JP2003331288A - Device and method for discriminating image - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To precisely discriminate similarity between image data. <P>SOLUTION: A CPU 2 provided in an information providing server 1 of the present invention divides the image data into a plurality of areas classified a color by a color, calculates a pseudo-joining quantity between the plurality of divided areas, and generates a mapping vector based on the image data, using the calculated pseudo-joining quantity between the plurality of divided areas. <P>COPYRIGHT: (C)2004,JPO

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an image discriminating apparatus and an image discriminating method.

[0002]

2. Description of the Related Art In recent years, with the development of image analysis technology, an image discriminating apparatus has been proposed which discriminates image data having the highest similarity from the imaged data of a discrimination object from a plurality of different image data. In the discrimination processing, a method is generally used in which a feature amount is extracted from a plurality of image data and the similarity between each image data and the imaged data is determined based on the feature amount. As the feature quantity, the periodicity, shape, texture, edge, texture edge, etc. of each area divided in the image data are used.

[0003]

However, the above conventional image discriminating apparatus has the following problems. That is, when the periodicity, the shape, the texture, the edge, the texture edge, or the like is used as the feature amount, the projective distortion and the imaging direction due to the deviation of the imaging position of the imaging data are easily affected. The shift of the image pickup position is particularly remarkable when an object is photographed by using an image pickup device having a small display screen like a mobile terminal.

In view of the above circumstances, an object of the present invention is to realize an image discriminating apparatus and an image discriminating method for discriminating the similarity between image data with high accuracy.

[0005]

[Means for Solving the Problems] In order to solve the above problems,
An image discriminating apparatus according to the present invention includes a dividing unit that divides image data into a plurality of color-based regions, and a joining amount calculation unit that calculates a pseudo joining amount between the plurality of color-based regions divided by the dividing unit. Generating means for generating a mapping vector from the image data by using the pseudo-joint amount between the plurality of color regions calculated by the joint amount calculating means.

According to the image discrimination method of the present invention, in the image discrimination method in which the image discrimination device discriminates the similarity of a plurality of image data, a division step of dividing the image data into a plurality of areas for each color, and the division step. The image is obtained by using a joining amount calculation step of calculating a pseudo joining amount between the plurality of color-separated regions and a pseudo joining amount between the plurality of color-separated regions calculated in the joining amount calculation step. Generating a mapping vector from the data.

The present invention focuses on the positional relationship between the color areas when the image data is divided into the color areas, in particular, the pseudo-joint amount. According to these aspects of the invention, the image data is divided into a plurality of color-based areas, the pseudo-joint amount between these color-based areas is calculated, and a mapping vector is generated from the image data using the pseudo-joint amount. Therefore, this mapping vector reflects the amount of pseudo joining between the color-based regions that form the image data as a feature amount.

By applying the above-described mapping vector generation process to a plurality of image data registered in advance and image data (imaging data described later) which is a target of similarity or identity determination, It is possible to perform image discrimination using a feature amount (pseudo-joint amount) that is unlikely to be affected by the projective distortion due to the shift of the image pickup position or the image pickup direction. Specifically, a discrimination vector and a bias, which will be described later, are further generated based on the mapping vector generated from the plurality of image data.
Then, the inner product of the discriminant vector and the mapping vector generated from the imaged data is taken, and a bias is added to this scalar value to perform binary discrimination, thereby determining the pseudo-joint amount for the similarity discrimination between the image data. Use as an indicator.

This makes it possible to accurately determine the similarity between image data. For example, the image data most similar to the imaged data can be selected and extracted with high accuracy from the plurality of image data. The deviation of the imaging position is remarkable when the object is imaged by using the imaging device having a narrow display screen, and the present invention is particularly effective for the image discrimination in the mobile terminal.

In the image discriminating apparatus according to the present invention, preferably, the pseudo joining amount is a color of the base point and a color of each pixel other than the base point in a plurality of pixels including one pixel on the image data as a base point. It is the result of counting the combination with.

In the image discrimination method according to the present invention, preferably, the pseudo joining amount is a color of the base point and a color of each pixel other than the base point in a plurality of pixels including one pixel on the image data as a base point. It is the result of counting the combination with.

According to these aspects, the pseudo-joint amount is the color of the base point and each pixel other than the base point in a plurality of pixels (boxes described later) including one pixel on the image data as a base point. It is the result of counting the combinations with colors (for example, one set of two colors of white and black). By moving the base point to any pixel on the image data, calculating the pseudo-joint amount for as many base points as possible, and generating a mapping vector, it is possible to accurately determine the similarity between a plurality of image data. It will be possible.

In the image discriminating apparatus according to the present invention, preferably, an inclusion amount calculation unit for calculating a pseudo inclusion amount between the plurality of color-separated areas divided by the dividing unit is further provided, and the generation unit includes the pseudo junction. A mapping vector is generated from the image data by using the amount and the pseudo inclusive amount between the plurality of color regions calculated by the inclusive amount calculating means.

In the image discriminating method according to the present invention, preferably, the image discriminating apparatus further includes a comprehensive amount calculating step of calculating a pseudo comprehensive amount between the plurality of color-based areas divided in the dividing step, and the generation is performed. In the step, a mapping vector is generated from the image data by using the pseudo joining amount and the pseudo inclusion amount between the plurality of color regions calculated in the inclusion amount calculating step.

The present invention focuses on the arrangement relationship between the color areas when the image data is divided into the color areas, in particular the pseudo comprehensive amount. According to these aspects of the invention, the image data is divided into a plurality of color-based regions, the pseudo-joint amount and the pseudo-inclusive amount between these color-based regions are calculated, and a mapping vector is generated from the image data using these. . Therefore, the mapping vector reflects the pseudo-joint amount and the pseudo-inclusive amount between the color-based regions forming the image data as the feature amount. Therefore, by combining the two types of feature amounts relating to colors, the discrimination accuracy is improved as compared with the case where the pseudo joining amount is used as the only feature amount.

In the image discriminating apparatus according to the present invention, preferably, the pseudo comprehensive amount is obtained by reconstructing the plurality of color-based areas as a plurality of closed areas and using one pixel on the image data as a base point. A closed region including the base point,
It is the result of counting the combinations with closed regions existing in all directions within the predetermined direction with the base point as the center.

In the image discriminating method according to the present invention, preferably, the pseudo comprehensive amount is obtained by reconstructing the plurality of color-based areas as a plurality of closed areas and using one pixel on the image data as a base point. A closed region including the base point,
It is the result of counting the combinations with closed regions existing in all directions within the predetermined direction with the base point as the center.

According to these aspects of the invention, the pseudo comprehensive amount is obtained when the plurality of color-based areas are reconstructed as a plurality of closed areas (labels described later) and one pixel on the image data is used as a base point. , A combination of a closed region including the base point and a closed region existing in all directions within a predetermined direction (for example, four directions of up, down, left, and right) around the base point (for example, the first label and the second label). 1 in two closed areas of label
It is the result of counting the set). By moving the base point to any pixel on the image data, calculating the pseudo-inclusive amount for as many base points as possible, and generating a mapping vector, it is possible to accurately determine the similarity between a plurality of image data. It will be possible.

In the image discriminating apparatus according to the present invention, more preferably, the image data is image data of a signboard. In the image discrimination method according to the present invention, more preferably, the image data is data obtained by capturing an image of a signboard.

According to these aspects of the invention, the image data is image data of a signboard. Signboards that are artifacts are often painted in multiple colors so that the creator can effectively convey information to others. Therefore, it is particularly effective to use the feature amount related to the color arrangement or the color arrangement relation such as the pseudo joining amount or the pseudo inclusion amount.

[0021]

DETAILED DESCRIPTION OF THE INVENTION An information providing system according to the present invention will be described below with reference to the accompanying drawings. First, Fig. 1
FIG. 3 is a diagram conceptually showing an information providing service realized by this system in order to suggest the outline of the present invention. Figure 1
As shown in, the user of the mobile phone 10 uses the signboard on the street corner photographed by the camera mounted on the mobile phone 10,
When a predetermined operation described later is executed, information related to the signboard is displayed on the display screen of the mobile phone 10. For example,
When the imaging target is the signboard A2 for restaurant advertisement, not only the name of the restaurant but also the reputation of the customer, the current waiting time, and the like are presented to the user in real time. As a result, the user can easily and quickly obtain desired information.

(First Embodiment) Hereinafter, a first mode in which a mode in which the mobile phone 10 can communicate with an information providing server (hereinafter referred to as "online mode") is applied to the information providing system according to the present invention. Will be described. As a premise of the description, in order to clearly distinguish the image data registered in advance in the information providing server 1 for determining the image capturing target and the image data to be the determination target captured by the mobile phone 10, the former is simply referred to as " The image data will be referred to as "image data", and the latter will be referred to as "imaging data".

First, the structure will be described. FIG. 2 is a schematic diagram showing an example of the overall configuration of the information providing system 100 in this embodiment. As shown in FIG. 2, the information providing system 100 is roughly configured by an information providing server 1 (corresponding to an image discrimination device) and a mobile phone 10. The information providing server 1 is wire-connected to the base station B1 and the network N. Furthermore, the mobile phone 10 can wirelessly communicate with the base station B1. Therefore, the mobile phone 10 is
It is possible to exchange information with the information providing server 1 via the network 1 and the network N.

The configuration of the information providing server 1 will be described with reference to FIG. FIG. 3 is a block diagram showing a functional configuration of the information providing server 1. As shown in FIG. 3, the information providing server 1 includes a CPU 2 (corresponding to a dividing unit, a joining amount calculating unit, a comprehensive amount calculating unit, and a generating unit), an input unit 3, a display unit 4,
The computer is composed of a communication control unit 5, a RAM 6, and a storage unit 7, and each unit is a computer connected by a bus.

CPU (Central Processing Unit) 2
Reads out the program stored in the storage unit 7 and expands it in the RAM 6, and centrally controls the operation of each unit according to the program. Further, the CPU 2 executes an information providing process (see FIG. 6) described later according to the program expanded in the RAM 6, stores the processing result in the RAM 6 and displays it on the display unit 4. Then, the processing result stored in the RAM 6 is stored in a predetermined storage destination of the storage unit 7.

The input unit 3 comprises a keyboard having character / number input keys, cursor keys, various function keys, and the like, and a mouse as a pointing device, and signals of keys pressed by the keyboard. And a mouse operation signal are output to the CPU 2 as input signals.

The display unit 4 is an LCD (Liquid Crystal Display).
play) and CRT (Cathode Ray Tube)
Various information is displayed on the screen in accordance with the instruction of the display signal input from the CPU 2. The communication control unit 5 is wire-connected to the base station B1 via the network N and transmits / receives data to / from the mobile phone 10.

A RAM (Random Access Memory) 6 is a C
In various processes controlled to be executed by the PU 2, it forms a temporary storage area for various programs, input or output data, parameters, etc. that can be executed by the information providing server 1 and read from the storage unit 7.

The storage unit 7 stores a program necessary for executing an information providing process described later. The storage unit 7 includes a database 71. The database 71 includes the mobile phone 10
A range in which the image capturing target can be captured (hereinafter, referred to as “service range”) and related information regarding the image capturing target are stored in association with each other in advance. In addition, the storage unit 7 stores the image data acquired (registered) in order to determine the imaging target,
It is stored in association with the service range and related information. The storage unit 7 stores a later-described discrimination vector and bias calculated based on the image data.

Hereinafter, the positions P1, P2 and P3 shown in FIG.
The data of the service range stored in the database 71 will be described in detail by taking the signboards A1, A2, and A3 installed in the above as an example. FIG. 4 shows signboards A1, A2.
It is an overhead view which shows typically the positional relationship of the service range of A3. As shown in FIG. 4, the service ranges C1 to C3 corresponding to the signs A1 to A3 are the positions P1 of the signs A1 to A3.
It is defined around P2 and P3. When the imaging target is a flat object such as a signboard or poster, the service range is generally fan-shaped due to the effects of the viewing direction and the obstruction, but for the sake of simplicity, the service range defined as a circle will be described as an example. To do.

The mobile phone 10 covers the service range C1 to C3.
It is set to be slightly extended from the actual visible range in consideration of the position estimation error due to. The error of position estimation is GP
When S is used, it is several m to several tens m, and the radius of the service range C1 to C3 is expanded accordingly. The position estimation error when using only the color code indicating the cell of the base station B1 or the service area is several hundreds to several kilometers.

The viewing angle of the imaging device of the mobile phone is φ [ra
d], and the width of the signboard is W _m [m] (identification symbol m = 1, 2, 3 of the signboard to be imaged), the distance R _m where the entire signboard fits on the display screen is R _m = W _m / 2 tan (φ / 2) [m]
Represented by Further, the error of the positions of the signboards A1 to A3 estimated at the time of registration of the image data is ΔR _m, and the positions of the signboards A1 to A3 estimated by the mobile phone 10 or the mobile phone 10 and the base station B1 at the time of determining the imaged data. Error of
When R _u, coverage C _m Signs A _m at the position P _m is the radius R _m around the _{P m '= R m + △} R m + △ R
It becomes the inside of the circle of _u [m].

Next, with reference to FIG. 5, the internal structure of the mobile phone 10 will be described in detail. First, the configuration will be described. FIG. 5 is a block diagram showing a functional configuration of the mobile phone 10. As shown in FIG. 5, the mobile phone 10 includes a control device 11,
Input device 12, RAM 13, display device 14, storage device 1
5, an imaging device 16, a wireless communication device 17, a position estimation device 18 such as GPS, and a voice processing device 19. Further, these respective devices are electrically connected to each other via a bus so that various signals can be input and output.

The controller 11 is a ROM (Read Only Memo).
ry) 11a is built in, a program stored in advance in the ROM 11a is read out, expanded in the RAM 13, and each unit is centrally controlled according to the program. That is, the control device 11 executes the information providing process (see FIG. 6) described later according to the input signal from the input device 12 and the program expanded in the RAM 13, and temporarily stores the processing result in the RAM 13 and displays it. Display on the device 14. And
The processing result stored in the RAM 13 is stored in a predetermined area in the storage device 15 as needed.

The input device 12 selects data and processing,
It is configured to include various operation buttons for instructing image capturing of an image capturing target, etc. These various operation buttons are pressed individually or in combination to output an input signal according to the instruction content to the control device 11.

A RAM (Random Access Memory) 13 is
The program read from the ROM 11a and the data read from the storage device 15 are temporarily stored in various processes executed by the control device 11 and are configured by a volatile semiconductor memory. Further, the RAM 13 is a VRA that temporarily stores the data displayed on the display device 14.
It also has the function of M (Video RAM).

The display device 14 is an LCD (Liquid Crystal).
Display), EL (Electro Luminescence), etc., and displays image data and related information on the screen according to a display signal input from the control device 11.

The storage device 15 is an EEPROM (Electric).
ally Erasable and Programmable ROM), etc., and stores data necessary for executing various processes, data generated as a result of executing various processes, and the like.

The image pickup device 16 has a known CCD (Charge Coupled Diode) camera 16a whose viewing angle is set to φ [rad]. The viewing angle φ is preferably π / 2. The image pickup device 16 converts an image of an image pickup target picked up by the CCD camera 16a into an electric signal, and stores the electric signal in a predetermined area in the storage device 15 as image pickup data. The imaging device 16 displays the imaging target on the display device 14 at the time of imaging, and the display device 1 is displayed when the imaging button (shutter) is pressed.
The image pickup target displayed in 4 is imaged. The position and size of the imaging target can be adjusted by appropriately moving the position of the CCD camera 16a or the mobile phone 10 itself in the horizontal or vertical direction.

The wireless communication device 17 controls wireless communication with the base station B1. Specifically, the wireless communication device 17 is a circuit that includes a modulation / demodulation unit (not shown) that modulates and demodulates a signal, and an encoding / decoding unit (not shown) that encodes and decodes the signal. , With an antenna A. Antenna A
Is provided on the upper part of the housing of the mobile phone 10 so as to be expandable and contractible, and transmits and receives radio waves to and from the base station B1.

The position estimating device 18 uses a GPS (Global Pos).
The current position (latitude, longitude, etc.) of the mobile phone 10 is estimated with a precision of several tens of meters using radio waves from a satellite, and is displayed on the display device 14. Position estimation technology
Since this is a well-known and commonly used two-dimensional measurement technique, detailed description (including illustration) of the technique is omitted.

The voice processing device 19 is composed of a converter, an amplifier and the like, and includes a microphone M and a speaker S. During a call, the voice processing device 19 converts voice data input from the control device 11 into an analog signal by a converter and outputs the sound from the speaker S via an amplifier. Further, the voice processing device 19 converts a voice signal input from the microphone M into a digital signal capable of wireless communication by a converter and outputs the digital signal to the control device 11 during a call. The voice processing device 19 is, for example,
Play the audio data if the related information includes the audio data.

Next, the information providing system 10 according to the present invention.
The operation of 0 will be described, and the image discrimination method according to the present invention will be described. Each process described below is performed by the storage unit 7 of the information providing server 1 or the ROM 1 of the mobile phone 10.
It is realized by the program stored in 1a being executed by the CPU 2 or the control device 11.

FIG. 6 is a flowchart showing the flow of the information providing process executed by the information providing system 100. First, in the image learning phase of registering image data for discriminating an imaging target, the information providing server 1 acquires image data for discriminating an imaging target (S1). The acquisition of the image data is performed by an electronic device having an imaging function, and is preferably performed by another mobile phone or the mobile phone 10 having the same imaging characteristics (screen size and viewing angle) as the mobile phone 10. It is desirable that the number of pieces of image data to be acquired is at least 2 or more from the viewpoint of maintaining and improving the accuracy of discrimination of the imaging target.

In S2, the information providing server 1 generates a mapping vector from the image data acquired in S1. The color space mapping feature amount extraction processing will be described below with reference to FIG. 7. As shown in FIG. 7, the image data of the signboard A1 (corresponding to m = 1) of the hamburger shop has three pieces (n =
1,2,3) are acquired, and the restaurant sign A2 (m = 2
Image data of 4 (n = 1,2,3,4) is acquired, and image data of the signboard A3 (corresponding to m = 3) of the motorcycle shop is acquired 2 (n = 1,2). ing.

Here, m is a symbol for identifying an image pickup target, and n is a symbol for identifying image data for the same image pickup target.
That is, m = 1, 2, 3 indicates that the image capturing targets are the signs A1, A2, A3, respectively. Also, n = 1,
2, 3 and 4, the image data P (m_1) Signs A _m, P
(M_2), P (m_3), and P (m_4) are acquired in this order.

As shown in FIG. 8, the color space mapping feature amount extraction process includes a color reduction process (S21) and a pseudo joining amount calculation process (S2).
2), pseudo comprehensive amount calculation processing (S23), and mapping vector generation processing (S24). Below, S21
Each process of S24 to S24 will be described in detail.

Since the color reduction processing of S21 is a well-known and commonly used image analysis technique, detailed description (including mathematical expressions) will be omitted, and a suitable method will be briefly described. For example, the literature "
Digital image processing (Makoto Nagao, Modern Science Company, 1978 12
8), using the median value of RGB as a threshold value, and reducing the total of 8 colors of black, blue, red, purple, green, water, yellow, and white. Image data P (m_
n) is divided into areas for each color.

The reason why the color reduction processing is applied to the image data will be briefly described below. It is extremely difficult to match the color of the image data with the color of the actual image pickup target, or to match the color of the image data with the color intended by the creator of the image pickup target. Therefore, in the color space mapping feature amount extraction processing, a large amount of color reduction processing is performed on the image data to roughly classify the colors of the image data, and then the feature amount is extracted. As this feature amount, a feature amount indicating the arrangement relationship (comprehension and joining) of each color area is used. Hereinafter, this feature amount will be referred to as “color space mapping feature amount”.

Regarding the pseudo-bonding amount calculation processing of S22,
Since this is a characteristic image analysis technique for realizing the discrimination of the imaging target according to the present invention, it will be described in detail below with reference to FIGS. 9 (a) and 9 (b). The pseudo-bonding amount calculation process is a process of obtaining the pseudo-bonding amount between the regions divided in S21.
That is, as shown in FIG. 9A, for example, the base point P
3 × 3 [p including _i (X _i , Y _i ) as the upper left pixel
ixel] box is moved on the image data P (m_n), the color combinations in the box are counted, and the count result is set as the pseudo-joint amount.

Specifically, in order to calculate the pseudo-bonding amount for the box shown in FIG. 9B, the following is done. That is, since the color of the base point P _i is black, the number of combinations of black and colors other than black, based on black, is sequentially counted. Because in the box white pixels exist three pixels, first - to hold a pseudo-bonding amount of 3 × W _C in the combination of "black white".

Note that W _C is a variable numerical value that is multiplied to give arbitrary weighting to the pseudo junction amount. Therefore, W _C = 1 is substituted when weighting is not performed. Similarly, there are 3 and 2 yellow and blue pixels in the box. Thus, - holding the pseudo bonding amount of 3 × W _C in the combination of "Black Yellow", - holding the pseudo bonding amount of 2 × W _C in the combination of "Black Blue". All the pixels of the box are image data P
All pixels are moved within the range of (m_n) as the base point P _i .

Regarding the pseudo inclusive amount calculation process of S23,
Since this is a characteristic image analysis technique for realizing the imaged object determination according to the present invention, it will be described in detail below with reference to FIGS. 10 to 12. The pseudo inclusion amount calculation process is a process of obtaining the pseudo inclusion amount between the regions divided in S21. First, the information providing server 1 reconfigures an area for each label (closed area) by giving an identification number to each divided area. As a result, as shown in FIG. 10, it is assumed that the label L1 (white area), the label L2 (blue area), and the label L3 (yellow area) are generated. Since the identification number is given to each of the divided closed areas, the areas of the same color do not always have the same label.

Subsequently, the information providing server 1 determines that the base point P
_{When i} (X _i , Y _i ) is moved in a plurality of directions (for example, four directions of up, down, left and right) on the image data P (m_n) and the same label is detected in all of the plurality of directions, the label It is determined that the label of the base point P _i is included in pseudo. Then, the combination of the color of the pseudo-included label and the color of the pseudo-included label is counted, and the count result is set as the pseudo-included amount. The movement direction of the base point P _i is not limited to vertical and horizontal directions, but may be any direction such as diagonal direction.

[0055] Specifically, as shown in FIG. 11, with respect to the base point P _1, on the base points P _1, below, there is no different labels on the right, only the different labels on the left side of the base point P ₁ (label L2 ) Exists. Therefore, the identification number "2" of the label L2 is recorded in the item D1 corresponding to the left side of the base point.

Returning to FIG. 10, the base point P _i is sequentially moved in the image data P (m_n) as indicated by the dashed line arrow Y in order to calculate the pseudo inclusion amount with respect to the base point other than the base point P ₁ . The movement interval of the base points P _i may be continuous or discrete (for example, 5 pixel intervals). Also, the base point P _i
Does not always move linearly, but may move curvilinearly or randomly.

When the base point P _i moves to the position of the base point P ₂ shown in FIG. 10, the pseudo comprehensive amount calculation process for the base point P ₂ is executed. In other words, on the base point P _2, below, there are different labels (label L1) is on the left, there are different labels (labels L1, L3) on the lower side of the base point P _2. Therefore, "1" is recorded in items D2, D3, and D5, and item D
In "4", "1, 3" is recorded. Hereinafter, similar processing is executed for the base points P _{3 to} P ₆ .

Subsequently, with respect to the base points P _{1 to} P ₆ , the logical product of the identification numbers of the labels existing in the up, down, left and right directions is obtained,
The identification numbers of the labels existing in all four directions are extracted. The extracted identification numbers correspond to the four directions (Du & Dr &
Dd & Dl) are stored in the corresponding areas of the item D6.

For example, with respect to the base point P ₁ , there is no label existing in all directions of up, down, left, and right, so that "-" indicating that there is no label is present in the item corresponding to the base point P ₁ of the four directions. Will be recorded. Similarly, with respect to the base point P ₂ , the labels existing in all the directions of up, down, left and right are the labels L1.
Therefore, "label L1" is recorded in the item corresponding to the base point P ₂ of the items in the four directions. Further, with respect to the base point P ₅ , the labels existing in all the directions of up, down, left and right are the labels L.
1 and L2, "labels L1 and L2" are recorded in the item corresponding to the base point P ₅ of the items in four directions. As a result, as shown in FIG. 11, the corresponding relationship between the pseudo inclusion label and the pseudo inclusion label is determined.

FIG. 12 is a diagram showing a correspondence relationship between the pseudo inclusion color and the pseudo inclusion color with respect to the base points P _{1 to} P ₆ . For example, referring to E11 and E12 in FIG. 11, the label L1 that pseudo-includes the label L2 is 1 for each item,
There are two in total. Therefore, the blue color (label L2 in FIG.
2 × W _O is recorded (voted) in the white item E1 (corresponding to the label L1 in FIG. 11) that pseudo-includes (corresponding to the item).
Similarly, 1 × W _O is recorded in a blue (corresponding to label L2 in FIG. 11) item E2 that pseudo-includes yellow (corresponding to label L3 in FIG. 11). Note that W _O is a variable value that is multiplied to give arbitrary weighting to the pseudo inclusive amount. Therefore, when weighting is not performed, W _O = 1
Is substituted. As a result, as shown in FIG. 12, the correspondence relationship between the pseudo inclusion color and the pseudo inclusion color is determined.

Here, the number of dimensions when the pseudo-joint amount and the pseudo-inclusive amount are vector-described will be described. The number of dimensions when the pseudo joining amount is described by a vector corresponds to the number of combinations of two colors selected from k colors. In addition, considering that there naturally exists a combination of the same colors, {(k ² −
k) / 2} + k = k (k + 1) / 2. Therefore, when k = 8, the pseudo-joint amount is represented by a 36-dimensional vector. On the other hand, the number of dimensions when the pseudo-inclusive amount is described by a vector is k × k = k ² , as is apparent from the chart of FIG. Therefore, when k = 8, the pseudo inclusive quantity is represented by a 64-dimensional vector.

Returning to FIG. 8, in the mapping vector generation process of S24, the information providing server 1 multiplies the pseudo-joint amount calculated in S22 and the pseudo-inclusive amount calculated in S23 by a weighting coefficient to obtain the color. A mapping vector V (m_n) as a spatial mapping feature is generated. Mapping vector V
(M_n) is the image data P (m_n) as shown in FIG.
One vector is generated for each n). Here, the dimension number of the mapping vector V (m_n) is obtained by summing the dimension number of the pseudo-joint amount and the dimension number of the pseudo-inclusive amount, and when k = 8, 36 + 64 = 100 dimensions.

In the color space mapping feature amount extraction process, the information providing server 1 executes only one of the pseudo joint amount calculation process of S22 and the pseudo inclusion amount calculation process of S23. It may be one. However, the number of dimensions of the mapping vector V (m_n) in this case is
As a matter of course, it depends only on the number of dimensions of the calculated feature amount. That is, when only the pseudo-joint amount calculation process is executed, the number of dimensions of the mapping vector V (m_n) is "k (k +
1) / 2 "dimensions, the number of dimensions of the mapping vector V (m_n) is" k × k = k "when only the pseudo inclusive amount calculation process is executed.
² ".

Returning to FIG. 6 again, the information providing server 1 executes S
One set of mapping vectors V (m_n) generated in 24 (2
Each vector) is mapped to a high-dimensional feature space (hereinafter referred to as "high-dimensional space") to generate a discrimination vector and a bias (S3). Regarding the processing for generating the discrimination vector and the bias from the mapping vector, for example, SVM
There are various methods such as (Suport Vector Machine), hidden Markov model, neural network, and decision tree. In this embodiment, SVM suitable for high-speed binary discrimination processing of high-dimensional vector with excellent generalization performance is used. It shall be.

SVM is based on the document "Support-Vector Network".
s, Corinna Cortes, Vladimir Vapnik, Machine Learnin
g, 20, pp.274-297, Mar. 1995. "and the like, which are well known conventional techniques, and detailed description (including mathematical formulas) is omitted.
Is an algorithm for converting a feature vector (mapping vector V (m_n)) into a high-dimensional space using a non-linear conversion function and performing linear discrimination in the high-dimensional space.

Next, the information providing server 1, as shown in FIG. 7, maps the mapping vector V in the high-dimensional space S (m1, m2).
Boundary dividing (m_n) (broken lines K1 to K3 in FIG. 7)
Of the discriminant vector V _SV
Calculate (m1, m2) and bias b (m1, m2). The calculated discriminant vector and bias are stored in a predetermined area of the storage unit 7. In FIG. 7, for example, the discrimination vector V _SV
(1,2) is the image data P (1_n) and the image data P (2
_n) is a vector for determining which is similar to the determination target data (imaging data). Bias b
The same applies to (1,2).

In FIG. 7, the mapping vector V (1_n) and the mapping vector V (2_2 in the high-dimensional space S (1,2) are used.
n) and an example in which a mapping vector corresponding to one image data (hamburger image data) and a mapping vector corresponding to one image data (knife and fork image data) are divided. Indicated. However, for example, in the high-dimensional space S (1,2 & 3), the mapping vector V
(1_n) is divided into the mapping vector V (2_n) and the mapping vector V (3_n), and the mapping vector corresponding to one image data and the mapping vector corresponding to other image data are divided. It may be done.

The process of S3 does not necessarily have to be performed for all combinations of image data, but may be performed only for combinations of image data of imaging targets whose service ranges geographically overlap. The reason for this will be described below. If the plurality of service ranges do not have areas that overlap each other, the number of signboards that the user of the mobile phone 10 located in any of the service ranges can image is specified as one. Therefore, the process of discriminating the image data similar to the imaged data from the plurality of image data already registered in the information providing server 1 becomes unnecessary. Therefore, it is effective to determine the presence or absence of overlapping areas in order to quickly determine the imaging target.

Therefore, referring to FIG. 13, the position P_m1, P
_m2Regarding the service range of the two signboards installed in
A method for determining the presence / absence of the area will be described. P_m1
(X _m1, Y_m1) And P_m2(X_m2, Y_m2) And the distance D
_{m1, m2}Is D_{m1, m2}= √ {(X_m1-X_m2)²+ (X_m1-X
_m2)}²It is expressed as [m]. At this time, the above two signs
Radius of each service range is R_m1', R_m2'Is
Then, the service ranges of these two signs overlap.
The condition (including contact) is D_{m1, m2}≤ R_m1'+ R_m2'
Is. Therefore, D_{m1, m2}≤ R_m1'+ R_m2Meet '
Only if the discrimination vector and buy
It suffices to carry out the process of generating asbestos.

Further, the information providing server 1 generates a discrimination vector and a bias as data for discriminating the similarity between the image data and the image pickup data prior to discriminating the image pickup object, and stores them in the storage unit 7. deep. That is, as described above, a large number of image data P are used to determine the imaging target.
(M_n) is required, but it is not always necessary to store all of these image data themselves (excluding image data for presentation). This facilitates data management.

When an image pickup target is added (or deleted) in order to improve the completeness and variety of information accessible by the user, data (discrimination vector and discrimination vector) relating to an area overlapping the service range of the image pickup target is added. Only the bias) needs to be updated. As a result, it is possible to easily update the imaging target, and it is possible to construct a highly flexible information providing system with no break in the service range.

When the image learning phase is completed, the process shifts to the imaging object discrimination phase. First, in step S4 of FIG. 6, the mobile phone 10 sets the signboard A2 as an image pickup target to the image pickup device 16
To image. The signboard A2 is imaged so as to fill the display area of the display device 14. Thereby, the correlation value between the imaged object and the imaged data is increased, and the imaged data can be converted into the image feature amount and used for the determination without being processed. As a result, the accuracy rate of object discrimination is improved.

When the image pickup target is picked up, it is feared that the aspect ratio of the image pickup target does not match the aspect ratio of the display area. However, if the image pickup is performed so that either one of the vertical and horizontal sides matches the display area. For example, the imaging data including the background of the imaging target is acquired.
Regarding image data as well, in the image learning phase, if the same image capturing method is used, the target determination can be performed without partially cutting out the image capturing target from the image capturing data.

In S5, the mobile phone 10 estimates P _U shown in FIG. 14 as the current position of the mobile phone 10 by the position estimation device 18. In S6, the mobile phone 10 transmits the image pickup data acquired in S4 and the position information indicating the image pickup position P _U estimated in S5 to the information providing server 1. In addition,
The position information is, for example, GPS coordinates, and may include the type of position estimation method and position estimation error. In S7, the information providing server 1 receives the imaging data and the position information from the mobile phone 10.

In S8, the information providing server 1 selects the service ranges C1, C2, C3 including the image pickup position P _U indicated by the position information, based on the position information received in S7.
In S9, the information providing server 1 determines the image data P (1_n) of the service range C1, C2, C3 selected in S8,
The discriminant vectors and biases corresponding to P (2_n) and P (3_n) are extracted from the storage unit 7. The discrimination vector and the bias are extracted in consideration of the position estimation error of the mobile phone 10. As a result, the discrimination vector regarding the imaging target that is not imaged at the imaging position P _U is not extracted, and the efficiency of the discrimination process can be improved.

That is, in S9, as shown in FIG.
The service areas C1 and C2 have areas overlapping with each other.
Therefore, the discrimination vector V_SV(1, 2) and bias b (1,
2) is extracted. Service ranges C1 and C3 also overlap each other
The discriminant vector V has an overlapping area._SV(1,3)
And the bias b (1,3) are extracted. Similarly, service
The areas C2 and C3 also have areas that overlap each other.
Therefore, the discrimination vector V _SV(1,3) and bias b
(2,3) is extracted.

In S10, the information providing server 1 generates a mapping vector from the image pickup data received in S7. The processing can be performed by executing the same color space mapping feature amount extraction processing (S21 to S24) as the image data on the imaged data. Detailed processing contents are omitted because they have already been described. As a result, as shown in FIG. 15, a mapping vector V (cap) is generated from the imaged data P (cap).
Here, (cap) indicates that the data indicated by the symbol relates to imaging data, and is distinguished from the data regarding image data P (m_n).

The mobile phone 10 may perform the processing of S10, that is, the processing of generating the mapping vector from the imaged data, and transmit the generated mapping vector to the information providing server 1. As a result, the processing load of the mobile phone 10 increases, but the communication capacity can be reduced.

In S11, the information providing server 1 discriminates the object to be imaged based on the discrimination vector extracted in S9 and the mapping vector generated in S10. Specifically, the scalar product is obtained by taking the inner product of the discrimination vector V _SV (1,2) and the mapping vector V (cap), and the bias b (1,2) is added to this. If the addition result is 0 or more, "1" is voted for the image data (1_n) of the signboard A1, and if the addition result is less than 0, "1" is voted for the image data (2_n) of the signboard A2.

Below, the discrimination vectors V _SV (1,3), V
The same process is repeated for _SV (2,3). That is, the discrimination vector V _SV (1,3) and the mapping vector V (cap)
The inner product of and and the bias b (1,3) are added. If the addition result is 0 or more, "1" is added to the image data of the signboard A1.
Is voted, and if the addition result is less than 0, "1" is voted for the image data of the signboard A3. Further, the discrimination vector V
The inner product of _SV (2,3) and the mapping vector V (cap) is calculated, and the bias b (2,3) is added. If the addition result is 0 or more, "1" is voted for the image data of the signboard A2, and if the addition result is less than 0, "1" is voted for the image data of the signboard A3.

As a result of the binary discrimination, the signboard corresponding to the image data having the largest number of votes is discriminated as the image pickup object. Here, when the type of image data registered in the image learning phase is M, the number of binary discriminations is M (M-1) / 2.
[Times]. Therefore, when three types of image data are registered as in the present embodiment, the number of binary discriminations is three. When M is 4 or more, the signboard is not limited to the signboard corresponding to the image data having the largest number of votes, and a plurality of signboards corresponding to the image data having the third largest votes are selected as candidates for the imaging target. You may.

As described above, both the discrimination of the image pickup target based on the pseudo-joint amount and the discrimination of the image pickup target based on the pseudo inclusion amount use the feature amount regarding the color of the image pickup data. Therefore, compared to the determination of the imaging target based on other feature amounts (periodicity, shape, texture, edge, texture edge, etc.), it is less likely to be affected by the projective distortion and the imaging direction due to the deviation of the imaging position. In particular, for signboards that are artifacts, since the creators are often painted in multiple colors to effectively convey information to others, use the feature values related to the color arrangement and color arrangement relationship. Is preferred. Further, when determining the imaging target in the present embodiment,
Since it is not necessary to recognize what the imaging target is, the imaging target does not matter. Further, there is an advantage that the discrimination accuracy is improved by combining the pseudo joining amount and the pseudo inclusion amount.

When the imaging object discrimination phase ends, the information presentation phase starts. First, in S12 of FIG. 6, the information providing server 1 extracts the image data and its related information from the database 71 based on the determination result of S11. At this time, only one of the image data and the related information may be extracted. In S13, the information providing server 1 transmits the extracted image data and related information to the mobile phone 10.

At S14, the mobile phone 10 receives the image data and the related information transmitted from the information providing server 1. In S15, the mobile phone 10 displays the image data and the related information received in S14 on the display device 14, and ends the series of information providing processing.

The related information displayed in S15 is, for example, a store name and business hours, but other than this, hyperlink information or keyword information may be displayed. The hyperlink information is information that enables access to another information source (Web site or the like) designated by the address via the network N. The keyword information is information that allows more relevant information to be obtained by a search using the related keyword. Further, if a function of automatically bookmarking related information is added, classification and management by category (for example, a restaurant when the related information is a hamburger shop) becomes possible, and reuse of related information becomes easy.

The form of the information presented in S15 is
For example, characters, still images, moving images, audio, etc. are arbitrary. Since the information to be presented is selected based on the above-mentioned image pickup data, the user of the mobile phone 10 does not need to perform a key or numeral input operation when acquiring the information. Therefore, the labor and time involved in the operation are saved, and the operability of the mobile phone at the time of obtaining the desired information is improved.

As described above, according to the information providing system 100 of the present invention, the image data and the related information are registered in the database 71 of the information providing server 1 in association with each other. The information providing server 1 discriminates the image data most similar to the imaged data with the acquisition of the imaged data captured by the mobile phone 10, retrieves and presents the information corresponding to the image data. As a method of discriminating image data, for example, a method of extracting a color space mapping feature amount based on the joining amount / inclusive amount of color regions acquired from the imaged data and comparing it with the color space mapping feature amount of existing image data is used. To use. The color space mapping feature amount contributes to highly accurate target determination because it is less affected by image distortion, image pickup position shift, and the like.

Further, even when the user of the mobile phone 10 lacks knowledge about desired information or forgets it, he / she can easily obtain the desired information by a simple operation of photographing and transmitting the surrounding object. Available at. Further, the terminal that the user can freely carry can timely acquire only the information that matches the taste and purpose of the user from a wide variety of information mixed in the real world and the network.
As a result, the affinity between the network where information is exchanged and the user of the mobile phone is increased, and it is possible to provide the information service that suits the individual.

In order to improve the completeness and convenience of information retrieval, it is desirable that the service is provided in a wide range.
However, if the service provision range is expanded too much, the number of imaging targets increases and the information retrieval time becomes huge. Therefore, by once narrowing down the number of candidates for the imaging target based on the position where the imaging target is imaged by the mobile phone 10, it is possible to perform an information search that combines the current position of the user and the imaging data. As a result, the search time is shortened.

The object to be imaged may be any object that can be imaged by the mobile phone 10 and converted into electronic data, and for example, a signboard or a poster is suitable. The reason is as follows. First, the signboard selected by the user is likely to match the user's taste and purpose of use. Second
In addition, as for the information advertised by the signboard, a lot of related information exists on networks such as the Internet. Third, most of the signs have a fixed geographic location,
It is easy to grasp the position information. Fourthly, there are a large number of outdoors, which are supposed to be the main places of use of highly portable mobile terminals.

(Second Embodiment) Next, the mobile phone 10
A second embodiment in which a mode in which communication with the information providing server 1 is unnecessary (hereinafter, referred to as “offline mode”) is applied to the information providing system will be described.

As a premise, the difference between the online mode and the offline mode is clarified. FIG. 16 is a diagram showing differences between the modes by comparing the features of the online mode and the offline mode. As shown in FIG. 16, the online mode is a mode in which communication with the information providing server 1 is required, and since the accurate position estimated by the mobile phone 10 is notified to the information providing server 1, it can be used outdoors in a wide area. Suitable for use.

On the other hand, the off-line mode is a mode in which communication with the information providing server 1 is unnecessary, and the related information peculiar to the area is mainly transmitted by using a notification of a specific area (area ID etc. as a trigger). Cell or service area). The specific area includes the inside of the building. At the same time, the offline mode is intended for use outside the service area where the information providing server 1 cannot communicate. By appropriately switching and using each of these modes according to the location of the user, it is possible to provide the imaging target determination technology according to the present invention in a wide range.

First, the configuration of the information providing system in the second embodiment will be described. FIG. 17 is a schematic diagram showing an example of the overall configuration of the information providing system 200 in this embodiment. As shown in FIG. 17, the information providing system 200
Is roughly configured by a data providing device 201 and a mobile phone 210 (corresponding to an image discrimination device). The data providing apparatus 201 is wire-connected to the access point B201. Furthermore, the mobile phone 210 can wirelessly communicate with the access point B201. Therefore, the mobile phone 210 can exchange information with the information providing server 201 via the access point B201.

The basic configuration of the data providing apparatus 201 is the information providing server 1 described in detail in the first embodiment.
Since the configuration is the same as that of 1., each component is denoted by the same reference numeral, and the illustration and description of the configuration are omitted. That is,
The data providing device 201 includes a CPU 202 and an input unit 20.
3, display unit 204, communication control unit 205, RAM 206,
And a storage unit 207, and each unit is connected by a bus.

The storage unit 207 has a database 217. In the database 217, image data of all the imaging targets A21 to A28 that can be imaged in the specific area Z and related information are stored in advance in association with each other. Further, the database 217 stores a discrimination vector and a bias (hereinafter collectively referred to as “discrimination data”) for discriminating image data similar to the imaged data from the image data. However, unlike the first embodiment, the data indicating the service range is not stored in the database 217.

The basic structure of the mobile phone 210 is the same as that of the mobile phone 10 described in detail in the first embodiment except that the position estimating device is not provided. The reference numerals are assigned and the illustration and description of the configuration are omitted. That is, the mobile phone 210 has the control device 211.
(Corresponding to the dividing unit, the joining amount calculating unit, the comprehensive amount calculating unit, and the generating unit), the input device 212, the RAM 213, the display device 214, the storage device 215, the imaging device 216, the wireless communication device 217, and the voice processing device 219. It is equipped with. Further, these respective devices are electrically connected to each other via a bus so that various signals can be input and output.

Next, referring to FIG. 18, the second embodiment of the present invention will be described.
The operation of the information providing system in the embodiment will be described, and also the image discrimination method according to the present invention will be described. In each step shown below, the program stored in the storage unit 207 or the control device 211 is executed by the CPU 202.
Alternatively, it is realized by being executed by the control device 211.

Since each processing of T1 to T3 is the same as S1 to S3 of FIG. 6 described in detail in the first embodiment, the description thereof will be omitted. At T4, a transmission request for the data stored in the database 217 is transmitted from the mobile phone 210 to the data providing apparatus 201. Access point B201 installed near entrance F of specific area Z
The notification information is always sent from. The data transmission request is automatically transmitted to the mobile phone 210 when the mobile phone 210 detects the presence of an application for discriminating an imaging target (hereinafter referred to as “discrimination application”) based on the notification information. To be done. Note that the data transmission request may be transmitted when the user of the mobile phone 210 performs an operation for instructing acquisition of the determination application.

At T5, the data providing apparatus 201 acquires the discrimination application and the discrimination data from the database 217 upon receiving the data transmission request via the access point B201. At the same time, the image data and the related information are acquired from the database 217 while maintaining the corresponding relationship. At T6, the data providing apparatus 201 transmits the discrimination application, the discrimination data, the image data, and the related information to the mobile phone 210.

At T7, the mobile phone 210 sends the discrimination application, the discrimination data, the image data, and the related information from the data providing apparatus 201 to the access point B20.
Receive via 1 (download). Note that these data are stored in the mobile phone 210 and the other mobile phones 220, 2
It may be distributed between 30 and 30. Such distribution of data between mobile phones is particularly effective when the purpose of using the data is limited.

After that, each processing of T8 to T12 is performed in S4, S10, S11, S12 and S15 described with reference to FIG.
Since it is the same as each processing of, the detailed description is omitted.
The mobile phone 210 uses the data acquired from the data providing apparatus 201 to perform the offline imaging target determination processing by the mobile phone 210.

The information acquisition processing in the present embodiment is intended for use in the offline mode of the information providing system according to the present invention. In the off-line mode, the use area and the use purpose are limited, but the user of the mobile phone surely exists in the area (specific area Z). Therefore, it is not necessary to perform the position estimation process of the mobile phone and the process of narrowing down the image data based on the position information, and a more immediate information acquisition service can be provided to the user.

Since the information acquisition process is executed on the side of the mobile phone 210, the storage capacity is a concern. In this respect, when the capacity of the discrimination data of 1 is managed by 4 [byte] and the discrimination vector V _SV of one pair (two) of the imaging target is expressed in 96 dimensions and the bias b _SV is expressed by the scalar, 100 kinds of imaging The capacity of the discrimination data capable of discriminating the target is (96 [dimension] +1) × 4 [b
yte] × ₁₀₀ C ₂ = about 2 [MB].

The description content in this embodiment is a preferred example of the mobile terminal according to the present invention, and the present invention is not limited to this. For example, the mobile phones 10 and 210 are
Any electronic device having a communication function such as a PDA (Personal Digital Assistance) may be used.

Finally, a program for realizing the information providing and acquiring technique according to the present invention and a computer-readable recording medium (hereinafter, simply referred to as “recording medium”) recording the program will be described. A recording medium causes a reading device provided in a hardware resource such as a general-purpose computer to cause a change state of energy such as magnetism, light, and electricity in accordance with the description content of a program, and to generate a signal corresponding to it. In the format, the description content of the program can be transmitted to the reader. As such a recording medium, for example, an IC card such as UIM, a magnetic disk, an optical disk, a magneto-optical disk that is detachably attached to a computer (including a mobile phone, a PHS, etc.), or is fixed to the computer. A hard disk (HD) built into the hard disk, a nonvolatile semiconductor memory such as firmware fixed integrally, and the like are applicable.

Further, a part or all of the above-mentioned program is transmitted from another device via a transmission medium such as a communication line.
The wireless communication device included in the mobile terminal according to the present invention may receive and record the received data. On the contrary, the program may be transmitted from the mobile terminal according to the present invention to another device via a transmission medium and installed.

[0108]

According to the present invention, the image data is divided into a plurality of color-based areas, the pseudo-joint amount between these color-based areas is calculated, and a mapping vector is calculated from the image data using the pseudo-joint amount. Is generated. Therefore, this mapping vector reflects at least the pseudo-joint amount between the color-based regions forming the image data as a feature amount.

By applying the above-described processing of generating the mapping vector to the image data and the image pickup data, the influence of the projective distortion and the image pickup direction due to the shift of the image pickup position of the image pickup data becomes less likely to occur. As a result, it becomes possible to accurately determine the similarity between image data.

[Brief description of drawings]

FIG. 1 is a diagram conceptually showing an information providing service realized by an information providing system according to the present invention.

FIG. 2 is a schematic diagram showing an example of the overall configuration of an information providing system in the first embodiment.

FIG. 3 is a block diagram showing a functional configuration of an information providing server according to the first embodiment.

FIG. 4 is a bird's-eye view schematically showing the positional relationship of the service range of the imaging target.

FIG. 5 is a block diagram showing a functional configuration of a mobile phone according to the first embodiment.

FIG. 6 is a flowchart showing a flow of information providing processing executed by the information providing system according to the first embodiment.

FIG. 7 is a diagram conceptually showing a series of processes from calculation of a discrimination vector from image data.

FIG. 8 is a flowchart showing a color space mapping feature amount extraction process.

FIG. 9A is a diagram conceptually showing a pseudo joining amount calculation process. FIG. 9B is a diagram showing an example of a box used in the pseudo joining amount calculation process.

FIG. 10 is a diagram conceptually showing pseudo inclusive amount calculation processing.

FIG. 11 is a diagram showing, for each base point, the correspondence relationship between the identification number of the pseudo inclusion label and the identification number of the pseudo inclusion label in the pseudo inclusion amount calculation process.

FIG. 12 is a diagram showing a correspondence relationship between a color that is included in a pseudo manner and a color that is included in a pseudo manner in a pseudo inclusion amount calculation process.

FIG. 13 is an overhead view schematically showing a positional relationship between two service ranges that partially overlap with each other.

FIG. 14 is an overhead view schematically showing a positional relationship between a service range of an image pickup target and a mobile phone.

FIG. 15 is a diagram conceptually showing a series of processes for determining the similarity between imaged data and image data.

FIG. 16 is a diagram showing a difference between an online mode and an offline mode.

FIG. 17 is a schematic diagram showing an example of the overall configuration of an information providing system according to a second embodiment.

FIG. 18 is a flowchart showing a flow of information acquisition processing executed by the information providing system in the second embodiment.

[Explanation of symbols]

1 ... Information providing server, 2 ... CPU, 5 ... Communication control unit, 7
… Storage, 71… Database, 10… Mobile phone, 11
... Control device, 14 ... Display device, 16 ... Imaging device, 18 ...
Position estimation device, A1 to A3 ... Signboard, C1 to C3 ... Service range, Z ... Specific area

Front page continuation (51) Int.Cl. ⁷ Identification code FI theme code (reference) H04N 1/60 H04N 1/40 D (72) Inventor Setsuyuki Hongo 2-11-1, Nagatacho, Chiyoda-ku, Tokyo F term in NTT Docomo, Inc. (reference) 5B057 BA11 CA01 CA16 CB01 CB17 CE09 CE10 CE17 CH01 DA08 DB06 5C077 MM03 MP08 PP32 PQ12 PQ17 PQ22 SS06 5C079 HB01 JA23 LA02 LA10 LB12 MA11 NA29 5L096 FAA02 CA02 FA02 CA02 FA02 CA02 FA02

Claims (10)

[Claims] 1. A dividing unit that divides image data into a plurality of regions for each color, a joining amount calculation unit that calculates a pseudo joining amount between the plurality of color regions, which is divided by the dividing unit, and the joining amount. An image discriminating apparatus comprising: a generating unit that generates a mapping vector from the image data by using the pseudo-joint amount between the plurality of color regions calculated by the calculating unit. 2. The pseudo joining amount is 1 on the image data.
Within multiple pixels including the pixel as the base point,
The image discriminating apparatus according to claim 1, wherein the image discriminating apparatus is a result of counting the combination of the color of the base point and the color of each pixel other than the base point. 3. An inclusion amount calculation unit for calculating a pseudo inclusion amount between a plurality of color-separated areas divided by the dividing unit, wherein the generation unit includes the pseudo joining amount and the inclusion amount calculation unit. The image discrimination apparatus according to claim 1, wherein a mapping vector is generated from the image data by using the calculated pseudo inclusion amount between the plurality of color regions. 4. The pseudo inclusion amount is a closed area including the base point when the plurality of color-based areas are reconstructed as a plurality of closed areas and one pixel on the image data is used as a base point. The image discriminating device according to claim 3, wherein the image discriminating device is a result of counting combinations with closed regions existing in all directions of a predetermined direction with the base point as a center. 5. The image discriminating apparatus according to claim 1, wherein the image data is data obtained by capturing an image of a signboard. 6. An image discriminating method in which an image discriminating apparatus discriminates the similarity of a plurality of image data, a dividing step of dividing the image data into a plurality of regions for respective colors, and a plurality of divided portions obtained in the dividing step. A mapping amount is calculated from the image data using a bonding amount calculation step of calculating a pseudo bonding amount between the color regions and a pseudo bonding amount between the plurality of color regions calculated in the bonding amount calculation step. An image discrimination method comprising: a generation step. 7. The pseudo joining amount is 1 on the image data.
Within multiple pixels including the pixel as the base point,
7. The image discrimination method according to claim 6, wherein the result is obtained by counting the combination of the color of the base point and the color of each pixel other than the base point. 8. The image discriminating apparatus further includes an inclusion amount calculation step of calculating a pseudo inclusion amount between the plurality of color-based areas divided in the dividing step, and in the generating step, the pseudo joining amount and The image discrimination method according to claim 6, wherein a mapping vector is generated from the image data using the pseudo inclusion amount between the plurality of color-based regions calculated in the inclusion amount calculating step. 9. The pseudo inclusive amount is a closed area including the base point when the plurality of color-based areas are reconstructed as a plurality of closed areas and one pixel on the image data is used as a base point. 9. The image discriminating method according to claim 8, wherein the image discriminating method is a result of counting combinations with closed regions existing in all directions of the predetermined direction with the base point as the center. 10. The image discrimination method according to claim 6, wherein the image data is data obtained by capturing an image of a signboard.