Image analysis – Applications – Manufacturing or product inspection
Patent
1996-10-16
1999-11-09
Chang, Jon
Image analysis
Applications
Manufacturing or product inspection
382288, G06K 900
Patent
active
059829192
DESCRIPTION:
BRIEF SUMMARY
FIELD OF THE INVENTION
This invention relates to an image processor having object recognition ability which is used for a LCD (liquid crystal display) tester for inspecting objects by using images, image inspection apparatus such as a CCD (charge-coupled device) tester, or image recognition apparatus for recognizing objects by using images.
BACKGROUND OF THE INVENTION
Conventionally, there have been many types of image processing apparatus for inspecting, recognizing, determining, or analyzing objects with use of images. Generally, most of them perform the image processing by using a computer which processes an image as a digitized image. Among the conventional types, there is also an apparatus which displays an image in two-tone such as black and white, or red and blue, by converting density of pixels into a binary system, as 0 and 1. The present invention relates to an image processor having object recognition ability which deals with the images which are processed by binary system, i.e., binary images. In this case, a pixel is defined as one of the points which compose a screen, i.e., an element, and an image is defined as a pixel having density value. In the binary image system, an image displayed as 1 or in black is called an active image. On the other hand, an image which is displayed as 0 or in white is called a non-active image.
Conventionally, a labeling system has been utilized so that the image processing apparatus is able to recognize an object which appears to be a mass of object to human eyes. In the conventional labeling system, it is arranged that the active image objects, which are adjacent each other and considered to be a mass, are labeled as "A, B, C . . .". The objects labeled in the same letter, for example as "A", are recognized to be in the same group.
However, the conventional labeling system is associated with the following disadvantage. Even though it is appeared to be a mass to human eyes, if there is one missing pixel in the object, the apparatus recognizes the object as the two separate objects. For example, as in FIG. 4, if there is a gap 55 between each pixel in the right and left sides of a golf club as a result of an interruption by noise or the like, the image processing apparatus recognizes the golf club as the two separate objects, a club head 51 and shaft 50. The image processing apparatus may recognize the left object as a hammer, the right object as a bar. Thus, it has been very difficult for the conventional image processing apparatus to recognize an object with gaps as a whole, even though it is appeared to be the golf club to human eyes, i.e., one continuous object.
SUMMARY OF THE INVENTION
Therefore, it is an object of the present invention to provide the image processor having an ability to recognize an object which is digitized in the binary system as a mass, as if it is appeared to be a mass to human eyes. Namely, the present invention is able to recognize one object as a whole even if it is divided to multiple labels, as if viewed in the human eyes.
In the image processing apparatus of the present invention, an object is labeled at first. Then, the object divided in the multiple labels is analyzed whether or not it is one object by a particular means/method so that the image processing apparatus is able to recognize the object in the same level of recognition ability as the human eyes.
The above method of the present invention is discussed in the following.
Digital image data is input by an image input device and stored in an input image memory by means of CCD camera or other communication means. If necessary, this digital image data is noise-removed by means of an average filter method or median filter method. Then, the density of each pixel is coded in the binary numbers by a threshold value. For example, if the density of the pixel is displayed in a range of 0-100, the threshold value is arranged to be a half of it, as 50. If the density of a pixel is either equal or more than 50, it is regarded to be a number 1, an active image, and displayed in bl
REFERENCES:
patent: 4484081 (1984-11-01), Cornyn, Jr. et al.
patent: 5301248 (1994-04-01), Takanori et al.
Advantest Corp.
Chang Jon
Patel Jayanti K.
LandOfFree
Image processor having object recognition ability does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Image processor having object recognition ability, we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Image processor having object recognition ability will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-1466896