Image analysis – Pattern recognition – Feature extraction
Reexamination Certificate
1999-12-09
2001-01-30
Johns, Andrew W. (Department: 2721)
Image analysis
Pattern recognition
Feature extraction
C382S286000
Reexamination Certificate
active
06181820
ABSTRACT:
BACKGROUND OF THE INVENTION
1. Field of the Invention
The present invention relates to an object extraction method and apparatus for extracting a specific image from color images or black-and-white images. Further present invention relates to a color-image recognition method and apparatus for determining whether the thus-extracted specific image is an image from given images. In practice, such a specific-image extraction is performed as follows: Data representing a specific image is extracted from data representing color images or black-and-white images. Further, data representing a specific image part is extracted from data representing a specific color image or a black-and-white image.
Each of the color-image recognition method and apparatus uses the thus-extracted specific image and then determines whether the specific image corresponds to an image included in given images, which are, for example, images constituting an image database, in particular, a color-image database.
In an example, such a technique may be applied to works in a police office. The police office may have a database of criminals' face photographs. When a montage picture (composite picture) is given by a witness, the police officer attempts to determine whether the image of the thus-given picture is an image included in the database.
In another example, an intelligent building may use such a technique. A microcomputer is provided inside a door of a room of a firm and the microcomputer uses a database of employees' face photographs. If a person comes in front of the door, the microcomputer uses a CCD camera to take image data of the person's face and, by using the thus-taken image data, the microcomputer attempts to determine whether the same image data is included in the database. If the microcomputer has succeeded in determining the same image data is in the employees' face database, the microcomputer opens the door for the person, and if not, the microcomputer does not open it. Thus, no real key is needed for the door.
In another example, such a technique may be used to prevent bank notes and securities from being replicated. A duplicator may have image data of bank notes and securities and if an operator sets an original image on the duplicator to obtain a copy thereof, the duplicator checks if data of the thus-set original image is included in the image data of bank notes and securities.
2. Prior Art
Apparatuses processing color images such as a color duplicator, a color printer, a color scanner, a color-image communication apparatus, and so forth are expected to be further developed in the future. It has become easier to handle the color images by hardware development, such as cost reductions and capacity increase of various sorts of memories, data-communication cost reduction and so forth. However, since data representing the color images has a huge data amount (for example, 96 Mbytes for an image of the A3 size) it is impossible to handle the color images similarly to handling data representing two-tone images.
In particular, in handling image data, a complicated process such as image recognition (such as recognition of a specific image, OCR or the like) requires a huge amount of data to be processed.
Therefore, it is further difficult to realize the image recognition for the color images.
Various methods have been proposed as an object extraction method for extracting an object to be recognized. In an example method, the method extracts black-continuity rectangles from a relevant image and compares values representing the thus-extracted rectangles with predetermined threshold values. The method classifies the rectangles into rectangles representing characters and rectangles representing line drawings. See Japanese Laid-Open Patent Application No.55-162177. However, this method neither classifies the thus-classified rectangles of line drawings into rectangles representing horizontal ruled-lines, nor has a capability to extract a specific image even if the relevant image is present but is in a rotated form.
As an apparatus and method for identifying a specific color image, a color-image processing apparatus (see Japanese Laid-Open Patent Application No.4-54681) was proposed, which apparatus extracts a hue and a distribution of the hue from a given region of an original image and thus detects whether or not a specific image is present. However, the above-described apparatus uses only 32 points to be sampled, the number of points to be sampled being a few in order to reduce a data amount to be processed. Such using of a few sampled points may not result in obtaining a sufficient amount of information required to extract characteristics of a relevant image. Therefore, it is not possible to obtain highly accurate image identification performance.
Further, a method was proposed for determining whether or not a specific color appears and then extracting image parts having the same characteristics (see Japanese Laid Open Patent Application No.4-180348). Each of patterns constituting an image has a particular stretch in a color space. The above-mentioned method uses such specific stretches, in the color space, of the patterns to identify a certain image part as being a specific image part, the certain image part having specific stretches in the color space the same as those of the specific image part. However, the method cannot distinguish image parts having the same stretch in the color space but a different color distribution inside the stretch. Thus, if a certain image part has a stretch in the color space, which stretch is the same as the stretch of a specific image part, but a color distribution inside the stretch is different from that of the stretch of the specific image part, the method may erroneously identify the certain image part as being the specific image part.
SUMMARY OF THE INVENTION
An object of the present invention is to provide an object extraction method and apparatus, each of which extracts an image region corresponding to an object from an input image with high accuracy.
Another object of the present invention is to provide a color-image recognition method and apparatus, each of which method and apparatus extracts an object from an input image and performs a color-image recognition operation on the object. Each of the color-image recognition method and apparatus obtains a sufficient amount of information required for the color-image recognition operation so that high-accuracy color-image recognition can be achieved with a data amount to be processed being compressed.
Another object of the present invention is to provide a color-image recognition method and apparatus, each of which method and apparatus uses a table-conversion manner to convert relevant data, instead of performing a calculation operation, so as to compress the amount of the relevant data. Thus, it is possible to achieve an effective data-amount compression.
Another object of the present invention is to provide a color-image recognition method and apparatus, each of which method and apparatus suppresses an adverse effect of noise image information. The noise image information may be included in relevant information as a result of, for example, an image scanner reading in information of not only a relevant image printed on a relevant sheet but also of another image which is printed on another sheet subsequent to the relevant sheet. The image of the subsequent sheet may be read by the scanner through the relevant sheet if it is thin. Each of the method and apparatus performs a vector quantizing operation only on the relevant image information. Thus, image-recognition capability and data processing speed are improved.
Another object of the present invention is to provide a color-image recognition method and apparatus, each of which method and apparatus determines a similarity between information of an input image and information of a predetermined code book information. If the thus-determined similarity is higher than a predetermined threshold value, no vector-quantizing value is assigned to the informa
Hirano Akihiko
Miyazawa Toshio
Tachikawa Michiyoshi
Johns Andrew W.
Oblon & Spivak, McClelland, Maier & Neustadt P.C.
Ricoh Company,. Ltd.
LandOfFree
Image extraction method and apparatus and image recognition... 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 extraction method and apparatus and image recognition..., we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Image extraction method and apparatus and image recognition... will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-2478108