JPH0735509A - Postage stamp detecting device - Google Patents

Abstract

PURPOSE:To detect postage stamps having several different colors and having different sizes with a high degree of accuracy by providing means for inputting a color image, scanning means for detecting a color variation and specifying a stamp existing area and the like. CONSTITUTION:A color image input device 4 inputs a color image on a postal matter P on which a postage stamp is applied. A CPU 1 removes a background part from the inputted color image so a5 to cut out of the postal matter P. Further, for detection of colors, the surfaces for red, green and blue components of the image are subjected to first-order differentiation such as first-order color differentiation, a maximum value of absolutes values of the first order differentiation, is set to a first-order differentiated value of the pixel thereof. The first- order differentiated color image is subjected to a binarization process, and first binary images of the postal matter 2 are searched in an oblique direction at an angle of 45deg. Through this search, the position of a first pixel is set at a starting position of the postage stamp. Further, the postage stamp is successively cut out, an entire pregiven area as a postage stamp search area is searched, and thereafter, the process is completed. These control precesses are carried out by the CPU 1.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a postage stamp detecting device for detecting a postage stamp attached to a postal matter conveyed at a constant speed.

[0002]

2. Description of the Related Art Conventionally, when a stamp on a postal matter is detected,
A stamp is discriminated by reading a specific line on the mail and detecting the state of change of the signal. For example, the stamp detecting device described in Japanese Patent Publication No. 47-18368 will be described. An ordinary postage stamp has a white portion of 1 mm or more around which a perforated cut is made and inside the stamp. Is drawn with an arbitrary pattern including chromatic colors depending on the type of stamp. Pay attention to this perforated white frame and the chromatic color of the very narrow part adjacent to it, regardless of the pattern inside the stamp, or conversely, that it is a stamp from the original pattern of the stamp. The type of stamp is determined from the color or combination of colors of the stamp.

[0003]

However, in the conventional device as described above, only the color information of a specific line is used to detect a stamp, and therefore, an advertisement having a similar color pattern on a mail is detected. If there is information or the like, it cannot be distinguished from it, and only stamps of color patterns registered in advance can be detected. Then, the two-dimensional area information of the stamp cannot be obtained, and the range in which the stamp exists is not clear.

Therefore, the present invention is applicable not only to stamps having a specified pattern such as ordinary stamps, but also to stamps having various color patterns, such as memorial stamps and special stamps, or stamps of different sizes. An object of the present invention is to provide a stamp detecting device that can detect with high accuracy.

[0005]

SUMMARY OF THE INVENTION A stamp detecting device of the present invention is a color image inputting means for inputting a color image on a postal matter on which a stamp exists, and a specific position of the color image input by the color image inputting means. Scanning means for detecting a color change point on the mail by scanning in an oblique direction from the above, and based on the color change point detected by the scanning means, the color image input means inputs the color change point. Area specifying means for specifying the area where the stamp is present on the postal matter.

Further, the stamp detecting device of the present invention has a color image inputting means for inputting a color image on a postal matter on which a stamp exists, and a diagonal direction from a specific position of the color image input by the color image inputting means. First scanning means for detecting a color change point on the mail by scanning with reference, and a position separated from the first color change point detected by the first scanning means by a predetermined distance, The second color change on the postal matter from the color image of the postal matter inputted by the color image input means by scanning with reference to the second diagonal direction which is the opposite direction to the first diagonal direction. Second scanning means for detecting a point, the first color change point detected by the first scanning means, and the second color change point detected by the second scanning means On the mail input by the color image input means And it includes a region specifying unit for specifying the presence area of the hand.

Further, the stamp detecting device of the present invention has a color image inputting means for inputting a color image on a postal matter on which a stamp exists, and an oblique direction from a specific position of the color image input by the color image inputting means. Scanning means for detecting color change points on the mail by scanning with reference, and on the mail inputted by the color image input means based on the color change points detected by the scanning means. Area specifying means for specifying the area where the stamp exists, inclination detecting means for detecting inclination information of the area based on position information of the area where the stamp exists specified by the area specifying means, and detection by the inclination detecting means Inclination correction means for correcting the position information of the area where the stamp specified by the area specifying means exists based on the obtained inclination information.

[0008]

[Function] By detecting the color change point from the color image on the mail and detecting the stamp as an area based on the color change point, not only the stamp in which the pattern like a normal stamp is specified, Stamps with various color patterns, such as commemorative stamps and special stamps, can be detected with high accuracy.

By detecting not only the start point but also the end point of the stamp area, for example, a memorial stamp or a special stamp having a different stamp size is not affected by the change in magnification of the optical system of the color image input means. Stamps can also be detected with high accuracy.

By detecting not only the start point but also the end point of the stamp area, the inclination of the stamp and further the inclination of the postal matter can be detected, and the inclination can be corrected. Further, since the two-dimensional image is used for detecting the stamp on the postal matter, the two-dimensional area information of the stamp is obtained and the range of the stamp becomes clear.

[0011]

Embodiments of the present invention will be described below with reference to the drawings. FIG. 1 shows the configuration of a stamp detecting device according to this embodiment. In the figure, a CPU 1 executes a program in a program storage area in a main storage unit 2, a storage device 3, a color image input device (for example, a color television camera or a color scanner device) 4, and The image output device 5 is controlled. Data input / output to / from the color image input device 4 and the image output device 5 is performed via an input interface unit 6 and an output interface unit 7, respectively. A data input device 8 such as a keyboard is provided as a means for inputting data from the outside.

The main memory 2 has a program storage area and a temporary storage area. In the program storage area, a processing program shown in a flowchart described later is stored in advance. The temporary storage area is
It is used as a temporary storage area for the CPU 1. When the image input request command is sent from the CPU 1 to the color image input device 4, the color image input device 4 optically outputs a color image prepared in advance, that is, a color image on the mail P conveyed at a constant speed. Read it. The color image data read by the color image input device 4 is sent to the storage device 3 via the input interface 6 and the CPU bus 9 and stored therein.

When the image output request command is output from the CPU 1, the color image data stored in the storage device 3 is sent to the image output device 5 via the CPU bus 9 and the output interface unit 7. , Output to any device. All of these controls are performed by the CPU 1.

Next, the stamp detecting process in the above-mentioned structure will be described with reference to the flow chart shown in FIG. First, in the postal matter color image input process of step S1, the color image on the postal matter P to which the stamp is attached is input by the color image input device 4, digitized, and stored in the storage device 3. FIG. 3 shows an example of a part of the color image data of the postal matter P stored in the storage device 3. FIG. 3 shows only the red component data in the color image, and the green component data and the blue component data have the same distribution.

Next, in the postal matter cutting process of step S2, the postal matter P is cut out by removing the background portion from the input color image of the postal matter P. For example, in the color image data of FIG. 3, if the background is black, the portion (hatched portion) 15 in which black data is continuously distributed is removed.

Specifically, for example, if the black data level range is set to "0 to 30", a region in which the data of this level range is continuous vertically or horizontally, for example, 10 pixels or more is removed. FIG. 4 shows an example of the cut-out image data of postal matter in which the background portion is removed from FIG.

Next, the color change point detection processing in step S3 will be described. This color change point detection processing, for example,
The process is performed according to the flowchart shown in FIG. The color change point is a point changing from a certain color to a certain color, for example, a point changing from red to blue. When a color change point is detected from the image shown in FIG. 6A, the edge of the image is extracted as shown in FIG. 6B.

Various methods are conceivable for detecting such a color change point, but in the present embodiment, the detection of the color change point by the primary color differential is performed by expanding the edge extraction by the primary differential of the grayscale image into the color image. I will describe. The primary color differentiation described here is the following processing. That is,
First, in step S11, the red component of the color image,
First-order differentiation is performed on each surface of the green component and the blue component. Next, in step S12, the maximum absolute value of the primary differential values in each color component surface of each pixel is set as the primary color differential value of that pixel.

More specifically, for example, consider a 3 × 3 first-order differential matrix as shown in FIG. Figure 7
(A) is a matrix 701 of the first-order differential in the horizontal direction, FIG.
(B) is a matrix 702 of the first-order differential in the vertical direction.
The first-order differentiation of the image is performed by the procedure shown in the flowchart of FIG. 8, for example.

In FIG. 8, description will be made by taking as an example the case where the first-order differentiation of the cut-out image data (red component) of the postal matter shown in FIG. 4 is performed. The first derivative is the coordinate position (5,
It starts from the pixel of 4). The coordinate position referred to here indicates the coordinates of the image stored in the memory. First, in step S21, a lateral differential matrix 701 is formed in a 3 × 3 area centered on the pixel of (5, 4).
Is multiplied and all the calculation results are added to obtain the horizontal first-order differential value of the pixel.

Next, in step S22, a vertical matrix 702 is applied to a 3 × 3 area centered on the same pixel, and all calculation results are added to obtain a vertical primary differential value. Next, in step S23, a value having a larger absolute value out of the horizontal primary differential value or the vertical primary differential value is set as the primary differential value of the pixel.

Next, in step S24, it is judged whether or not the above processing is completed up to all pixels, in this example, the coordinate position (15, 19), and if so, this primary differentiation processing is carried out. To finish. If not completed, the process proceeds to the next pixel in step S25, and the process from step S21 is repeated.

The point to be noted is that the coordinate position (4,
3), (4, 4) and other pixels at the edges of the image cannot be subjected to the first-order differentiation, and there is no first-order differentiation value. FIG. 9 shows the calculation result of the primary differential value of the red component data of the pixel at the coordinate position (5, 4). The red component data 901 of the target pixel is extracted from the cut-out image of the postal matter, and the lateral direction differential matrix 701 is obtained.
The respective corresponding pixels are multiplied by respective corresponding pixels, and all the calculation results are added to obtain a horizontal first-order differential value 903.

Similarly, in the vertical direction, the red component data 901 of the target pixel is extracted from the cut-out image of the postal matter, each component of the vertical differential matrix 702 is multiplied by each corresponding pixel, and all calculation results are added. Vertical direction first derivative value 90
Set to 5. Finally, the larger of the absolute values of the horizontal primary differential value 903 and the vertical primary differential value 905 is the primary color differential value 906 of the red component in the pixel at the coordinate position (5, 4).
And

The same applies to the green component and the blue component,
Finally, the maximum absolute value among the primary color differential values of the red, green and blue components is set as the primary color differential value of this pixel. FIG. 10 shows an arithmetic expression of the first derivative.

The square sum of each color component may be used as the primary color differential value of the pixel, instead of the maximum value of the primary differential value of each color component. Also, here, the first-order differentiation of the red component, the green component, and the blue component was performed.
It may be converted into a C value or the like, and first-order differentiation with respect to H, that is, the hue may be performed to obtain the first-order color differential value of the pixel.

By the above processing, the primary color differential image of the color image is obtained. Then, in step S13,
Binarization processing of this primary color differential image is performed. Although various methods have been proposed for the binarization process, the threshold value for binarization is fixed in this embodiment. For example, when the first-order differential matrix has 3 × 3 values as shown in FIG. 7, the threshold value is “130”. FIG. 11 shows an example of a binarized image of postal matter obtained by binarizing the cut-out image of postal matter of FIG.

Next, the processing for cutting out the first stamp in step S4 will be described. The processing for cutting out the first stamp is performed, for example, according to the flowchart shown in FIG. First, in step S31, a diagonal search is performed. That is, the binarized image of the postal matter obtained earlier is searched in the direction of an oblique 45 °, as indicated by the arrow in FIG. In this case, the starting point 1201 of the search is the pixel at the upper left near the stamped position in the postal matter, that is, the pixel at coordinates (4, 3) in FIG.

If a "1" pixel is detected in step S32, the process proceeds to step S33, as shown in FIG.
As shown in (b), the pixel position is set to the first stamp 12
The start point 1203 of 02 is set. Generally, since the size of a stamp is constant, the size of the stamp in the image is obtained in advance from the magnification and reading density of the optical system of the color image input device 4, and the first stamp is processed in step S34. A region of that size is cut out from the starting point 1203 of 1202.

This processing will be described in detail with reference to the binary image of the postal matter shown in FIG. 11. First, the search is performed at the pixel position (4, 3).
Start with. In the diagonal search, the pixel position (5,
3), (4,4), (6,3), (5,4), (4
5), (7, 3), ... will be searched in this order.

By this search, the pixel position (5, 7) is first found as the pixel of "1". This position is the starting point of the first stamp. In this image, the size of the stamp is 3 horizontal
It is known in advance that the pixel is 4 pixels long,
A region of this size is cut out from the pixel position (5, 7) and is used as the first stamp.

Further, the first stamp detecting section 1 in step S4
As the first stamp stamping process according to No. 1, for example, as shown in FIG.
A process like the flowchart shown in FIG. Diagonal direction search in step S41, '1' in step S42
The process up to the pixel detection and the process of setting the start point of the first stamp in step S43 is the same as the process of steps S31 to S33 in FIG.

When the starting point of the first stamp is obtained in step S43, the process proceeds to step S44, and a position separated from the starting point by a predetermined distance is set as the starting position of the end point search of the first stamp. This predetermined distance is an amount that is obtained from the magnification and reading density of the optical system of the color image input device 4 and has a margin for the size of the stamp in the image. FIG. 15 (a)
Shows an image of the start position of the end point search. Stamp 1
When the starting point 1502 of 501 is calculated, the starting point 150
Point 1 separated from 2 by a predetermined distance 1503 in the horizontal and vertical directions
504 is set as the end point search start position of the first stamp 1501.

Next, in step S45, a search is performed from the obtained end point search start position 1504 in the diagonally opposite direction, as in the case of the start point search. This search is performed in the direction opposite to the direction at the time of searching the start point, as indicated by the arrow in FIG.

If a "1" pixel is detected in step S46, the process proceeds to step S47, as shown in FIG.
As shown in (b), the pixel position is set to the first stamp 15
The end point of 01 is 1505. Next, in step S48, the start point 1502 and end point 15 of the first stamp 1501
Cut out area 05.

This processing will be specifically described with reference to the binarized image of postal matter in FIG. Since the starting point of the first stamp is the same as the method obtained in the description of the processing of FIG. 12, a specific example of detecting the end points of the subsequent stamps will be shown here. The starting point position of the first stamp is (5, 7), and a predetermined distance from here,
In this example, a point (9, 13) that is 4 pixels in the horizontal direction and 6 pixels in the vertical direction is set as the end point search start position of the first stamp. From this point (9, 13), the search is performed in the diagonally opposite direction.

The search is performed at pixel positions (8, 13), (9, 1).
2), (7,13), (8,12), ...
By this search, the pixel position (7, 10) is first detected as the pixel of “1”. Therefore, set this pixel position to 1
The end of the first stamp. Then, the starting point (5,
The areas of 7) and the end point (7, 10) are cut out and used as the first stamp.

The first stamp detecting section 1 in step S4
As the first stamp stamping process according to No. 1, for example, as shown in FIG.
A process like the flowchart shown in FIG. First, in step S51, as shown in FIG. 17A, projection data of a stamp search area 1702 including a stamp 1701 is obtained from the binarized image of the postal matter. Then, the vertical projection data 1703 and the horizontal projection data 1704 become as shown in the figure.

Alternatively, as shown in FIG. 17B, it is conceivable that the vertical projection area 1705 and the horizontal projection area 1706 are separately defined, and the vertical projection data 1707 and the horizontal projection data 1708 are respectively obtained. . It is also conceivable to define a projected area as shown in FIGS. 17A and 17B from the starting point of the first stamp obtained from FIG. In any case, this specific area projection processing provides vertical projection data and horizontal projection data.

Next, the first stamp is cut out from the both projection data. First, in step S52, the vertical and horizontal projection data are binarized. Next, in step S53, as shown in FIG. 18, the area 18 where the vertical and horizontal projection data 1801 and 1802 intersect
Cut out 03 and use it as the first stamp.

This processing will be specifically described with reference to the binarized image of postal matter in FIG. For example, when the stamp search area is extracted from FIG. 11, it becomes as shown in FIG. When the projection data of this image is obtained, the vertical projection data 1 as shown
901 and horizontal projection data 1902 are obtained.

Next, this projection data is binarized at level "2", for example. In this example, the binarized image is also the same as the projected image in FIG. Then, the area (5, 5) where the respective binarized projection data intersect
7) to (7, 10) are cut out as a stamp area 1903.

Further, as the first stamp cutting process by the first stamp detecting unit 11 in step S4, for example, a process shown in the flowchart of FIG. 20 can be considered.
First, in step S61, labeling of a specific area is performed. That is, the binarized image of the postal matter is labeled, for example, in the stamp search area. For example, for 8-neighbor labeling, one pixel in the image is searched and that pixel is labeled. If there is a 1 pixel in 8 neighborhoods of that pixel, that pixel is also labeled with the same label. Further, if there is a pixel of 1 in 8 neighborhoods of the pixel, the same label is added to the pixel, and the same label is attached to one area.

Next, in step S62, as shown in FIG. 21, the area 210 labeled as described above is used.
The area is cut out with a rectangle 2102 including 1 and used as the first stamp.

This processing will be described in detail with reference to the binarized image of postal matter in FIG. As the specific area, here, the stamp search area is an area from pixel positions (4, 3) to (9, 13). First, labeling in this area is performed. Then, the same labels are given to the regions from pixel positions (5, 7) to (7, 10). Once labeled, the area with the same label is then cut out as a rectangle. To cut out as a rectangle, in the same labeled area,
It is only necessary to find the point having the maximum and the minimum pixel position as viewed in the vertical and horizontal directions, and to obtain the four sides of the rectangle from each point. When the labeled area of FIG. 11 is cut out into a rectangle according to FIG. 21, pixel positions (5, 7) to (7,
Area 10) is cut out.

Next, the inclination detection process of the stamp in step S5 will be described. For example, suppose that the starting point 2201 and the ending point 2202 of a stamp are obtained as shown in FIG. Since the size of the stamp is known from the magnification and density of the optical system of the color image input device 4, if the stamp is stuck straight, the starting point 2201 is the pixel position (1,
In the case of 1), the end point 2202 should be the pixel position (3, 4) in this example.

However, the end point 2202 is actually the pixel position (4, 3). Therefore, from this difference, it can be seen that the stamp is tilted. It can be seen that the end of the inclined direction is large in the horizontal direction and small in the vertical direction, so that the end is inclined to the left.

The inclination detection process of the stamp is performed according to the flowchart shown in FIG. 23, for example. First, in step S71, the coordinates of the starting point are subtracted from the coordinates of the ending point of the stamp. Next, in step S72, the subtraction result is compared with the reference data, that is, the size of the stamp of the image obtained from the magnification and reading density of the optical system of the color image input device 4.

As a result of this comparison, if the subtraction result matches the reference data (S73), it is determined that the stamp is attached straight without inclination (S74). If the subtraction result is larger in the horizontal direction and smaller in the vertical direction than the reference data (S75), it is determined that the reference data is tilted to the left (S7).
6). Further, if the subtraction result is smaller in the horizontal direction and larger in the vertical direction with respect to the reference data (S77), it is determined that the reference data is tilted to the right (S76).

Further, as shown in FIG. 22B, the inclination of the stamp can be detected also by the inclination of a straight line 2203 connecting the starting point 2201 and the ending point 2202 of the stamp. From the size of the stamp, the amount of inclination of the straight line connecting the start point and the end point when the stamp is attached in a straight line is known. Therefore, the amount of inclination and the actually detected stamp start point 2201 and end point 2
By comparing the amount of inclination of a straight line 2203 connecting 202 and 202, it is detected whether or not the stamp is inclined and in which direction.

The process of detecting the inclination of the stamp is performed according to the flowchart shown in FIG. First,
In step S81, the amount of inclination of the straight line connecting the start point and the end point of the stamp is calculated. Next, in step S82, the calculated tilt amount is compared with reference data (reference tilt amount set in advance).

As a result of this comparison, if the calculated inclination amount matches the reference data (S83), it is determined that the stamp is attached straight without inclination (S84). Also,
Since the tilt amount of the reference data is a negative value, if the calculated tilt amount is larger than the reference data (S85), it is determined that the tilt is left (S86). Further, if the calculated tilt amount is smaller than the reference data (S87), it is determined that the tilt is right (S88).

More specifically, FIG. 25A shows an example of the inclination of a straight line connecting the starting point and the ending point of a stamp having no inclination. For example, the inclination amount −1.33 is used as the reference data. Then, when the inclination amount of the straight line connecting the start point and the end point of the stamp of FIG. 25 (b) is calculated, it becomes −0.75 from the figure.
This value is larger than the reference data (-1.33). Therefore, it can be seen that the stamp of FIG. 25 (b) is tilted to the left.

By the way, since the stamps are generally stuck straight, the fact that the stamps are tilted means that the postal items are tilted rather than the stamps being stuck to the postal items. It is thought that Therefore, the inclination of the postal matter itself can be detected by detecting the inclination of the stamp. As the inclination detection process for mail,
It is realized by the same processing as the inclination detection of a stamp.

Next, the inclination correction process of the stamp in step S6 will be described. The inclination correction process of the stamp is performed according to the flowchart shown in FIG. 26, for example. First, in step S91, for example,
By the process based on FIG. 24, the tilt direction and the tilt amount of the stamp are detected. Next, in step S92,
The correction is performed so as to approach the standard data of the stamp around the starting point of the stamp, for example, in FIG. 25, the image is rotated clockwise around the starting point of the stamp.

More specifically, there is a difference of about 0.58 between the non-tilted stamp shown in FIG. 25A and the tilted stamp shown in FIG. 25B. Therefore, FIG.
For each pixel position of the stamp shown in (b), the calculation as shown in Expression 1 is performed.

[0057]

[Equation 1]

Here, a value in which the difference in angle of inclination between the reference data and the actual data is expressed in radians is substituted for θ.
In the example of FIG. 25, θ = −0.58 × π / 4 = −0.45
5, and substituting this numerical value into the mathematical expression 1, for example, as shown in FIG.
When the coordinate position (3,3) of the inclined stamp of 7 is corrected to (2,3), the coordinate position (5,4) is corrected to (4,4).
5), respectively.

Next, the process of cutting out a plurality of stamps in step S7 will be described. The processing for cutting out a plurality of stamps is performed according to the flowchart shown in FIG. 28, for example. First, in step S101, N = 2 is set, and a position separated by a predetermined distance from the starting point of the first stamp is set as the starting point of the second stamp search. The predetermined distance referred to here is an amount obtained from the magnification and reading density of the optical system of the color image input device 4 with some margin in the size of the stamp.

Next, in step S102, a diagonal search is performed. That is, the binarized image of the postal matter is searched diagonally from the starting point of the second stamp search, as in FIG. Next, in step S103, "1"
If the pixel is found, the process proceeds to step S104, and the pixel position is set as the starting point of the second stamp. Then, in step S105, a stamp-sized region in the image, which is obtained from the magnification and reading density of the optical system of the color image input device 4, is cut out from the starting point.

After that, stamps are sequentially cut out in the order of N = 3, 4, 5, ... And when the search has been completed for all the areas previously given as the stamp search range (S106), the process is terminated.

This processing will be specifically described with reference to the binarized image of the postal matter shown in FIG. 11. First, the starting point of the first stamp is found to be the pixel position (5, 7), and it is fixed from there. In this example, the distance is 0 pixels horizontally and 7 pixels vertically (5, 1
4) Let the position be the starting point of the second stamp search. And
Pixel positions (6,14), (5,15) in the diagonal direction,
(7, 14), (6, 15), (5, 16), ... Are sequentially searched, and the pixel of '1' is first found at the position of (6, 16).

Therefore, using this position as the starting point of the second stamp, a fixed area, an area of 3 pixels in the horizontal direction, and an area of 4 pixels in the vertical direction are cut out.
In this example, the postage stamp search range is from pixel positions (4, 3) to (9, 20), and the process ends because there is no third postage stamp.

Further, as the processing of cutting out a plurality of stamps in step S7, for example, a processing as shown in the flowchart of FIG. 29 can be considered. This processing is an application of the stamp cutting processing of the first sheet in FIG. 14 to the detection of a plurality of sheets, and as shown in FIG. 30, the second sheet separated by a predetermined distance from the end point 2903 of the first stamp 2901. The search starting point 2904 of the stamp 2902 is obtained, and the search is performed in the same manner as in FIG.

That is, first, in step S111, N = 2 is set, and a position separated from the starting point of the first stamp by a predetermined distance is set as the starting point of the second stamp search. Next, in step S112, a diagonal search is performed. That is, the binarized image of the postal matter is searched diagonally from the starting point of the second stamp retrieval, as in FIG.

Next, in step S113, "1" is set.
If the pixel is detected, the process proceeds to step S114, and the pixel position is set as the starting point of the second stamp. Then, in step S115, the position distant from the starting point by a predetermined distance is set as the starting position of the end point search of the second stamp. next,
In step S116, the search is performed in the diagonally opposite direction from the previously obtained end point search start position, as in the start point search. This search is performed in the opposite direction to the direction used when searching for the starting point.

Now, in step S117, "1" is set.
If the pixel is detected, the process proceeds to step S118, and the pixel position is set as the end point of the second stamp. Next, in step S119, the areas of the start point and the end point of the second stamp are cut out. After that, stamps are sequentially cut out in the order of N = 3, 4, 5, ... And when the search has been completed for all the areas previously given as the stamp search range (S120), the processing is terminated.

This processing will be specifically described with reference to the binarized image of the postal matter in FIG. 11. The end point of the first sheet is the pixel position (7, 1).
0), the second stamp is displayed at this pixel position (7, 1
The pixel at the pixel position (4, 13), which is advanced by a fixed amount from 0) to -3 pixels in the horizontal direction and 3 pixels in the vertical direction in this example, is searched.

When the search is advanced in the same manner as in FIG. 14, the starting point of the second stamp is found at the pixel position (6, 16), and a fixed amount is found from the pixel position (6, 16), in the horizontal direction in this example. Three
The point advanced by 4 pixels in the vertical direction is used as the starting point for searching the ending point of the second sheet, and the ending point of the second sheet is searched. As the search proceeds, the pixel position (8, 18) is found as the end point of the second stamp.

Therefore, from the pixel position (6, 16) to (8,
The area of 18) is cut out. In this example, the stamp search range is from pixel positions (4,3) to (9,20), and 3
The process ends because there is no stamp of the first sheet.

Further, as the process of cutting out a plurality of stamps in step S7, for example, a process as shown in the flowchart of FIG. 31 can be considered. The flow of this processing is as follows: Steps S121 to S122: Steps S51 to S of FIG.
Similar to 52, except that a plurality of areas are cut out from the projection data after that.

This processing will be specifically described with reference to the binarized image of postal matter in FIG. For example, when a plurality of stamps search area is taken out from FIG. 11, it becomes as shown in FIG. When the projection data of this image is obtained, it becomes vertical projection data 3101 and horizontal projection 3102 as shown in FIG. Next, this projection data is binarized, for example, at level "2". In this example, the vertical projection binary image is 310
It becomes like 5.

Then, by the horizontal projection area cutout processing in step S123, the image image of FIG. 32 is divided into two areas as shown in FIGS. 33 (a) and 33 (b) based on the horizontal projection data. Next, in the vertical projection area cutout process of step S124, vertical projection data 3201, 3202 is obtained for each of these areas, the projection data 3201, 3202 is binarized, and the intersecting areas (5, 7) From (7, 10) to (6, 16) to (8, 18) regions 3103 and 3104 are cut out as stamp regions.

In this example, since the division by the vertical projection data cannot be performed first, the division by the horizontal projection data is performed first, but the division by the vertical projection data may be performed first.

Further, as the process of cutting out a plurality of stamps in step S7, for example, a process as shown in the flowchart of FIG. 34 can be considered. The flow of this process is the same in step S131 as in step S61 of FIG.
The difference is that a plurality of rectangular areas are cut out from the labeling area in step S132.

This processing will be specifically described with reference to the binarized image of postal matter in FIG. As a specific area, here, the plural stamps search area is defined as an area from pixel positions (4, 3) to (9, 20). First, 8-neighbor labeling in this region is performed. In FIG. 11, pixel positions (4, 3) to (9,
When labeling is performed in the regions up to 20), the same labels are attached to the regions from pixel positions (5, 7) to (7, 10), and further, from pixel positions (6, 16) to (8, 1).
The same label is attached to the areas up to 8).

After the labeling, the area with the same label is cut out as a rectangle. When the labeled area in FIG. 11 is cut out into a rectangle according to FIG. 21, the area from pixel positions (5, 7) to (7, 10) and the area from (6, 16) to (8, 18) are extracted. It is cut out.

When a plurality of stamps are pasted, the number of stamps may be specified by the keyboard of the data input device 8 as well as the range of stamp search is determined. Further, as shown in FIG. 35, it may be possible to detect a plurality of stamps after detecting in advance whether or not a plurality of stamps are present on the right side or in the downward direction. FIG. 36 shows a flowchart of the processing procedure.

That is, when the first stamp 3401 is detected, first, in step S141, the starting point 3
Find the center point 3404 between the 402 and the end point 3403. Next, in step S142, the midpoint 340
A point 3405 or 3407, which is a predetermined distance from 4, is obtained. Next, in step S143, a certain range is defined around that point, and the number of pixels of "1" of the binarized image of the postal matter is obtained within the certain range.

Then, in step S144, if the number of pixels previously obtained is equal to or larger than a predetermined threshold value, it is determined that a stamp exists in that direction (S145), and the stamp search in that direction can be advanced. In step S144, if the number of pixels obtained previously is not greater than or equal to the predetermined threshold value, it is determined that no stamp exists in that direction (S14).
6), the process ends.

This processing will be described in detail with reference to the example of the binarized image of postal matter shown in FIG. For example, it is assumed that the area from pixel positions (5, 7) to (7, 10) is detected as the first stamp. First, by the process of step S141, 1
Find the midpoint (6, 8) of the first stamp. Next, a point that is a fixed amount, which is 9 pixels in the vertical direction in this example, is obtained from the obtained middle point (6, 8), and the point around that point, in this example, 3 pixels in each of the vertical and horizontal directions. 'Pixel is calculated. There are 6 pixels of "1" in this range, and the threshold value is 4
If it is a pixel, the result is that a stamp exists in that direction, that is, in the vertical direction.

Here, the determination of whether there is a stamp on the right side of the first stamp or in the downward direction has been described, but this can be applied to any stamp of the second stamp or more. .

Finally, in step S8, the position information of the stamp area cut out as described above is output, and all stamp detection processing is completed. Based on the output position information of the postage stamp area, for example, pattern matching processing of the postage stamp image can be performed by the composite similarity method to determine the type of postage stamp. Further, the position for stamping a stamp can be obtained from the position information of the stamp area.

As described above, according to the above embodiment,
Since the two-dimensional image is used for cutting out the stamp on the mail, the two-dimensional area information of the stamp can be obtained, and the range of the stamp becomes clear.

Further, since the position of the stamp is obtained by detecting the color change point of the color image on the postal matter, not only a stamp with a specified pattern such as a normal stamp but also a memorial stamp or a special stamp. Stamps of various color patterns can be cut out with high precision.

Further, by detecting not only the start point but also the end point of the stamp, it is not affected by the magnification change of the optical system of the color image input apparatus, and commemorative stamps and special stamps having different stamp sizes are high. It can be cut out with accuracy.

Further, by detecting not only the starting point but also the ending point of the stamp, the inclination of the stamp and further the inclination of the postal matter can be detected, and the inclination can be corrected. Also, by cutting out a plurality of stamps, the range in which the stamps are present becomes clear.

In the above embodiment, the search for detecting the starting point of the stamp is performed in the diagonal direction.
The direction of this search can be reverse. Also,
The angle in the oblique direction is not limited to 45 ° and may be any angle. The search for detecting the end point of a stamp is similar.
Further, the starting point of the search is the upper left near the position where the stamp is attached in the mail, but the search may be started from any position.

[0089]

As described above in detail, according to the present invention, not only stamps whose patterns are usually specified, but also stamps having various color patterns such as commemorative stamps and special stamps, Alternatively, it is possible to provide a stamp detection device that can detect stamps of different sizes with high accuracy.

[Brief description of drawings]

FIG. 1 is a block diagram showing a configuration of a stamp detecting device according to an embodiment of the present invention.

FIG. 2 is a flowchart illustrating a stamp detection process.

FIG. 3 is a diagram showing an example of color image data obtained from a postal matter.

FIG. 4 is a diagram showing an example of postal image cut-out image data.

FIG. 5 is a flowchart illustrating color change point detection processing.

FIG. 6 is a diagram showing a color change point detection image.

FIG. 7 is a diagram showing an example of a first-order differential matrix.

FIG. 8 is a flowchart showing a procedure of first-order differentiation.

FIG. 9 is a diagram showing a first-order differential image.

FIG. 10 is a diagram illustrating the calculation of the first derivative.

FIG. 11 is a diagram showing an example of a binarized image of postal matter.

FIG. 12 is a flowchart illustrating a stamp cutting process for the first sheet.

FIG. 13 is a diagram illustrating detection of the starting point of the first stamp.

FIG. 14 is a flowchart illustrating another example of the stamp cutting processing for the first sheet.

FIG. 15 is a diagram illustrating detection of the end point of the first stamp.

FIG. 16 is a flowchart illustrating still another example of the stamp cutting processing of the first sheet.

FIG. 17 is a diagram illustrating a method of generating projection data.

FIG. 18 is a diagram for explaining stamp cutting processing using projection data.

FIG. 19 is a diagram for explaining an example of stamp cutting using projection data.

FIG. 20 is a flowchart illustrating still another example of the stamp cutting process for the first sheet.

FIG. 21 is a diagram illustrating a rectangular cutout process using a labeling area.

FIG. 22 is a diagram for explaining tilt detection of a stamp.

FIG. 23 is a flowchart illustrating a tilt detection process of a stamp.

FIG. 24 is a flowchart illustrating another example of the inclination detection process of a stamp.

FIG. 25 is a diagram illustrating another example of detecting the inclination of a stamp.

FIG. 26 is a flowchart illustrating a tilt correction process of a stamp.

FIG. 27 is a diagram for explaining tilt correction of a stamp.

FIG. 28 is a flowchart illustrating a process of cutting out a plurality of stamps.

FIG. 29 is a flowchart for explaining another example of the processing for cutting out a plurality of stamps.

FIG. 30 is a diagram for explaining detection of a plurality of stamps.

FIG. 31 is a flowchart for explaining still another example of the multiple-sheet stamp cutting process.

FIG. 32 is a diagram illustrating an example of cutting out a plurality of stamps based on projection data.

FIG. 33 is a diagram illustrating an example of region division based on horizontal projection data.

FIG. 34 is a flowchart for explaining still another example of the multiple-sheet stamp cutting process.

FIG. 35 is a diagram for explaining detection of a plurality of stamps.

FIG. 36 is a flowchart for explaining a multiple stamp detection process.

[Explanation of symbols]

P: Mail, 1 ... CPU, 2 ... Main memory, 3 ...
Storage device, 4 ... Color image input device, 5 ... Image output device, 6 ... Input interface unit, 7 ... Output interface unit, 8 ... Data input device.

─────────────────────────────────────────────────── ─── Continuation of the front page (51) Int.Cl. ⁶ Identification code Internal reference number FI Technical display location G06T 1/00 G06K 9/00 H 8623-5L 8837-5L G06F 15/70 330 A (72) Invention Toshiyasu Hirasawa 70 Yanagicho, Saiwai-ku, Kawasaki-shi, Kanagawa Prefecture, Toshiba Yanagimachi Plant (72) Inventor Teruhiko Uno 70, Yanagicho, Sachi-ku, Kawasaki, Kanagawa Prefecture

Claims (6)

[Claims] 1. A color image input means for inputting a color image on a postal matter having a stamp, and scanning from a specific position of the color image input by the color image input means with reference to an oblique direction, Scanning means for detecting the color change point on the mail, and based on the color change point detected by the scanning means,
A stamp detecting device comprising: a region specifying unit that specifies a region where a stamp on the mail input by the color image input unit exists. 2. The scanning means increases / decreases the X coordinate and decreases the Y coordinate with reference to the coordinates (X1, Y1) of the specific position.
2. The stamp detecting device according to claim 1, further comprising a unit for performing scanning based on an oblique direction from the specific position by sequentially increasing / decreasing the coordinates. 3. The area specifying means generates projection data for projecting color change points detected by the scanning means, and projection data generation means for generating the projection data on the postal matter from the projection data generated by the projection data generation means. The stamp detecting device according to claim 1, further comprising: a cutting-out unit that cuts out an area where the stamp exists. 4. The area specifying means labels the color change points detected by the scanning means, and cuts out the area labeled by the labeling means as a rectangle, thereby stamping on a postal matter. The stamp detecting device according to claim 1, further comprising a cutting unit that cuts out an existing region. 5. A color image input means for inputting a color image on a postal matter having stamps, and scanning from a specific position of the color image input by the color image input means with reference to an oblique direction, The first scanning means for detecting a color change point on the mail and the first diagonal direction from a position separated from the first color change point detected by the first scanning means by a predetermined distance. A second scan for detecting a second color change point on the mail from the color image of the mail input by the color image input means by scanning with the second diagonal direction being the opposite direction as a reference. Means, and the first color change point detected by the first scanning means and the second color change point detected by the second scanning means, which are input by the color image input means. Identify areas of stamps on postal items A stamp detecting device comprising: an area specifying unit; 6. A color image input means for inputting a color image on a postal matter having stamps, and scanning from a specific position of the color image input by the color image input means with reference to an oblique direction, Scanning means for detecting the color change point on the mail, and based on the color change point detected by the scanning means,
From the area specifying means for specifying the area of the postage stamp on the mail input by the color image input means and the position information of the area of the postage stamp specified by the area specifying means, the inclination information of the area is obtained. A tilt detecting unit for detecting the tilt; and a tilt correcting unit for correcting the position information of the region where the stamp specified by the region specifying unit is present, based on the tilt information detected by the tilt detecting unit. Characteristic stamp detection device.