Image analysis – Applications – Manufacturing or product inspection
Reexamination Certificate
1998-08-31
2001-07-10
Bella, Matthew C. (Department: 2621)
Image analysis
Applications
Manufacturing or product inspection
C382S147000, C382S172000, C382S199000, C382S281000
Reexamination Certificate
active
06259809
ABSTRACT:
BACKGROUND OF THE INVENTION
1. Field of the Invention
The present invention relates to a system for and a method of recognizing a feature contained in an image or picture such as, for example, a pattern image on the surface of a semiconductor wafer, a printed wiring pattern or printed circuit pattern of a printed board, or the like, and more particularly, to such image information recognition system and method suitable for use in recognizing linear information on a digital image or picture.
2. Description of the Related Art
There have been previously known in the art many image processing methods and systems which perform an inspection, recognition, judgment or diagnosis of an object or article using an image or picture thereof. Almost all of the processing methods and systems utilize a digital image processing in which an image is digitally treated or processed and they perform the image processing using a computer.
For example, the occurrence of a defect or defects or a commingled foreign substance or substances during the manufacture of a printed board or a semiconductor wafer comes to the cause of a faulty or defective product. Herein, the term “printed board” is intended to refer generically to a single-sided, a double-sided and a multi-layer printed boards inclusive of flexible printed circuits. It is thus necessary that the occurrence of any defect inclusive of a foreign substance is rapidly detected. In the conventional practice, if a defect occurs on a printed board or a wafer surface, the defect is observed by use of, for example, a scanning electron microscope (SEM), and the result of observation (an image of the printed board or wafer surface obtained by taking it) is stored as image data.
A multiplicity of image information recognition methods and systems have been also proposed in the art for detecting, recognizing or rendering a decision about the presence of a defect and/or an alien substance as mentioned above on the basis of a photograph of a printed board, or an image or picture of a semiconductor wafer obtained by having taken it using the scanning electron microscope or the like and supplied thereto.
It is noted that there exist a number of images having a very similar structure among the stored defective images thus obtained. It happens that during the observation of a wafer image, one may want to see a past image which exhibited a similar structure. However, at present, he can find no other alternative but to retrieve image files one after another relying on his memory. In addition, since features appearing on the image data do not be readily expressible by words, frequently he cannot find the exact past image of similar structure. That is, the exact past image can be easily found only by one who has actually observed this past image. Thus, it is not a simple matter to share such image data with others. To resolve such problem, there is an increasing need for a system which enables an efficient retrieval of image data of printed boards and/or semiconductor wafers.
Conventionally, a retrieval of image data is performed principally by utilizing a pattern matching technique which compares an image to be detected with the stock of images. The comparison between two images means a processing in which values of picture elements of the one image are taken from values of picture elements of the other image for their corresponding picture elements to create a difference image with respect to picture element values between the two images. When such a difference image is created, if the two images being compared are all the same, the resulting difference image will be a flat image having its all picture element values of 0 throughout the entire image. However, if defects or foreign matters of the two images are different in shape from each other, picture elements having values other than 0 appear concentratedly in the region that the shapes are different. In this instance, the total number of picture elements which have values other than 0 or the like is measured and the measured value is compared with a preset threshold, and if the measured value is less than the threshold, a decision is rendered that the two images are similar and the decision is outputted as a result of retrieval.
In this manner, only similar images can be extracted from the stored image data.
In case image data is an image of a semiconductor wafer or a printed board, the background of the image includes a printed wiring pattern which runs vertically and horizontally or obliquely. The wiring pattern is normally aligned in a given direction in regular manner, and accordingly, the use of the pattern matching technique with the wiring pattern results in canceling out the background of the image. As a result, only differences in unique areas (features) of the two images such as defects and/or foreign substances can be extracted.
Where it is desired to detect linear components regularly aligned such as components of the wiring pattern in the image data, the pattern matching technique mentioned above can be simplified and performed at high speed in the following manner.
Initially, an edge detection procedure is applied to an original image to create a binarization image of binary values comprising only edge components. Edge components of linear components (the wiring pattern) in the original image remain among the created binarization image. As a template for the binarization image is prepared a binarization image which depicted a single straight line for every inclination and every intercept. The mathematical product is taken between the template and a binarization image to be detected, and the number of active picture elements in the product image is counted. When a linear component which is identical with the linear component depicted in the template exists in the binarization image to be detected, the active picture elements in the template overly those in the binarization image to be detected, whereby the number of active picture elements in the product image increases. If the count exceeds a predetermined threshold, a decision is rendered that the linear component depicted in the template does exist in the original image.
In this manner, the presence of any linear component in the original image can be detected.
While the described defect detecting technique is already established, it suffers from a disadvantage in that the comparison of the two images by using the pattern matching technique requires a register or matching of corresponding locations of two images such as wiring patterns on the two images. The register of corresponding locations must be done accurately and such work requires a considerable labors.
The stored image data which is to be retrieved has been reserved in the state that an image was rotated at a predetermined angle and/or enlarged (scaled up) or reduced (scaled down) in order to allow an observer to analyze detailed features of defects. Accordingly, the orientation and/or size of the wiring pattern often varies from image to image, and this adds a further difficulty to the registering operation.
In such way, if a modification such as rotation, magnification or minification should be applied to the same image or picture, even the retrieval of the same image using the pattern matching technique becomes very difficult.
It will then be seen that when an angle of rotation as well as a magnification/minification of each wiring pattern in the images of semiconductor wafers and/or printed boards which are stored as a stock in a database are quantified and if the background of individual images can be matched to each other by using the quantified values, or in other words, if a normalization of images can be achieved, a subsequent registering of a pair of images can be dispensed with, and an image retrieval is enabled in a facilitated manner by a simple technique such as the pattern matching technique, for example.
Further, in case that the linear components in a binarization image comprising only edge components are detected using the pattern matching technique as described abo
Advantest Corporation
Bella Matthew C.
Chawan Sheela
Gallagher & Lathrop
Lathrop David N.
LandOfFree
System and method for recognition of image information does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with System and method for recognition of image information, we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and System and method for recognition of image information will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-2485070