JPH053632B2 - - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION [Field of Industrial Application] The present invention relates to a threshold range detection device, and more particularly to a threshold range detection method for determining the threshold of a multi-value pattern in character recognition preprocessing of an optical character reading device. .

[Conventional technology]

Conventionally, this type of threshold detection method has determined the optimal threshold by detecting changes in the density of character patterns and line element widths, and evaluating them based on certain evaluation criteria.

[Problem that the invention seeks to solve]

The conventional threshold detection method described above does not evaluate the pattern after determining the threshold, so as a result, noise may remain in the pattern sliced by the threshold, or some line elements may be missing. There are drawbacks. Therefore, a phenomenon occurs in which recognition becomes unrecognizable due to noise, or unrecognizability or erroneous recognition occurs due to lack of line elements.

[Means for solving problems]

The threshold range detection device of the present invention has a first method that detects the number of line element blocks of each binary pattern by sequentially slicing a multi-value pattern obtained from a scanning device using a plurality of threshold values.
means for calculating the local maximum value and local minimum value of the number of line element blocks according to a change in the number of line element blocks, and a group of threshold values in the vicinity of a threshold value for giving the local minimum value between two of the maximum values; and second means for detecting, as an appropriate threshold range, any one of a threshold value group giving the number of line element blocks in the vicinity of the number of line element blocks corresponding to the minimum value between the maximum values.

[Operation/principle of the present invention]

The present invention aims to evaluate the entire binary pattern sliced by a threshold value and obtain an appropriate threshold range based on the result. When a multi-value pattern is sliced by changing the threshold value, The authors focus on the fact that the number of noise increases, and if the threshold value is too high, some line elements are missing, and as a result, the number of line element blocks in the entire binary pattern including noise changes.

〔Example〕

Next, the present invention will be explained with reference to the drawings. FIG. 1 is a block diagram showing an embodiment of the threshold range detection device of the present invention.

In the figure, a multi-value scanning signal 100 is obtained from an external scanning device and temporarily stored in a multi-value pattern storage means 10, and then a multi-value pattern signal 101 is outputted to one input of a slice circuit 30. The other input of the slicing circuit 30 has a threshold specifying unit 2.
A threshold signal 102 is given from 0. This threshold signal 102 is also applied to one input of the detection information buffer 50. The slice circuit 30 receives the threshold signal 10
2, the multi-value pattern signal 101 is sliced to output a binary pattern signal 103, which is input to the line element block number detection section 40. The line element block number detection unit 40 temporarily stores the entire binary pattern of the specified threshold value and sequentially retrieves it, detects the number of line element blocks included in the entire binary pattern including noise, and generates a block number signal. 104 is output and applied to the other input of the detection information buffer 50. The detection information buffer 50 temporarily stores the specified threshold value θ and the number B of line element blocks. The output 105 of the detection information buffer 50 is input to the detection information storage means 60.

In this embodiment, since the multi-value pattern has 9 levels of values, the threshold value can specify a range of 8 levels. Therefore, the threshold specifying unit 20 is controlled to change the threshold 8 times from 1 to 8, and extract 8 types of binary patterns from one multi-value pattern. This result detection information storage means 60
Eight sets of data are stored in . When the output 106 of the detection information storage means 60 is given to the calculation section 70, the calculation section 70 calculates the maximum value and minimum value from the 8 sets of specified threshold values θ and the number of line element blocks B, and then
Find an appropriate threshold range.

Next, the processing contents of this embodiment will be explained in detail.

Figures 2 and 3 are diagrams showing the first and second examples of the relationship between the threshold value, the binary pattern, and the number of line element blocks, respectively, and Figure 4 is a general representation of the relationship between the threshold value and the number of line element blocks. FIG.

In Figures 2 and 3, which respectively show pattern examples of the number "0" and the katakana "ru", the threshold θ is set so that the smaller the number, the darker the binary pattern becomes. When the threshold value θ is between 1 and 3, noise appears on the paper surface, and when the threshold value θ is between 6 and 8, lines are missing (faint). The number B of line element blocks in each pattern is shown as the sum of the number of independent line elements increased or reduced as a result of line elements being separated due to the number of independent noises and blurring. The situation is similar in FIG. 3, but the absolute number of line element blocks B is different due to the difference in the pattern.

The detection information storage means 60 shown in FIG. 1 stores eight sets of data for each input pattern in pairs of the threshold value .theta. and the number of line element blocks B as described above.

Figure 4 shows such a threshold value θ and the number of line element blocks B.
The relationship is generalized and expressed, and the number of line element blocks B is illustrated as a function B(θ) of the threshold value θ.
In the same figure, the maximum value G ₁ of the function B(θ) has the largest number of noises, and at a threshold θ below this, connection also occurs to the noises, and the number of line element blocks of the noise B
is decreasing. Further, at the maximum value _G2 , line elements are separated most often due to blurring, and when the threshold value θ is higher than this, the line elements become too blurred and disappear. These phenomena are also shown in FIGS. 2 and 3.

The minimum value L is between the maximum value G ₁ and the maximum value G ₂ ,
The threshold value θ ₀ indicating the minimum value L is not necessarily the optimal threshold value.
This is because when character line elements are too close together as shown in FIG. 3, lowering the threshold value θ may cause the line elements to connect. However, it is clear that an appropriate threshold value is in the vicinity of the threshold value that gives the minimum value to the function B(θ), and this is defined as the appropriate range. Also, as can be seen from the examples in Figures 2 and 3 above, in the range between the local maximum values G ₁ and G ₂ , it is possible to obtain useful patterns by performing other processing, such as noise removal or smoothing. Therefore, it can be thought of as an effective range that includes the appropriate range.

Once the minimum value L is determined in this manner, the definition of the vicinity of the minimum value can be changed in accordance with the characteristics and functions of the product to which the present invention is applied.

Next, FIG. 5 is a flowchart showing the processing operation of the calculation unit 70 in _{FIG .}
Then, the minimum value L is calculated, and a group of threshold values near the threshold that gives the minimum value L (here, for example, ±1 level with respect to the minimum value L) is set. Thereafter, this threshold value group is detected and output as an appropriate threshold value range. Note that the neighborhood of the number of line element blocks B ₀ having the minimum value L may be defined as, for example, B ₀ ±1 block, and a threshold group giving this may be detected as an appropriate threshold range.

〔Effect of the invention〕

As explained above, the present invention has two methods that are effective for recognition.
Since the threshold value can be determined by pre-evaluating the state of the value pattern, that is, noise and lack of line elements, it not only enables more accurate recognition but also reduces waste such as redoing recognition processing. Furthermore, since the evaluation results of patterns for which appropriate thresholds have been detected are known in advance, more effective thresholds can be specified by using this as reference information when retrying. Furthermore,
By changing the application method of other preprocessing, such as noise removal or smoothing, based on the preliminary evaluation results, the accuracy can be further improved.

[Brief explanation of the drawing]

FIG. 1 is a block diagram showing one embodiment of the threshold range detection device of the present invention, and FIGS. 2 and 3 show first and second examples of the relationship between the threshold, the binary pattern, and the number of line element blocks, respectively. 4 is a diagram showing a general representation of the relationship between the threshold value and the number of line element blocks, and FIG. 5 is a flowchart showing the processing operation of the calculating section in FIG. 1. DESCRIPTION OF SYMBOLS 10...Multi-value pattern storage means, 20...Threshold value specification section, 30...Slice circuit, 40...Line element block number detection section, 50...Detection information buffer, 6
0...Detection information storage means, 70...Calculation unit.

Claims (1)

[Claims] 1. A first means for detecting the number of line element blocks of each binary pattern by sequentially slicing the multi-value pattern obtained from the scanning device using a plurality of threshold values; After calculating the local maximum value and local minimum value of the number, a group of threshold values in the vicinity of the threshold value giving the local minimum value between the two local maximum values, and a group of threshold values in the vicinity of the line element block number corresponding to the local minimum value between the two local maximum values. a second means for detecting any one of the threshold values giving the number of line element blocks as an appropriate threshold range.