JPH0368431B2 - - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION The present invention relates to a character separation device that separates a character string image written on paper into individual characters.

When recognizing a series of characters in a device that optically reads various groups of printed characters (hereinafter referred to as OCR), it is necessary to separate each character one by one and send it to a character recognition unit. By the way, when reading mail or a large amount of documents using OCR, there are many different qualities and printing styles of printed matter, and it is necessary to handle these as objects to be read. In such a case, characters may come into contact with each other in the character string image on the paper, or one character may be separated into two or more characters, and a character separation method that can efficiently handle these situations is required. Conventionally, when human data with relaxed constraints on the reading target is included, character separation methods have been applied that add functions that are considered effective in each case. However, adding functions corresponding to individual cases as appropriate may reduce the accuracy of character separation, or it may become necessary to develop character separation devices with different functions for each individual object. .

SUMMARY OF THE INVENTION In order to solve the above-mentioned conventional problems, it is an object of the present invention to sequentially set separation candidate sections using character pitch and blank information, and to determine the distance between each separation candidate position between each separation candidate section. By using dynamic programming to calculate a series of character separation candidate positions that minimize the variance of the variance and the variance of the deviation between the average distance and character pitch, an image of contact between characters was created. To provide a character separation device capable of easily and accurately determining a character separation candidate position without adding any special function even when a character is separated into two or more characters.

Hereinafter, the present invention will be explained in detail using the drawings.

FIGS. 1a, b, and c are diagrams for explaining an example of a separation candidate section setting method for character separation in the present invention using an example of a partial character string image.
Note that when performing character separation in a reading target where there are various print qualities and printing styles, it is necessary to first detect the character pitch. As a character pitch detection device, for example, Japanese Patent Application No. 160763/1983 filed by the same applicant, “Character Pitch Detection Device”
There is a character pitch detection device as shown in (hereinafter referred to as Cited Example 1), and by using such a device, character pitch can be detected from a series of character line images.

In FIG. 1a, a character string image including a contact image between characters and a case where one character is split into two characters is shown by diagonal lines, and P in the figure shows a character pitch. When the partial character string image in FIG. 1a is vertically projected, the projection distribution shown in FIG. 1b is obtained, and it can be separated into a black background area (shaded area in the figure) and a white background area. The character pitch P may be determined from the character string image shown in FIG. A value may also be used. In order to correctly separate the touching characters (amu) and the split character h shown in FIG. 1A into individual characters, it is necessary to accurately predict the separation start position of the characters. For example, in the figure, between the touching character images "a" and "m", since a is slightly smaller, its extraction start position is slightly to the left of the starting edge of the touching character image (amu). Conventionally, correction of such positional deviation of touching characters has been performed by, for example, referring to character images. Therefore, depending on the degree of contact between the character images, it may not be possible to correctly determine the separation position, even though it takes processing time to refer to the images. Therefore, in the method of determining character separation positions in the present invention, first, in a character line image as shown in FIG.
The positions and sizes of white background areas (hereinafter referred to as blanks) and black background areas (hereinafter referred to as character blocks) extracted from the projection distribution etc. as shown in FIG. 2 are extracted. For example, the character block width Vi as shown in Figure 1b
(i=1,...4) and blank size Wi(i=1,...
4) Furthermore, their location information is obtained using known techniques. Next, using the previously obtained character pitch P and threshold values T ₁ and T ₂ , an interval (hereinafter referred to as an allowable interval) in which a character separation position can be set is determined, for example, under the following conditions. It can be set using (1) and (2).

Condition (1) Blank sections are allowed sections.

Condition (2) For any character block width Vi, Vi>P+
Among the black areas that satisfy T ₁ , the section excluding the black area from both ends of the character block width Vi to T ₂ is an allowable section.

For example, the permissible interval that satisfies the above-mentioned conditions (1) and (2) is as follows for the character string image in Figure 1a:
These are obtained as sections A ₁ , A ₂ , A ₃ , A ₄ , and A ₅ in the diagram of FIG. 1c. In addition, the threshold values T ₁ and T ₂ mentioned above are
It may be given as a function of character pitch P. Further, the threshold value T ₁ may be set based on the estimation error when estimating the character pitch P shown in Cited Example 1 above. Next, character separation positions in the present invention are determined within the above-mentioned allowable interval using the character pitch P and the threshold value T ₃ in order of separation candidate interval k (however, k≧0 ₎ , and _each separation is The distance d (k, k+1; i _k , i _k+1 ) between candidate positions is calculated, and the distance d (k, k
+1; i _k , i _k+1 ) and ^the squared error (μ _{d - P) of the deviation between the average value μ d} ^and _{the character} pitch P. This is done by finding a sequence of separation candidate positions x(k, i _k ) (k≧0) that minimizes Therefore, regarding an example of the method of setting the separation candidate section k described above, the first
This will be explained using Figure c.

The black dots in FIG _. The positions are sequentially set as positions that satisfy the relationship of equation (3) shown below.

|d(k, k+1: i _k , i _k+1 )−P |≦T ₃ ...(1) In equation (1), the distance d(k, k+1; i _k , i _k+1 )
is the separation candidate position x(k+
1, i _k+1 ) and separation candidate position x of separation candidate section k
Distance x(k+ ₁ ,i _k+1 )−x(k+1,
i _k ). For example, the separation candidate position x (1, 1) and the separation candidate position x (2,
2), at the distance d(1,2;1,2), |P
-d(1,2;1,2)|≦P+T ₃ is satisfied. Also, separation candidate position x(0,
From 1) to the position that satisfies the relational expression shown in equation (1),
There are six positions as shown by the white dots in Figure 1c, but there are two white dots that belong to the above-mentioned tolerance interval as shown by the black dots in Figure 1c. Separation candidate positions are x (1, 1), x (1, 2)
It can be found as two positions. In addition, the threshold value mentioned above
T ₃ can be given as a function of the character pitch P, like the threshold T ₁ described above. Next, when determining the optimal separation candidate position using the evaluation scale U, the separation candidate section that becomes the terminal state may be set only within the blank space that is the end of the character line, or the blank size described above may be set. It is also possible that W _i can be set within a permissible interval that satisfies Wi>T ₄ ·P (where T ₄ is a threshold) with respect to character pitch P. If the latter method is adopted, for example, the first
Region E in FIG. c is detected as a permissible section in which the terminal state occurs. Furthermore, in calculating the evaluation scale U, the separation candidate section (k+1) to be set next is
is set using the above-mentioned formula (1) from the separation candidate interval k, the separation candidate interval (k+1) that satisfies formula (1)
If all of the separation candidate sections k are not permissible sections, the separation candidate section k can be set as the separation candidate section k that becomes the terminal state. On the other hand, the separation candidate section that becomes the starting end state for determining the optimal separation candidate position using the evaluation scale U can be set based on the position of the separation candidate section that becomes the end state that has already been detected.

It goes without saying that the separation candidate section setting method for character separation of the present invention can be set based on the character pitch P, the character block width Vi, and the blank size Wi in addition to the conditions described above.

FIG. 2 is a diagram for explaining the principle for extracting optimal character separation positions in the present invention.
In the figure, the positions indicated by black dots indicate the values of each separation candidate position x(k, i _k ) in the separation candidate section k (k=0, . . . 4) shown in FIG. 1c. Further, the case where the character pitch P is 20 is shown. Furthermore,
In order to simply explain the principle of the present invention, character string images to be separated are from separation candidate section 0 to separation candidate section 4 in FIG. 1c. First, let's define the symbols. Symbol μ _d (r, n; i _r , i _o )
(However, 1≦r<n) represents each separation candidate from separation candidate position x (r, i _r ) of separation candidate section r to separation candidate position x (n, i _o ) of separation candidate section n. n−r+1 separation candidate positions x(r, i _r ), x(r+1, i _r+1 ), ..., x(n, i _o
)
n-r distances d(r, r+1; i _r ,
i _r+1 ) d(r+1, r+2; i _r+1 , i _r+2 )…, d(n
-1, n; _io-1 , _io ). symbol σ ² _d
(r, n; i _r , i _o ) (1≦r≦n) is calculated from the separation candidate position x (r, i _r ) in the separation candidate section r to the separation candidate position x (n, n−r+ arbitrarily selected in each separation candidate interval up to i _o )
One separation candidate position x(r, i _r ), x(r+1,
n-r distances _d (r, r+1; i _{r ,} i _r+1 ), d(r+1, r+
2; i _r+1 , i _r+2 ), the above-mentioned average value μ _d (r, n; i _r ,
i _o ). Therefore, the character separation position from the separation candidate section r (r = 0 in Fig. 2) which is the start state to the separation candidate section n (n = 4 in Fig. 2) which is the end state is , separation candidate positions x(r, i _r ), x(r+1, i _r+1 ), ..., which minimize the evaluation scale U shown in equation (2).
It can be obtained by finding x(n, _io ).

U(r,n)=β・σ ² _d (r, n; i _r , i _o ) +(1+β)・(μ _d (r, n; i _r , i _o )−P) ² ... (2) The weighting coefficient β in equation (2) satisfies 0≦β≦1. Here, a more specific method for realizing separation candidate positions that minimizes Equation (2) can be performed without consuming memory capacity using the dynamic programming method described below. Therefore, at any separation candidate position x (k+1, i _k+1 ) in separation candidate section k+1, the previous separation candidate section k (however, if k=0,
This is the separation candidate section that includes the start state. ) for any separation candidate position x(k, i _k ) (where i _k =1,2...h _k
The optimal separation candidate position x (k, i _k ) that satisfies equation (2) is determined from the separation candidate position x (k + 1, i _k ) from h _k ≥ 1) using the recurrence formula described below. can be found. First, the distance d(k, k+1; i _k ,
i _k+1 ) (where i _k =1, 2...h _k ), and calculate equations (3-1), (3-2), and (3-3) shown below.

μ _d (0, k+1; i ₀ , i _k+1 )=1/k+1 {k・μ ^* _d (0, k; i ₀ , i _k ) +d(k, k+1; i _k , i _k+1 ) } ...(3-1) D(k+1)=D ^* (k)+d ² (k, k+1; i _k , i _k+1 ) ...(3-2) U(0, k+1)=β・( D(k+1)/k+1−μ ² _d (0, k+1; i ₀ , i _k )) ² +(1−β)(μ _d (0, k+1; i ₀ , i _k )−P) ²
...(3-3) h _k separation candidate positions x(k,
1), ... x (k, h _k ), the separation candidate position x (k, i k ) that minimizes the evaluation scale U (0, k + 1) of equation (3-3) is the separation candidate position x (k, i _k ) of the separation candidate section k + 1. This becomes the optimal separation candidate position for an arbitrary separation candidate position x (k+1, i _k+1 ).

Here, each separation candidate position x (0, i ₀ ) in separation candidate section 0, which is the start state (i ₀ in FIG.
= 1), the optimal average value μ ^* _d (0, 0; i ₀ , i ₀ ) = 0 shown in equation (3-1), equation (3-2)
Assume that the cumulative sum of squares of the optimal distance d (-1, 0; i _-1 , i ₀ ) shown in is D ^* (0)=0. Each separation candidate position x (k, i _k ) of separation candidate section k has the above-mentioned optimal average value μ ^* _d (0, k; i _k , i ₀ ) and optimal distance d (k−
1, k; i _k-1 , i _k ) cumulative sum D ^* (k-1,
k), the next separation candidate section k+1
The optimal separation candidate position x (k, i _{k ) of separation candidate section k at each separation candidate position x (k+1} , i _k+1 )
is required. Note that the first term in equation (3-3) is the variance σ ² _d ((0, k+1; i ₀ , i _k+1
)
It has become a different way of expressing.

Next, using Figure 2, equation (3-1) and equation (3
-3) The calculation process will be explained. In the figure, the values shown in brackets are for each separation candidate section k (k
=0,1,2,3,4) for each separation candidate position x(k,
i _k ), the average value μ ^* _d (0, k; i ₀ , i _k ) and the evaluation scale U (0, k ), which is calculated as the optimal value from the separation candidate position x (k-1, i _k-1 ). In addition, in this explanation, formula (3-3)
The case where the weighting coefficient β in is set to 0.5 will be described. Further, each arrow in the figure indicates a series of optimal separation candidate positions. for example,
The separation candidate position x (2, 1) is the position 39, and the distance from the separation candidate position x (1, 1) is d (1, 1; 1,
1) becomes 19. Therefore, the separation candidate position x(1,
The average value μ _d (0,2) at the separation candidate position x(2,1) passing through 1) is 1/2·(1×20+19) from equation (3-1) and the figure, and has a value of 19.5. Next, for the separation candidate position x (1, 1), formula (3-2) is used.
D ^* (1) = 20 ² shown by is stored (in the figure,
(omitted), using equation (3-2), D(2)=
20 ² + 19 ² . Therefore, the separation candidate position x(1,
The evaluation scale U (0, 2) at the separation candidate position x (2, 1) passing through 1) is U (0, 2) = 0.5 (20 ² + 19 ² /
2 -19.5 ² ) + 0.5・(19.5 - 20) ² , giving a value of 0.38. Similarly, the evaluation scale U(0,
2) is also calculated (however, the calculation is omitted), and the value
It becomes 1.26. Therefore, separation candidate position x (2, 1)
If the minimum value is taken from the two evaluation scales U(0,2) for
Average value μ ^* _d (0, 2) = 19.5, evaluation scale U (0, 2) =
0.38 is selected. Below, similar operations are performed using the formula (3-
1), equation (3-2), and equation (3-3), each separation candidate position x(k, i _k ) (k=0,1,
2, 3, 4), the evaluation scale U(0, k) (where k=0, 1, 2, 3, 4) is calculated.

Next, as described above, separation candidate positions x(4,2), x(4,
3), among x(4,4), the evaluation scale U(0,4)
The separation candidate position x (4, 2) with the minimum is selected as the end position of character separation. Therefore, we set the series of optimal separation candidate positions to the character separation end position x(4,
2) By tracing backwards, x(4,2)=
81, x (3, 3) = 60, x (2, 2) = 20, x (1,
1)=20, x(0,1)=0.

FIG. 3 is a logic block diagram showing a specific embodiment of the present invention. A scanning unit 1 optically scans a character string image written on a paper surface, converts it into an electrical signal, and writes it into a character string image memory 2 after binary quantization. The character block extraction unit 3 sequentially extracts character blocks from the character string image stored in the character string image memory 2, and stores the position, width, and height of each character block in the character block information register 21. Note that such a character chunk extraction unit 3 can be obtained using a known technique. The character pitch detection unit 4 estimates the character pitch P using the position and character block width of each character block stored in the character block information register 21, and furthermore, the height of the character.
Store in 2. Incidentally, such a character pitch detection section 4 can be obtained using the technique shown in the specification of Cited Example 1 by the same applicant, and if the character pitch P is known in advance, The character pitch P may be used. The parameter information register 30 stores parameters T ₁ , T2, T ₃ , T ₄ , which are various threshold values and weighting coefficients used in the present invention.
Store T ₅ and β. The permissible interval extraction unit 5
A permissible interval that satisfies condition (1) and condition (2) described using the diagram is extracted. First, the allowable blank interval shown in condition (1) is based on the positions of multiple character blocks and the character block width Vi stored in the character block information register 21.
Using this, the blank position and blank size are extracted by a comparison circuit or the like. Next, the permissible interval within the black background area shown in condition (2) is determined by the width of each character block.
It is compared whether Vi is larger than the sum P+T ₁ of the character pitch P stored in the character pitch information register 22 and the parameter T ₁ stored in the parameter information register 30, and if it is, the width of each character block Vi is Excluding the values from both ends to the value indicated by the parameter information register _T2 , the section including the character block width Vi is extracted as a permissible section. As described above, the blank allowable interval and conditions that satisfy the extracted condition (1)
Tolerable sections within the black background area that satisfy (2) are extracted, and the position and width of each extracted allowable section are stored in the allowable section information register 23. The terminal candidate section extraction unit 6 extracts the parameters T 4 and 4 stored in the parameter information register 30 for the blank allowable section Wi among the allowable sections sequentially stored in the allowable section register ₂₃ corresponding to the character line image. By calculating the product T ₄・P with the character pitch P and comparing the product T ₄・P with the blank allowable interval Wi, detect the allowable interval Wi that is larger than the product T ₄・P. . Next, from the start of the allowable interval Wi, the product T ₅ · P of the parameter T ₅ and the character pitch P (however, T ₅ ≦
Calculate the allowable interval up to T ₄ ) and further calculate the allowable interval Wi
Calculate the permissible interval within the sum P + T ₁ of the character pitch P and the parameter T ₁ from the character block width Vi that exists immediately before, and use the logical sum of the two permissible intervals described above as the end candidate interval, and sequentially calculate It is stored in the end candidate section register 24.

FIGS. 4a and 4b show an example of a terminal candidate section extracted by the terminal candidate section extraction section 6. Fourth
In the case of figure a, the terminal section is determined as the section indicated by _T5 ·P in the figure. In the case of Fig. 4b, the terminal section is the section indicated by the last arrow in the figure,
Of the logical sum of T ₅ ·P and P + T ₁ , it is a blank allowable interval. Separation candidate section extraction unit 7
is the separation candidate position x(k, i _k ) of the separation candidate area k as explained using FIG.
Using the tolerance interval and parameters stored in
Extract sequentially. Note that each separation candidate position x (0, i ₀ ) (however, i ₀ = 1, 2...h ₀ ) of the starting edge separation candidate section 0 including the character separation start position is determined by the control unit 10 first. It is assumed that the separation candidate positions are determined from each separation candidate position within a certain range of blank allowable intervals set based on the character pitch P from the starting end of the column image, and are stored in the optimum separation position information register 26. Therefore, the separation candidate section extraction unit 7 has already
Separation candidate section k extracted and stored in the optimal separation candidate position register 26 (k = 0, 1, 2...)
Separation candidate position x (k, i _k ) (where i _k = 1, 2...
From h _k ), a separation candidate position x (k+1, i _k+1 ) belonging to the allowable interval that satisfies equation (1) is calculated. That is, the first separation candidate position x(k,
1), x( _k ,
1) Calculate _the position where +P-T ₃ , and then calculate x( _k ,
Calculate the position where h _k )+P+T ₃ . The above two positions x (k, 1) + P - T ₃ , x (k, h _k ) + P + T ₃
Of each separation candidate position in the section obtained by
By performing a logical product, the separation candidate positions belonging to the above-mentioned permissible interval are determined as each separation candidate position x (k+1, i _k+1 ) in the separation candidate interval k+1 (where i _k+1 =1, 2
... h _k+1 ) and stored in the separation candidate position information register 25. When the contents of the separation candidate position information register 25 are input to the evaluation scale calculation unit 8, the optimum separation position information register 26 contains the already calculated separation candidate positions x from separation candidate section 0 to separation candidate section k. (0, i ₀ ) (however, i ₀ = 1...
h ₀ ), x (1, i ₁ ) (where i ₁ = 1...h ₁ ) _, ...x(k,
i _k ) (where i _k =1...h _k ) are stored. Furthermore, corresponding to each separation candidate position x (m, i _n ) (where i _n =1... h _n ) of the separation candidate section m (m=0...k), the evaluation scale calculation unit 8 calculates the following: The average value μ ^* _d (0, m; i ₀ , i _n ) calculated from equation (3-1), equation (3
−2) Cumulative sum of squared distances D ^* (k),
Evaluation scale U (0, m) calculated from formula (3-3)
and the optimal separation candidate position x(m-1,in _-1 ) of the immediately preceding separation candidate section m-1 are stored. still,
When each separation candidate position x (0, i ₀ ) of separation candidate section 0 is stored by the control unit 10, the average value μ stored corresponding to each separation candidate position x (0, i ₀ ) ^* _d
(0, 0; i ₀ , i ₀ ) and the cumulative sum of the square of the distance D ^*
It is assumed that 0 is stored in (0). Therefore,
The evaluation scale calculation unit 8 receives separation candidate positions x(k+1,
i _{k )} , _the distance _d ( k, k+1; i _k , i _k+1 ), and further calculate its average value μ ^* _d (0, k; i ₀ , i _k ), and the cumulative sum of squares of the distances.
Using D ^* (k) and the parameter β stored in the parameter information register 30, formula (3-1)
The recurrence formula μ _d (0, k+1; i ₀ , i _k+1 )=
1/k+1{k・μ ^* _d (0, k; i ₀ , i _k )+d(k, k
+ 1; i _k i _k+1 ), recurrence formula D (k
+1) = D ^* (k) + d ² (k, k+1; i _k , i _k+1 ), equation (3
-3) recurrence formula U(0,k+1)=β・
(D(k+1)/k+1−μ ² _d (0, k+1; i ₀ , i _k+1
)) ² + (1 − β)・(μ _d (0, k+1); i ₀ , i _k+1 ) − P) ² , the separation candidate position x of the previous separation candidate section k is calculated. An evaluation scale U(0, k+1) for (k, i _k ) (where i _k =1...h _k ) is calculated.

Next, among the h _k separation candidate positions x(k, i _k ),
The separation candidate position x (k, i _k ) where the evaluation scale U (0, k + 1) is the minimum is determined as the separation candidate position x (k + 1, i _{k +1} )
The optimal separation candidate position x(k,
i _k ), and further, the evaluation scale U(0, k+1)
The average value μ ^* _d (0, k + 1; i ₀ , i _k ) at which ^the minimum value of It is stored in the optimum separation position information register 26 together with the position x(k+1, i _k ). When the evaluation scale calculation unit 8 performs the above-mentioned calculation processing on all separation candidate positions x (k+1, i _k+1 ) sequentially transferred from the separation candidate position information register 25, the control unit 10 The separation candidate section extraction unit 6 inputs each separation candidate position x (k+2, i _k+2 ) of the next separation candidate section k+2.
, and the same operations described above are repeated. Here, the control unit 10 uses the evaluation scale calculation unit 8 to control the optimum separation position information register 2.
The separation candidate position x (k+1, i _k+1 ) of the separation candidate section k+1 transferred to the end section information register 2
It is checked whether the end candidate section stored in section 4 has been reached, and if it has not been reached, only the above-mentioned request is output to the separation candidate section extraction section 6. On the other hand, when the separation candidate position x (k+1, i _k+1 ) arrives at the terminal candidate section, the control unit 10 outputs the above-mentioned request to the separation candidate section extraction section 6, and selects the next separation candidate section k+
2, each separation candidate position x (k+2, i _k+2 ) (however,
After _the separation candidate position _x ( k, i _k ), the separation candidate position x at which the evaluation scale U (0, n) of the plurality of separation candidate positions x (n, i _k ) within the above-mentioned terminal candidate section is the minimum.
Separation candidate position x (n, i _o ) that is detected as the end point position in the interval where (n, i _o ) is evaluated and becomes the end point position
The series of optimal separation candidate positions that reach the separation candidate position x (n-1, _{i o -1 )} , .

Next, the control unit 10 controls the aforementioned end point position x(n,
A separation candidate that will be the start of the substring image to be separated next in a certain range set based on the character pitch P within the blank allowable interval from i _o ) to the start of the first detected character block. It is stored in the optimum separation position information register 26 as section 0, and commands are given to perform the operations described above. In this way, the character separation position of the character string image stored in the character string image memory 2 is stored in the character separation position register 27, and the height and character of each character block stored in the above-mentioned character block information register 21 are stored. By using the character separation position stored in the separation position register 27, it is possible to separate each character.

FIG. 5 is a logical block diagram showing a specific embodiment of the evaluation scale calculating section 8 in FIG.

As mentioned above, the separation candidate position information register 25
, each separation candidate position x(k+
1, i _k+1 ) (where i _k+1 =1...h _k+1 ), the control unit 10 shown in _FIG . ₊₁ ) is transferred to a predetermined position in the distance calculation unit 81 and separation candidate position group register 261, and separation candidate section k+1 is stored in a predetermined position in the stage register 80 and separation candidate position group register 261. When the separation candidate position x (k+1, i _k+1 ) is stored in the distance calculation unit 81, the control unit 1
0, each separation candidate position x(k,
i _k ) (where i _k =1...h _k ) are sequentially calculated by the distance calculation unit 81
will be forwarded to. Here, the optimum separation position information register 26 shown in FIG. 3 is the separation candidate position group register 2.
61, a connection information group register 262, an optimal statistics group register 263, and an optimal evaluation value group register 264. The distance calculation unit 8 calculates the separation candidate position x
Distance d(k, k+ ₁ ; i _k , i _k+1 )=x(k+1,
i _k+1 )−x(k, i _k ).

The statistics calculation unit 82 calculates the average value μ _d (0, k+1; i ₀ , i _k+1 ) and Calculate the cumulative sum of squared distances D(k). That is, the average value μ _d (0, k+
1; i ₀ , i _k+1 ) is the separation candidate position x (k, i _k ) stored in the read optimal statistical group register 263
The average value μ ^* _d (0, k; i ₀ , i _k ), the output d (k, k+1; i _k , i _k+1 ) of the distance calculation unit 81 and the separated sections k+1 and k which are the contents of the stage register 80 Using the calculation formula μ _d (0,k+1;i ₀ ,i _k+1 )=1/k+1{k・μ ^* _d (0,k;i ₀ ,i _k )+d(k,k +1; i _k , i _k+1 ). On the other hand, the cumulative sum of squared distances D
(k+1) is the read optimal statistical group register 2
Using the squared cumulative sum D ^* (k) of the distance at the separation candidate position x(k, i _k ) stored in 63 and the output d(k, k+1; i _k , i _k+1 ) of the distance calculation unit 81, It is calculated using the calculation formula D(k+1)=D ^* (k)+d (k, k+1; i _k , i _k+1 ). The average value μ _d (0, k+1; i ₀ , i _k+1 ) and the cumulative sum of squares of distances D(k+1) calculated by the statistics calculation unit 82 are stored in the statistics storage register 82, respectively. Evaluation value calculation unit 84
calculates the value of the evaluation scale U(0, k+1) based on the above-mentioned formula (3-2). That is, the evaluation value U(0, k+1) is calculated using the character pitch P shown in FIG. Calculation formula U(0,k+1)=β・(D(k+1)/k+1−μ ² _d (0, k+1; i ₀ , i _k+1 )) ² +(1−β) ・(μ _d ( 0, k+1; i ₀ , i _k+1 )−P) ² . Next, in the comparison section 85,
The evaluation value output from the evaluation calculation unit 84 is compared with the contents of the minimum evaluation value register 86, and the evaluation value calculation unit 84
If the output is smaller than the contents of the minimum evaluation value register 86, the output signal 8 of the output signal line 851 is
Turn 51S “ON”. It is assumed that the contents of the minimum evaluation value register 86 are initially set to a very large value.

When the output signal line 851S is turned “ON”, the gate circuit 53 is opened and the output of the evaluation value calculation unit 84 is transferred to the minimum evaluation value register 86.

Also, when the output signal 851S turns “ON”,
Average value μ _d stored in statistics storage register 83
(0, k+1; i ₀ , i _k+1 ) and cumulative sum of squares of distance D
(k+1) is stored in the minimum statistic register 88 by opening the gate circuit 52.

Furthermore, when the output signal 851S becomes “ON”,
By opening the gate circuit 51, the position information k and i _k at the separation candidate position x (k, i _k ) of the separation candidate section k transferred to the distance calculation unit 81 is stored in the connection information register 87.

The above operation is performed at the separation candidate position x (k, i _k ) of the separation candidate section k stored in the optimal separation position information 261.
(However, this is performed for i _k =1...h _k .

Next, _the control unit 10 shown in FIG ^.
(0, k+1; i ₀ , i _k+1 ) and the minimum statistics register 8 as the cumulative sum of squares of the optimal distance D ^* (k+1)
8 is transferred to the minimum statistics group register 263, and the contents of the minimum evaluation value register 86 are transferred to the minimum evaluation value group register 264 as the optimal evaluation value of the separation candidate position x (k+1, i _k+ 1 ). , Furthermore, the separation candidate position x (k+1, i _k ) in the separation candidate section k
The contents of the linkage information register 87 are transferred to the linkage information group register 263 as the optimal separation path information to the linkage information register 263.

Next, the contents of the minimum evaluation value register 86 are set to an initial value (a very large value).

By repeating the above operations, all separation candidate positions x(k+1,
i _k+1 ) (where i _k+1 =1, 2...h _k+1 ), an optimal evaluation value and an optimal separation path can be obtained.

It goes without saying that another specific method of implementing the present invention is to use an ordinary microcomputer.

As described above, by applying the present invention, even if there is contact between characters or if one character is split into two or more characters, it is possible to easily and stably It becomes possible to separate the

[Brief explanation of drawings]

Figure 1 uses an example of a partial string image,
A diagram for explaining an example of a separation candidate section setting method for character separation in the present invention, FIG. 2 is a diagram for explaining the principle for extracting an optimal character separation position in the present invention, and FIG. FIG. 4, a logical block diagram illustrating a specific embodiment of the invention, shows an example of a terminal candidate section extracted by the terminal candidate section extracting section 6 in FIG. FIG. 5 is a logical block diagram showing a specific embodiment of the evaluation scale calculating section 8 in FIG. In the figure, 1 is a scanning unit, 2 is a character string image memory, 3 is a character block extraction unit, 21 is a character block information register, 4 is a character pitch detection unit, 22 is a character pitch information register, 30 is a parameter information register, 5 23 is a permissible interval extraction unit, 23 is an allowable interval information register, 6 is a terminal candidate interval extractor, 24 is a terminal candidate interval register, 7 is a separation candidate interval extractor, 25
8 is a separation candidate position information register, 8 is an evaluation scale calculation unit, 26 is an optimal separation position information register, 27 is a character separation position register, 10 is a control unit, 80 is a stage register, 81 is a distance calculation unit, 261 is a separation candidate position group register, 262 is a connection information group register, 263 is an optimal statistics group register, 264 is an optimal evaluation value group register, 82 is a statistics calculation unit, 83
is a statistics storage register, 84 is an evaluation value calculation unit, 8
5 is a comparison section, and 86 is a minimum evaluation value register.

[Claims]

1. A cash storage section that can be pulled out to the outside of the device when collecting cash; a cash accounting mechanism that stores, dispenses, and transports cash between the cash storage section and a cash deposit/withdrawal port; and the cash transaction mechanism. a discrimination section that discriminates the type of cash being conveyed on the cash conveyance route of the mechanism; and a counting means that counts the amount of cash stored in the cash storage section by receiving a discrimination result signal from the discrimination section. A cash handling device having: a detection means for detecting that the cash storage section has been pulled out of the device; and a clearing means for clearing the count value of the counting means upon receiving a detection output emitted by the detection means. A cash handling device comprising: means.