JP2600375B2 - Threshold value determination device - Google Patents

Description

Description: BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to determination of a threshold value, for example, when binarizing image density in image recognition.

Hereinafter, for convenience of explanation, a case where a grayscale image is used as an input signal pattern will be described.

[Conventional technology]

FIG. 10 shows a conventional example of this type of apparatus.

FIG. 10 is a block diagram showing an outline of a conventional threshold value determining apparatus. In FIG. 10, (1) shows a signal pattern (2) obtained by converting a grayscale image into a signal, A histogram creating unit for creating a level density histogram (hereinafter, appropriately referred to as a histogram), and (3) for converting a grayscale image into an optimal binary image based on the density histogram created by the histogram creating unit (1). It is a threshold value determination unit that determines a threshold value.

As a method of determining a threshold based on the density histogram performed in the threshold determining unit (3), there are a P tile method, a mode method, and a discriminant analysis method.

In the P-tile method, as shown in FIG. 11, a density level at which the cumulative frequency from the lower or higher density level in the density histogram accounts for t percent of the entire distribution is determined, and the density level is determined as a threshold value. It is a method.

The modal method finds the valley of the density histogram, that is, the density level that is the minimum value of the graph, based on the premise that the density histogram shows bimodality with sufficient depression, and thresholds the density level. It is a method to make it a value.

The discriminant analysis method is a method in which, when a histogram is divided into two by a certain threshold value, a threshold value that maximizes the variance between the two groups is determined as an appropriate threshold value.

[Problems to be solved by the invention]

In the above-described conventional threshold value determination apparatus, a method of determining a threshold value using a histogram, such as a P tile method, a mode method, and a discriminant analysis method, is used.
In the tile method, it is necessary to know in advance the proportion of the object in the grayscale image, and in the mode method, it is difficult to determine the threshold value when the density histogram is not smooth or the valley is not noticeable. The discriminant analysis method can be applied even when the valley of the histogram is not conspicuous. However, when the ratio between the object and the background in the image is extremely different, or when the density difference between the object and the background is small, etc. There was a problem that an appropriate threshold could not be determined.

SUMMARY OF THE INVENTION The present invention has been made to solve the above-mentioned problems, and the input signal pattern has a very different ratio of the number of pixels indicating the signal level to be divided, or the level of the signal level to be divided. It is an object of the present invention to obtain a threshold value determining device capable of appropriately determining a threshold value even when the difference between the two is very small.

[Means for solving the problem]

A threshold value determining apparatus according to the present invention includes: a binary pattern trial production unit that converts a grayscale image into a plurality of binary images corresponding to the plurality of threshold values by using the plurality of threshold values; A feature value calculation unit that extracts a feature value relating to the arrangement state of pixels “0” and “1” in the vicinity area of each pixel on the binary image, and the plurality of thresholds by evaluating the feature quantity A threshold value determining unit that selects one of the values, and in the feature value calculating unit, the number F _{1 of} pixels that is “1” among the pixels on the binary image with respect to the threshold value T
(T) is determined, and the number of pixels “1” is determined from the pixel “1” as a central pixel and the number of pixels “1” among pixels adjacent to the central pixel, and this number is summed up for all the central pixels.
F ₂ (T) is calculated, and the upper threshold value determination unit calculates the evaluation function P (T) = F ₂ (T) / F ₁ (T) to calculate the threshold value of the evaluation function P (T). In this case, one of the threshold values T is selected by evaluating the change state of T.

[Action]

In the present invention, the threshold value determining device converts the grayscale image into a plurality of binary images corresponding to the plurality of threshold values in the binary pattern prototype unit, and the feature amount calculation unit converts the grayscale image into a binary image corresponding to the threshold value T. The number F ₁ (T) of pixels that are “1” among the pixels on the value image is obtained, and the pixel that is “1” is the center pixel and “1” is the pixel that is adjacent to the center pixel.
Is obtained, and the sum F ₂ (T) is calculated by summing this number for all central pixels.
An evaluation function P (T) = F ₂ (T) / F ₁ (T) is calculated and evaluated from a change state of the evaluation function P (T) with respect to the threshold value T, and one of a plurality of threshold values is evaluated. Select a value to determine the optimal threshold.

〔Example〕

FIG. 1 is a block diagram showing an outline of a threshold value determining apparatus according to an embodiment of the present invention.

In this embodiment, as in the conventional apparatus, a case will be described in which a grayscale image is used as an input signal pattern and the grayscale level is binarized.

In FIG. 1, (2) indicates a signal pattern obtained by converting a grayscale image into a signal, and (4) indicates a plurality of different set values set in advance. The input signal pattern (2) is based on each set value. A binary pattern prototype unit as a multi-value pattern prototype unit for creating and outputting a value image pattern (hereinafter, appropriately referred to as a binary pattern), (5) a plurality of types of binary patterns output from the binary pattern prototype unit (4) A binary pattern is input, and for each binary pattern, a feature quantity relating to an array of two types of information “1” and “0” included in a plurality of pixels constituting the binary pattern is calculated. (3) is a threshold value deciding unit, which is different from that of the device of the prior art. Of the grayscale image input as the signal pattern (2) based on the Determining a threshold value for performing a binary patterned.

Next, the operation of the threshold value determining device according to this embodiment will be described with reference to FIGS.

_First , in the binary pattern prototype unit (4), a plurality of different set values T are set, for example, as T ₁ , T ₂ ,..., Tn in order from a lower density level to a higher density level. Normally, the minimum value of the density level of the input grayscale image is T ₁ , and the maximum value is Tn.

Next, a binary pattern is created by binarizing the input signal pattern (2) with each set value T,
Output all created binary patterns.

Each binary pattern is composed of a plurality of pixels that represent the grayscale image in a subdivided manner, and each pixel has a darker level than the set value T based on two levels of grayscale levels divided by the set value T. Part is “1”, light level part is “0”
Respectively.

The plurality of output binary patterns are used as a feature value calculation unit (5).
Is input to In the feature value calculation unit (5), first,
For each of the value patterns, for a plurality of pixels constituting the binary pattern, a feature amount relating to the arrangement state of “0” and “1” in the vicinity of 3 × 3 of each pixel is calculated.

As shown in FIG. 2, point X = (x, y) on the binary pattern
Is the central pixel (the data is bo (X)) and 3 ×
3 Neighbor (pixels adjacent to the center pixel) data is represented by b _i (X) (i
= 0, 1, ..., 8) is represented by bo (X) and b _i (X) is "0"
Or, it has a value of “1”, and there are 2 ⁹ = 512 different arrangement states of “0” and “1” near 3 × 3.

Next, for a block whose central pixel in the vicinity of 3 × 3 is “1”, the total number F ₁ (T) and the eight neighbors b _i (X) (i = 1, 2,. The total sum F ₂ (T) of the number of “1” pixels in 8) is calculated.

F ₁ (T) and F ₂ (T) are represented by the following equations (1) and (2).

In order to calculate the feature value F ₁ (T), as shown in FIG. 3, the total sum of the number of blocks whose central pixel is “1” among the blocks of 3 × 3 neighboring data may be obtained. Further, in order to calculate the feature value F ₂ (T), as shown in FIG. 4, the number of “1” s near the center pixel of a block whose center pixel is “1” is counted, and the number of “1” is counted. What is necessary is just to find the sum.

The feature values F ₁ (T) and F ₂ calculated by the feature value calculation unit (5)
(T) is input to the threshold value determining unit (3). The threshold value determining unit (3) calculates the feature amount P (T). Feature P
(T) is represented by the following equation (3).

P (T) = F ₂ (T) / F ₁ (T) (3) However, when F ₁ (T) = F ₂ (T) = 0, P (T) = 0.

For example, consider the case where the density level of each pixel of the signal pattern of a grayscale image divided into 8 × 8 pixels is as shown in FIG. In FIG. 5, 0, 1, 4,
Numbers such as 5 indicate the density level of each part of the grayscale image corresponding to each pixel, and 0 is the lightest level, and the higher the numerical value, the darker the level.

When this signal pattern is converted into a binary pattern based on, for example, seven set values from 0 to 6, a binary pattern as shown in FIG. 6 is obtained for each set value T. In FIG. 6, a portion indicated by a hatched portion (6) is a portion indicating a pixel indicating "1", and a portion indicated by a white background is a portion indicating a pixel indicating "0".

The feature values F ₁ (T) and F ₂ (T) are obtained from the binary pattern corresponding to each set value T, and the feature value P (T)
Is calculated as shown in FIG. 7 (set values T and P
FIG. 8 is a graph showing the relationship with (T).
Next, in order to examine the change state of P (T) with respect to the set value T, a first derivative P '(T) and a second derivative P' (T) of P (T) are obtained.
Is obtained, the values shown in FIG. 7 are obtained.

P ′ (T) and P ″ (T) are expressed by the following equation (4) and (5)
It is expressed by an equation.

However, in this embodiment, ΔT = 1.

In this embodiment, the set value at which the change rate of the connected state of the pixel which is “1” is minimal is the point where the density rise between the background and the object is the steepest in the grayscale image, that is, the grayscale image is 2 It is based on the theory that it is an optimal threshold value for value patterning.

That is, a set value at which the value of P "(T) changes from positive to negative may be determined as the optimum threshold value. In this embodiment, as is apparent from FIGS. 7 and 8, the optimum threshold value is obtained. The value is T = 3.

As described above, in the threshold value determining apparatus according to the present embodiment, the threshold value for performing the binary patterning process on the signal pattern of the grayscale image is determined by a plurality of preset threshold values. A plurality of binary patterns corresponding to different set values are prototyped, and this is performed based on an analysis of the plurality of binary patterns.

This eliminates the need to create a density histogram, and when the ratio of the target area to the background area in the grayscale image is very different, or when the difference between the target density level and the background density level is very small. However, the optimum threshold can be determined.

Note that, in the above embodiment, a case was described in which one binary pattern prototype unit (4) and one feature amount calculation unit (5) were provided, as shown in FIG. 1, but as shown in FIG. A plurality of binary pattern prototyping units (4) and feature amount calculation units (5) are arranged in parallel, and binary patterns corresponding to different set values are respectively applied to one input signal pattern (2). It is also possible to create them at the same time. In this case, there is an advantage that the time required for determining the threshold can be greatly reduced.

Further, in the above embodiment, as an example of the specific number of set values used for determining the threshold value in the threshold value determining unit (3), the case where all the preset set values are used has been described. It is not always necessary to use the total number of set values as the specific number of set values used to determine the threshold value. For example, the threshold value is determined by, for example, selecting and using a set value of the density level in a specific range. It is also possible to appropriately select and use a necessary and sufficient number of set values.

Further, in the above embodiment, a case where a grayscale image is used as a pattern of an input signal has been shown, and a case has been described in which the array of pixels constituting the created binary pattern is two-dimensional. The present invention can be applied to a case where the arrangement of pixels is one-dimensional, for example, and the same effect can be obtained.

〔The invention's effect〕

As described above, the present invention uses a plurality of threshold values to convert a grayscale image into a plurality of binary images corresponding to the plurality of threshold values. "0" in the area near each pixel on the binary image
A feature value calculating unit that extracts a feature value related to the arrangement state of the pixels “1” and “1”; and a threshold value determining unit that evaluates the feature value and selects one of the plurality of threshold values. , In the above-mentioned feature value calculation unit,
Number F ₁ (T) of pixels that are “1” in each pixel on the value image
And the number of pixels "1" among pixels adjacent to the pixel "1" is determined as the center pixel, and the sum F is obtained by summing this number for all the center pixels.
₂ (T) is calculated, and the evaluation function P (T) = F ₂ (T) / F ₁ (T) is calculated in the threshold value determining unit to obtain the evaluation function P
Since one value among a plurality of threshold values is selected by evaluating from the change state of the threshold value T of (T), the optimum threshold value is set even when the area ratio between the target and the background is significantly different. The value can be determined, and there is an effect that a good binary image can be obtained.

[Brief description of the drawings]

FIG. 1 is a block diagram showing an outline of a threshold value determining apparatus according to an embodiment of the present invention, FIG. 2 is an explanatory view showing 3 × 3 neighborhood data of a pixel of a binary pattern, FIG. 3 and FIG. FIG. 5 is a diagram for explaining a method of obtaining a feature amount. FIG. 5 is a diagram showing an example of a signal pattern divided into 8 × 8 pixels. FIG. 6 is a diagram showing a signal pattern shown in FIG. FIG. 7 is a diagram showing each binary pattern when a binary pattern is formed by values, FIG. 7 is a correspondence diagram showing a relationship between each set value and a feature amount calculated based on the set value, and FIG. 8 is a set value and a feature amount. FIG. 9 is a block diagram showing an outline of a threshold value determining apparatus according to another embodiment of the present invention, and FIG. 10 is a block diagram showing an outline of a conventional threshold value determining apparatus. FIG. 11 is a block diagram showing the P-tile method, which is one of the conventional threshold value determination methods. Is a diagram showing an example of a graph for. In the figure, (3) is a threshold value determination unit, (4) is a binary pattern trial production unit, and (5) is a feature value calculation unit. In the drawings, the same reference numerals indicate the same or corresponding parts.

Claims (1)

(57) [Claims] 1. Each pixel of a gray-scale image is set to "0" and "1" by 2
In a threshold value determining device for determining a threshold value for conversion, a plurality of threshold values are used to convert the grayscale image into a plurality of binary images corresponding to the plurality of threshold values. A binary pattern prototyping unit, a feature amount calculating unit that extracts a feature amount relating to the arrangement state of pixels “0” and “1” in a region near each pixel on the binary image for each of the binary images,
And a threshold value determining unit that evaluates the feature value and selects one of the plurality of threshold values, wherein the feature value calculation unit calculates a value of each pixel on the binary image with respect to the threshold value T. Among them, the number F ₁ (T) of the pixels that are “1” is obtained, and the number of the pixels that are “1” among the pixels adjacent to the pixel “1” is determined with the above-mentioned pixel “1” as the center pixel. The sum F ₂ (T) of all the above central pixels is calculated, and the threshold value determining unit calculates an evaluation function P (T) = F ₂ (T) / F
₁ (T) is calculated and evaluated from the state of change of the evaluation function P (T) with respect to the threshold T, and one of the thresholds T is selected. Decision device.