METHOD AND APPARATUS FOR DETECTION
OF PARALLEL EDGES IN IMAGE PROCESSING
FIELD OF THE INVENTION 5
This invention relates generally to image processing and particularly to the detection of parallel edges in image analysis.
BACKGROUND OF THE INVENTION 1°
The occurrences and applications for various forms of image processing have experienced an overwhelming growth with the advent of high power computing systems. Such image processing systems are utilized in a 15 variety of fields including computer image enhancement, robotic viewing and monitoring systems, object and target recognition in surveillance and weapons systems, and character recognition in character scanning systems such as optical character readers and the 20 like. This is by no means an exhaustive list of image processing environments but is indicative of the diversity of uses for image processors. While image processing systems are required to deal with a great diversity of problems and challenges characteristic of the specific 25 environment of the image processing system, most image processing systems are subjected to two basic or overall challenges which may be generally referred to as recognition and noise filtering.
Recognition basically involves the determination of 30 the nature and extent of the image elements presented to the system. For example, an aerial or satellite photo presents a collection of different contrast areas. However, little information is present as to which areas are objects and which are shadows. Similarly, little or no 35 information may be available as to which elements are to be grouped in a collective object. Also a determination of which elements are manmade and which are surrounding topography may also be difficult to determine. Another example of a recognition challenge is found in the analysis of a section of test material. To the image processing system, test material presents a plurality of elements or strokes which must be grouped into characters, words and so on.
45
Noise filtering, on the other hand, involves distinguishing between image information from the viewed object and extraneous information added by the various system components.
Recognition and noise filtering are, of course, interre- J0 lated in any image processing system. For example, attempts to reduce noise using techniques such as threshold filtering often sacrifice information which may be deemed valuable in the recognition process.
One of the more basic aspects in image processing is 55 the detection of parallelism among the image elements. The detection of parallelism provides a basic tool in image processing because parallelism in present in a substantial number and variety of image elements. For example, a picture of a tree contains a large number of 60 parallel edges. At the largest scale, the trunk possesses a left and right vertical border which are highly parallel to each other. At finer scale, the branches and even the twigs display a similar parallelism. While naturally occurring image elements provide substantial parallelism, 65 manmade images possess an even greater degree of inherent parallelism. One example of manmade image parallelism is found in text material which generally
comprises a number of strokes, each of which usually defines two parallel edges.
In addition to a strict definition of parallelism, other objects which do not possess classic parallel edges may be generally considered parallel to derive valuable image processing information. For example, objects which possess an axis of symmetry such as an hourglass or cone may possess an overall characteristic within an image which is similar to parallelism and which may be desired to be treated in the manner parallel edges are treated.
The intensity of effort by practitioners in the image processing art has resulted in the publication of a number of related articles and treatises. The following are included in such relevant publications: Binford, Thomas. 1971. Visual Perception by Computer. IEEE conference on Systems and Control. December, 1971, Miami. Blum, H. 1973. Biological shape and visual science, part
1. J. Theor. Biol. 38, 205-287. Canny, John. 1986. A computational approach to edge detection. IEEE PAMI, Vol PAMI-8, no. 6., November 1986.
Hildreth, Ellen. 1984. The Measurement of Visual Motion. MIT Press. Marr, David. 1982. Vision. W. H. Freeman. New York. Rao, Kashipati and Nevatia, Ramakant. 1989. Descriptions of complex objects from incomplete and imperfect data. Proceeding of the DARPA image understanding conference. May 1989, Palo Alto, Calif. The foregoing provide a sampling of the various methods of image processing techniques which are known. In addition, other examples are found in the following U.S. Patents: U.S. Pat. No. 4,618,989 issued to Keisuke, et al. sets forth a METHOD AND SYSTEM FOR DETECTING ELLIPTICAL OBJECTS in which the geometric properties an ellipse are determined separately on three parameter sub-spaces obtained on the basis of edge vector field.
U.S. Pat. No. 4,424,588 issued to Takashi, et al. sets forth a METHOD FOR DETECTING THE POSITION OF A SYMMETRICAL ARTICLE by converting image signals of the article into binary information and detecting the position of the article based upon the binary information. A first median point is determined from the binary information which is between two points of intersection between a first straight line intersecting the article and the sides of the article. A second median point is determined from the binary information which is between two points of intersection between a second straight line passing through the first median point crossing the first straight line at right angles and crossing the sides of the article.
U.S. Pat. No. 4,896,279 issued to Yoshida sets forth a METHOD AND APPARATUS FOR INSPECTING AN EXTERNAL SHAPE OF AN OBJECT HAVING A SYMMETRY in which an object is sensed by a photoelectric conversion sensor having a photoelectric conversion screen expressed as an XY coordinate grid such that the symmetrical axis of the object is parallel to one axis of the XY coordinate. Coordinates at a plurality of right and left points with respect to the other axis of the XY coordinate at which an external contour of the object intersects a plurality of straight lines perpendicular to the symmetrical axis are determined.
U.S. Pat. No. 4,435,836 issued to Ruben sets forth a TECHNIQUE FOR EXTRACTING FEATURES FROM IMAGES in which attributes or features of a
