Image analysis – Pattern recognition – Feature extraction
Reexamination Certificate
1998-06-02
2001-10-23
Boudreau, Leo (Department: 2621)
Image analysis
Pattern recognition
Feature extraction
C382S170000, C382S173000, C382S209000, C382S218000, C382S219000
Reexamination Certificate
active
06307963
ABSTRACT:
BACKGROUND OF THE INVENTION
1. Field of the Invention
The present invention relates to an image detection method, image detection apparatus, image processing method, image processing apparatus, and medium.
2. Description of the Related Art
Because performances of a color copying machine, a color scanner serving as an input/output unit of a personal computer, and a color printer have been remarkably improved in recent years, it has been possible to obtain a high-accuracy color copy easily. Therefore, because bills and securities may be forged by abusing the above high-accuracy units, the necessity for taking actions for the above abuse has been raised.
That is, it is detected whether a specific image printed on these bills and securities is included in an image to be copied in a copying machine. When a specific image is detected, it is necessary to stop image reading or inhibit normal printing to prevent a forgery from being produced. Moreover, it is necessary to prevent the above forgery from occurring by for example, using means such as a network and thereby notifying the forgery to a supervisor.
The position and direction of a specific image included in an input image are unknown.
As a conventional method for detecting a specific image from a color image under the above condition, there is a method using the matching of spread of the entire image in a color space, which is disclosed in the official gazette of Japanese Patent Application Laid-Open No.Hei 4-207465. However, in such method, unrelated image having a color tone which is generally similar to that of the specific imgae may be detected by mistake. Then in the case of this method, it is effective to verify a specific mark in an image. As the specific mark, a read seal on a bill can be listed.
To verify the specific mark in an image, a method disclosed in the official gazette of Japanese Patent Application Laid-Open No.Hei 8-115022 is used which finds a specific-color circular image and verifies the frequency pattern and the tone in the image. However, the above method may not be used because a specific-color figure such as a red seal is not always included in a manuscript which must be prevented from being forged.
In this case, the case of detecting a region in any circle on an image is shown in the official gazette of Japanese Patent Application Laid-Open No.Hei 6-268859 as a method capable of detecting the presence or absence of a specific image whose arrangement is unknown from an input image even if there is no specific-color pattern.
Moreover, as a conventional method for detecting a specific image from a color image, there is a method of narrowing down a range to a small region in which a specific image may be included and noticing only color information disclosed in the official gazette of Japanese Patent Application Laid-Open No.Hei 7-131636 in addition to the above method disclosed in the official gazette of Japanese Patent Application Laid-Open No.Hei 4-207465.
Moreover, as a method for detecting a specific image, an image processing apparatus is disclosed in Japanese Patent Application Laid-Open No.Hei 7-177349 separately from the above method. This conventional example is described below by referring to
FIGS. 38
to
40
.
In
FIG. 38
, symbol
101
denotes a position state detection section,
102
denotes a rotation processing section,
103
denotes a blocking section, and
104
denotes a block matching judgment section.
FIG. 39
shows a purposed image and a specific image
201
(rectangle at whose center a character B
1
is drawn) included in the purposed image. The position state detection section
101
detects the position of the specific image
201
in the purposed image in FIG.
39
. For detection, an edge of the rectangle is detected and the rectangular is calculated in accordance with the position and angle of the edge. The rotation processing section
102
rotates the image so that the specific image
201
so that it matches up to the horizontal direction in accordance with the positional information detected by the position state detection section
101
.
FIG. 40
shows an image obtained by rotating the specific image
201
shown in FIG.
39
. The blocking section
103
divides the image obtained by rotation (see
FIG. 40
) into blocks. The block matching judgment section
104
compares the average density of each of the blocks obtained by division with a predetermined value to judge whether the specific image
201
is included in the purposed image or not.
However, in the case of the above conventional method of using an optional portion of an image as purposed data, it is necessary to repeat obtainment of purposed data, extraction of a characteristic value, and judgment on the characteristic value while repeating the shift for each pixel and therefore, a lot of throughput is required. Particularly, to obtain the data of a region in a circle, it is necessary to sequentially perform the processing such as address calculation of data to be obtained, causing the throughput to further increase. Because the throughput increases proportionally to the square of the radius of a circle, it is better to use a circle as small as possible in order to decrease the throughput. However, because of the narrow range of characteristic values obtained from a circular region having a small radius, there is a problem that the accuracy for detection of a specific pattern is deteriorated.
Moreover, in the case of the above conventional method of noticing only the color information necessary for detection but using no positional information such as color arrangement, if an image having a tone similar to that of a specific image to be detected is input, the image may be erroneously detected as the specific image at a high probability. Moreover, in the case of the method of narrowing down a range to only a part of an image and observing only the point, if there is a similar image on the portion, an image whose entire image is completely different may be erroneously detected as a specific image. Therefore, there is a problem that the accuracy for detecting a specific image is low.
In the case of the conventional method described for
FIGS. 38 and 40
, it is a premise to accurately detect the position of the specific image
201
. Therefore, when other images including an edge are present around a specific image, there are problems that it is difficult to accurately detect the edge of the specific image, it is impossible to accurately detect the position of the specific image, and resultingly the accuracy for detecting the specific image is deteriorated. Moreover, to handle not an image having a simple shape comprising straight lines such as a rectangle but an image having a complex shape comprising curves, it is necessary to newly prepare a technique for detecting the position and tilt of the image having a complex shape. Furthermore, though it is possible to use a specific-color mark (e.g. red mark) to specify a position, there is also a disadvantage that images other than an image including a specific-color mark cannot be detected.
The present invention is made to solve the above problems and its object is to provide an image detection method, image detection apparatus, image processing method, image processing apparatus, and medium capable of more efficiently processing data to detect a specific image and preventing a detection accuracy from deteriorating.
SUMMARY OF THE INVENTION
To achieve the above objects, an aspect of the prevent invention is an image detection method comprising the steps of:
holding image data including image to be detected;
scanning said held image data, extracting data from image in a substantially-circumferential region having a radius r around said scanning position by data extraction means;
calculating an extracted-data characteristic value in accordance with said extracted data;
accumulating reference characteristic value which is obtained based on each image in (1) a substantially-circumferential region having a radius corresponding to said radius r on said image to be detected and in (2
Imagawa Taro
Mekata Tsuyoshi
Nishida Yukihiro
Boudreau Leo
Mariam Daniel G.
Matsushita Electric - Industrial Co., Ltd.
Ratner & Prestia
LandOfFree
Image detection method, image detection apparatus, image... 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 detection method, image detection apparatus, image..., we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Image detection method, image detection apparatus, image... will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-2576721