected statistics define the actual characteristics for the blob within the original image.
• the blobs are dilated, step 653 and an "annular" region around each blob is determined, step 654. This region is typically along the perimeter of the blob.
• an XOR (Exclusive OR) operation of the dilated labelled binary blob with the labelled binary blob defines the annular region.
• Other techniques may also be used to define the annular region surrounding the periphery of the labelled binary blob, such as a "Top Hat" or closing algorithm. Of course, the particular technique often depends upon the application.
• Figs. 8A-8C illustrate the process of defining an annular region. As illustrated in Fig.
• blob 800 in threshold image 810 is identified, as illustrated in Fig. 8A.
• blob 800 is dilated to form blob 820, as illustrated in Fig. 8b.
• the dilation step uses a spatial convolution of the labelled binary blob image.
• a spatial convolution may be carried out by use the following convolution kernel, with a subsequent threshold level of one. 1 1 1
• blob 800 and 820 are then Exclusive-ORed (XOR) to form annular region 830.
• pixels in the original image corresponding to where an annular region of a blob has been found are determined, step 655.
• Statistics for these pixels are then calculated The annular region is considered representative of a selected background region of the blob from the original image.
• statistics for the blob area within the original image are compared to the annular area within the original image, step 656. The statistics determine whether a Mura defect is present.
• the comparing step identifies a Mura defect by way of brightness nonuniformity (BNU) and obtains a percentage difference.
• BNU brightness nonuniformity
• the mean value may be supplemented by a standard deviation. If the comparison falls under user selected criteria, no Mura defect is returned, step 657.
• line specific attributes such as orientation, position, width, length, BNU, end point coordinates, and others are calculated in step 657. These parameters are subsequently stored in a data file or the like, step 658. Based upon previously stored data, the line specific attributes may later be used to identify the particular type of line Mura defect and potential sources (or causes) thereto, step 659.
• Fig. 6D illustrates a preferred embodiment of a flow diagram for step 600 defect analysis.
• Fig. 6D includes steps 661-665.
• defect analysis for line Mura defects includes identification of a defect position on the panel 661, a defect angle 662, a defect length and width 663, a line Mura curvature 664, a BNU 665, and the like.
• the detection of the defect position on the panel identifies whether a line Mura is positioned along a driver block boundary, or positioned in a center of the panel.
• the detection of the defect angle identifies line Mura angle or orientation as 0 degree, 90 degrees, or an angle which coincides with a rubbing angle of the panel.
• a defect line Mura width is also detected to identify potential defect sources.
• a narrow line Mura is often a candidate for an alignment layer rubbing particle defect.
• a wide line Mura is often a candidate for an alignment layer rubbing roller/pressure non-uniformity.
• a line curvature, a line length, or any other line dimension for each defect line Mura is detected. Lines with high . curvature are often attributable to fiber contamination. Other attributes of the defect line Mura were disclosed by the Line Mura Pattern section, above.
• the image processor performs selective post-processing operations, step 670.
• the selective post-processing operations include steps to eliminate false detection of pixel line defects by the line Mura detection steps, steps to eliminate duplicate detection of the same Mura defect found at two or more sub-sampling rates, and to remove individual pixels defects.
• steps to eliminate false detection of pixel line defects by the line Mura detection steps steps to eliminate duplicate detection of the same Mura defect found at two or more sub-sampling rates, and to remove individual pixels defects.
• other post-processing operations may be performed, and the present invention is not limited to the described post-processing operations. Further details regarding post processing will be discussed in conjunction with the spot Mura section below.
• FIG. 7 illustrates a simplified flow diagram 700 of a method for detecting spot Mura according to a preferred embodiment.
• Flow diagram 700 includes steps
• the detection for line Mura defects generally includes steps of image acquisition, step 710, sub-sampling the image, step 720, performing defect specific filtering, step 730, image thresholding to create blobs, step 740, blob analysis, step 750, defect analysis 760, selective post processing 770, and others.
• spot Mura detection may also be briefly outlined as follows. I. Spot Mura Detection Method
• Image Acquisition Acquire original image(s) from a flat panel display (FPD) using any combination of the methods described in this document
• the determining factor is often spot width.
• a first step for spot Mura detection method includes a step of image acquisition, step 710. This step is typically performed according to the description in conjunction with step 610. After image acquisition, an image processor produces a plurality of sub- sampled images, step 720. This step is typically performed according to the description in conjunction with step 620. The smallest spot dimension ranges from about 5 pixels to about 7 pixels.
• steps 730-770 operate upon the original image, then operate upon a first sub-sampled image, then operate upon a second sub ⁇ sampled image, etc.
• step 770 data from each of the images operated upon are combined, as will be discussed.
• the step of performing defect specific filtering begins by defining omni-directional Laplacian kernels.
• convolution kernels for enhancing images may also be used depending upon the application.
• An exemplary convolution kernel is as follows: -1 -1 -1 -1 -1
• frequency filtering of an image can be performed in the time domain by convolving the image with convolution kernels, or alternatively in the frequency domain by multiplying the fourier transform of the image with an image of the filter.
• convolutions in the time domain are preferred.
• the images are passed through a threshold to form thresholded images, step 740.
• blobs are identified and characterized, step 750.
• Mura defects are then determined in response to the blob, step 760, identified in step 750. This step preferably occurs in a similar manner as described in the line Mura defect case.
• the method then includes a post processing step 770.
• One aspect of post processing is to separate individual pixel defects and line defects from area Mura defects. This step enables the user to focus attention upon area Mura defects and ignore other types of defects.
• the pixel and line defects may be removed by way of techniques known in the art, such as low-pass filtering.
• Post processing enhances the user's ability to detect Mura defects, as disclosed in both the line and spot Mura detection sections, above.
• Fig. 9 illustrates a flow diagram of one embodiment of a post processing method.
• Fig. 9 includes steps 790-820.
• each of the thresholded subsampled images is filtered to remove pixel defects, step 860.
• the filter is a low pass filter.
• each of the thresholded subsampled images may be passed through an erosion filter. The effect of step 860 is to eliminate individual pixel defects from consideration as a Mura defect.
• Fig. 10 illustrates an example of an embodiment of post processing on an image.
• Fig. 10 includes images 900, 910, and 920.
• Image 900 includes Mura defect 930
• image 910 includes Mura defect 940
• image 920 includes Mura defect 950.
• Fig. 10 also includes filtered images 960, 970, and 980 and cross section 990, 1000, and 1010.
• Cross section 990 includes potential defects 1020 and 1030
• cross section 1000 includes potential defects 1040 and 1050
• cross section 1010 includes potential defect 1060.
• image 900 illustrates an acquired image of a substrate.
• images 910 and 920 are formed and represent subsampled images from image 900.
• Mura defect 930 in image 900 appears, subsampled, as Mura defect 940 in image 910 and appears as Mura defect 950 in image 920.
• each of the images 900-920 are filtered using edge detection techniques.
• filtered images 960-980 correspond to edges of objects within images 900-920.
• cross sections 990-1010 correspond to cross-sections of intensity values versus position in each of the filtered images 960-980.
• Each cross sections 990-1010 are scaled to the same size.
• two potential defects 1020 and 1030 are shown. Potential defects 1020 and 1030 correspond to the edges of Mura defect 870.
• two potential defects 1040 and 1050 are shown, corresponding to the edges of Mura defect 940.
• potential defect 1060 is shown, corresponding to Mura defect 950. Because image 920 is formed by subsampling image 900, Mura defect 950 appears as only one potential defect 950 in filtered image 980.
• images 1070-1100 are detected blobs and images 1110-1130 are corresponding annular regions.
• the preferred embodiment of the present invention cross references potential defects 1020-1060 with respect to each other looking for overlapping potential defects between images. For example, it can be seen potential defect 1020 overlaps with potential defect 1040, and potential defect 1030 overlaps with potential defect 1050. Further potential defects 1060 overlaps with a potential, defect 1050, and potential defect 1050 overlaps with potential defect 1030. Because the size of the Mura defect decreases with successive sub-sampling, any Mura defect present in the original image appears smaller in each successive subsample.
• Overlapping potential defects between successive subsampled images therefore indicates that the potential Mura defects map onto the same Mura defect.
• potential defects 1020-1060 all report the same Mura defect 930 in the acquired image.
• overlap is illustrated along a row of the image.
• overlap along a column of the image is analyzed, or overlap in any direction (omnidirectional) is analyzed.
• Fig. 10 includes Mura defects 1070 and 1080 in image 900 and 1090 in image 910.
• Mura defects 1070 and 1080 are part of one line Mura defect.
• image 910 a result of subsampling image 900, only one defect, Mura defect 1090, appears.
• the user determines that Mura defects 1070 and 1080 are part of the same line Mura defect.
• the user extrapolates along a Mura defect such as 1080 to detect discontinuities between portions of a line Mura defect. In this case, Mura defect 1070 is found and Mura defects 1070 and 1080 coalesce to one defect.
• the original image is filtered using different frequency cut-off filters to form the filtered images, as an alternative to subsampling of the image and then filtering the subsampled images.
• larger kernel sizes are used for convolutions in order to achieve lower frequency cut-offs.
• the subsampled image where a line defect first disappears may indicate the width of a line because of the frequency cut-off.
• the user can also determine and look for frequency specific defects, for example, from thin line defects (several pixels) to thicker line defects.
• the presently claimed inventions may also be applied to other areas of technology that require optical inspection of a substrate for example, (non-exclusive) cathode ray tubes, semiconductor wafers, web inspection systems, medical imaging systems, and the like.

Landscapes

• Physics & Mathematics (AREA)
• Nonlinear Science (AREA)
• Engineering & Computer Science (AREA)
• General Physics & Mathematics (AREA)
• Crystallography & Structural Chemistry (AREA)
• Optics & Photonics (AREA)
• Chemical & Material Sciences (AREA)
• Theoretical Computer Science (AREA)
• Manufacturing & Machinery (AREA)
• Computer Hardware Design (AREA)
• Investigating Materials By The Use Of Optical Means Adapted For Particular Applications (AREA)
• Image Processing (AREA)
• Liquid Crystal (AREA)
• Testing Electric Properties And Detecting Electric Faults (AREA)
• Testing Of Optical Devices Or Fibers (AREA)
• Image Analysis (AREA)

Priority Applications (1)

Applications Claiming Priority (4)

Publications (2)

Family

ID=26667315

Family Applications (1)

Country Status (4)

Cited By (7)

Families Citing this family (87)

Family Cites Families (10)

• 1996
  • 1996-05-22 US US08/651,417 patent/US5917935A/en not_active Expired - Fee Related
  • 1996-06-11 KR KR1019970700951A patent/KR100399244B1/ko not_active Expired - Fee Related
  • 1996-06-11 JP JP50330797A patent/JP3914570B2/ja not_active Expired - Fee Related
  • 1996-06-11 WO PCT/US1996/010219 patent/WO1997000452A2/en not_active Ceased

Also Published As

Similar Documents

Legal Events