JP2646569B2 - Multi-value image evaluation device and threshold value determination device using the device - Google Patents

Description

Description: TECHNICAL FIELD The present invention relates to an image processing technique applied to an eye, a visual inspection device, and the like of an intelligent robot. In particular, the present invention relates to a grayscale image, a distance image, and the like. Multi-valued image evaluation device used to evaluate the image quality of the multi-valued image, and a threshold value determination device for determining a threshold value for processing the multi-valued image using the multi-valued image evaluation device And about.

<Conventional technology> Conventionally, in order to evaluate the image quality of a multi-valued image, based on an image of a specific shape that is known in advance, the image of the specific shape is compared with a multi-valued image captured by a camera. A method for evaluating the image quality of the multi-valued image has been proposed (Japanese Patent Laid-Open No. Sho 62-62).
No. -80771).

In addition, as a conventional method of determining a threshold value for performing the binarization processing on the multi-valued image, a p-tile method and a mode method are generally known.

The p-tile method is an effective method when the proportion p of the object in the image is known, and when the grayscale image is binarized, the number of pixels that becomes “1” is relative to the total prime number. The threshold value t is determined so that the ratio becomes p.

Figure 13 shows a histogram h ₁ of the brightness of the gray-scale image, the brightness on the horizontal axis of, the number of pixels on the vertical axis are taken. p-
To perform calculation of the threshold in tile method, after initially seeking brightness histogram h ₁ such as shown in the figure, in the cumulative number of pixels (figure from the larger brightness value, by hatching The threshold value t is determined so that the ratio (p.) Shown in FIG.

Next, the mode method focuses on the fact that two peaks appear in the brightness histogram when there is a contrast between the object and the background, and sets a threshold value at the position of an intermediate trough. Is what you do.

Figure 14 shows the two ridges 2 and its intermediate position troughs 3 and brightness appeared in the histogram h _2, to set the threshold t to the position of the valley 3 Separates the object from the background.

As another method for calculating the threshold value relatively well, a large-scale threshold method is known.

This method does not group the brightness histogram into bright and dark areas by valleys,
By using the discriminant analysis binarization method, the brightness histogram is divided into two groups so that each group is strongly united when seeing both sides from a certain boundary, and the centers of both groups are far apart. The value is set as an optimum threshold value. Here, “strongly united” means that the sum of the variances in each group is small, and “far away from the center” means that the variance between each group is large. A threshold k that maximizes the represented λ is determined.

λ = σ _B ² / σ _W ^{2 where} σ _W ² is the average of the variances of the two groups (referred to as “intra-class variance”), and σ _B ² is the average of the two groups The square of the distance between values (variance between classes)
In the histogram h ₃ brightness normalized shown in FIG. 15, the standard deviation of the first group G1 sigma _1, the first group G1
Is ω ₁ , the standard deviation of the second group G2 is σ ₂ ,
When the occurrence probability of the group G2 and omega _2, are respectively given by the following expression.

σ _W ² = ω ₁ σ ₁ ² + ω ₂ σ ₂ ² ... σ _B ² = ω ₁ ω ₂ (μ ₁ −μ ₂ ) ² ... Specifically, it can be obtained by finding a threshold value k that maximizes the following equation.

Note in the above formula, the mu _T average value of the entire histogram of brightness normalized, omega (k) is zero-order cumulative moments of the histogram, mu (k) is a first-order cumulative moments of the histogram.

In addition to the above methods, a method has also been proposed in which poor image quality is used as a threshold calculation standard.

This method uses a multi-level image (for example, 256 levels)
Each of the brightness values is used as a threshold to generate a binary image by binarizing the image, and for each binary image, the quality of the image is determined by one of the number of connected components, the length of the boundary line, and the complexity of the DF expression. After the calculation, the optimum threshold value is determined from the calculation result.

FIG. 16 shows the poor image quality of the binary image with respect to each threshold value, and the threshold t at which the poor image quality is minimized in the curve 1 in FIG. 16 is extracted as the optimal threshold value. .

<Problems to be Solved by the Invention> The above-described method of evaluating a multi-valued image is effective only when the size and shape of the evaluation target are known, and cannot be applied when those are unknown. Therefore, it has been considered that it is impossible to evaluate the image quality of a multivalued image obtained by imaging an unknown object.

In addition, with respect to the above-described threshold value determination method, the p-tile method cannot be applied unless the size of an object in an image is known, and there is a disadvantage that its use is limited.

In addition, the mode method has a disadvantage that when many undulations appear in the brightness histogram or when no undulations appear at all, it becomes difficult to specify a valley, and it is difficult to determine a threshold value.

Further, in the rule of the large threshold, not only the calculation of the equation takes time, but also the calculation is repeatedly performed while changing k, so that the calculation of the optimum threshold requires a large amount of calculation time. Moreover, this method aims at finding a threshold for clearly dividing the brightness histogram into two groups, and for finding a threshold for obtaining a good binary image. However, the threshold value obtained in this way does not always lead to a good binary image.

In this regard, a method using the poor image quality as a reference for calculating the threshold value can theoretically always produce a good binary image. However, in this type of conventional method, a smooth image as shown in FIG. An ideal curve is not always obtained, and there are cases where it becomes difficult to determine a threshold value because local minima appear at a plurality of locations. Further, since a binary image is generated many times (for example, 256 times) and the image quality is calculated each time, the processing time is long. In addition, the definition of the image quality is not preferable in terms of computational efficiency. There is a problem that it takes time.

The present invention has been made in view of the above problems, and enables evaluation of a multi-valued image even when the shape and size of an object are unknown, and also enables fast and stable determination of an optimal threshold value. It is an object of the present invention to provide a multi-valued image evaluation device capable of performing the above-mentioned, and a threshold value determining device using the device, so that a good binary image is always generated.

<Means for Solving the Problems> The present invention is a multi-valued image evaluation apparatus for evaluating the image quality of a multi-valued image, comprising: a window scanning means for setting and scanning a window of a predetermined shape on the multi-valued image; Extracting a brightness level using predetermined conditions for an image in a window for each scanning position, and generating first information indicating the number of windows corresponding to each brightness level in the multi-valued image based on the extraction result. A first information generating means for determining a smoothness of an image in a window at each scanning position by using a binary pattern obtained using the brightness level extracted under the predetermined condition as a threshold value; After determining whether or not the pattern matches a legal pattern or an illegal pattern, by associating the determination result with the threshold value, the level of each brightness level is set. A second information generating means for generating second information indicating the degree to which a binary pattern obtained as a threshold matches a legal pattern or an illegal pattern; and And third information generating means for generating third information for giving an image quality evaluation value using information generated by each of the second information generating means.

According to a second aspect of the present invention, there is provided a threshold value determining apparatus for determining a threshold value for processing a multi-valued image. Is determined.

<Operation> In the multi-valued image evaluation apparatus, the window scanning unit sets and scans a window of a predetermined shape on the multi-valued image.

In this case, the first information generating means generates information indicating the number of windows corresponding to each brightness level in the multi-valued image based on the brightness level extracted for the image in the window for each scanning position. Further, the second information generating means is arranged such that a binary image obtained by using the brightness level as a threshold value for the image in the window at each scanning position matches a legal pattern or an illegal pattern for determining smoothness. After the determination as to whether or not to perform the binarization process, by associating the result of the determination with the threshold value, a binarization process is performed for each brightness level using the level as a threshold value. Information indicating how frequently the value pattern matches the legal or illegal pattern is generated. When the scanning of the window is completed, the third information generating means generates third information for giving an image quality evaluation value based on the information generated by the first and second information generating means. Can evaluate the image quality of the multi-valued image.

Further, in the threshold value determination device of the second invention,
The optimum threshold value that gives the best image quality is determined on the basis of the above information, whereby the optimum threshold value for processing a multi-valued image by one scan can be obtained.

According to the present invention, it is possible to evaluate a multi-valued image and determine a threshold value even when the shape and size of an object are unknown, and to evaluate and evaluate a multi-valued image by one scan of a window. Fast and stable determination of the optimum threshold value is possible.

<Embodiment> Fig. 1 shows an example of the overall configuration of a threshold value determining device using a multi-valued image evaluation device according to the present invention. It includes a legal pattern detection unit 12, an illegal pattern histogram generation unit 13, a brightness histogram generation unit 14, an image quality histogram generation unit 15, a histogram smoothing unit 16, and an optimal threshold search unit 17.

The window scanning unit 11 controls the microcomputer 10
Address generator 19 connected via U bus 18
And an image memory 24 connected to the address generator 19 via an image bus 20, and a first shift register 25 of a FIFO (first-in first-out) system connected to a data bus 23 of the image bus 20. And the first group of latch circuits 26
To 29 and the same FI connected to the last-stage latch circuit 29.
It comprises a second shift register 30 of the FO system and a second group of latch circuits 31 to 34.

The window scanning unit 11 in the illustrated example is for setting a rectangular window 60 as shown in FIG. 5 on the grayscale image stored in the image memory 24 and performing raster scanning. Predetermined pixels A _{0 to} A _{7 for} the image in
Brightness data d ₀ to d ₇ is adapted to be taken out simultaneously. In the figure, the crosses indicate pixels that are not to be processed.

That is, the address generator 19 sequentially generates the address of each pixel in the image memory 24, and outputs the brightness data of the pixel corresponding to the address to the first shift register 25 via the image pulse 20. Of the image bus 20,
The address bus 21 transmits an address, the data bus 23 transmits brightness data, and the control bus 22 transmits an address strobe signal and a read / write signal. The first shift register 25 sets the number of pixels in the horizontal direction of the image memory 24 to n.
(See FIG. 5), the number of stages of the first group of latch circuits 26 to 29 is subtracted from n (n-4), and the output of the brightness data is delayed by the number of stages. Perform the action that causes First-stage four-stage latch circuits 26 to 29
Performs the operation of the shift register as a whole, and the output of the last-stage latch circuit 29 is output to the second shift register 30. The second shift register 30 also has (n-4) stages, and outputs the brightness data to the second group of four-stage latch circuits 31 to 34 by delaying the output by the number of stages.

Thus, the first and second group latch circuits 26-28, 31
To 34 and in which at the same time the brightness data from the second shift register 30. is taken out, the output of the latch circuit 26 to 28 of the first group of luminance data d of the pixel A ₀ to A ₂ in the window 60 ₀ to d ₂ gives also the output of the second shift register 30 and the second group of latch circuits 31 to 34 to give the brightness data d ₃ to d ₇ of the pixel a ₃ to a ₇ in the window 60 become.

In this embodiment, a rectangular window 60 composed of 2 rows × 5 columns of pixels is set. However, the window shape is not limited to this. For example, as shown in FIG. The window 60 may be a rectangular window 60 composed of pixels, or may have another shape.

Brightness data d ₀ to d ₇ of the respective pixels A ₀ to A ₇ together with the output to the illegal pattern detecting section 12, illegal pattern histogram generation unit 13 and the brightness for the brightness data d ₁ of the specific pixel A ₁ It is also output to the histogram generation unit 14.

The illegal pattern detection unit 12 is for determining whether the image in the window 60 is a legal pattern or an illegal pattern. When the image is an illegal pattern, the illegal pattern detection unit 12 outputs “1”. A flag is output, and if the pattern is a legal pattern, a flag of “0” is output. Illegal pattern detector 12 of the illustrated embodiment, as shown in FIG. 2, the eight latch circuits 35-42 which latch each luminance data d ₀ to d _7, luminance data d ₁ of the specific pixel A ₁ And other pixels A ₀ and A ₂ ~
It contains seven comparison circuit 43 to 49 to magnitude comparing the brightness data d ₀ and d ₂ to d ₇ of A _7. These comparison circuits 43 to
49, the brightness data d ₀ and d ₂ to d ₇ of the respective pixels A ₀ and A ₂ to A ₇ of the window 60 brightness data d ₁ specific pixel A ₁ as a threshold for binarizing The brightness data d _i (i = 0, 2, 3, ‥‥, 7) is “1” when d _i ≧ d ₁ and “1” when d _i <d _1. 0 ”binary data f _i
Are output. Note binary data f ₁ of the specific pixel A ₁ is always "1".

Figure 7 shows the process of being converted brightness data d ₀ to d ₇ of the respective pixels A ₀ to A ₇ in the window 60 is binarized into binary data f ₀ ~f _7.

The binary data f _i (i = 0, 2, 3,..., 7) is output to the detection logic unit 50, and the detection logic unit 50
Determines whether or not the image in the window 60 is illegal pattern by the value data f _i, if not legally pattern flag of "1", the flag of "0" if legal patterns, respectively output I do. Here, the legal pattern means a binary pattern which frequently appears on a smooth image as shown in FIG. 8, and an illegal pattern and a binary pattern which frequently appears on an unsmooth image as shown in FIG. It means a pattern. 8 and 9, hatched portions indicate pixels having binary data of “1”, and in the binary pattern of FIG. 9, a hole 61 of “0” is formed inside the pixel region of “1”. ing.

The following equations are the two types of 2 shown in FIGS. 10 (1) to (6).
This is a flag generation logic that assumes a value pattern as a legal pattern, determines whether the binary pattern is a legal pattern or an illegal pattern, and outputs the determination result as a flag.

flag =! _{(! F 0 & f 2 &} ! F 3 & f 4 & f 5 & f 6 ♯! F 0 & f 2 &! F 3 &
_{! f 4 & f 5 & f} 6 ♯! f 0 & f 2 &! f 3 &! f 4 &! f 5 & f 6 & f 7 ♯f 0 &!
_{f 2 & f 4 & f 5} & f 6 &! f 7 ♯f 0 &! f 0 & f 4 & f 5 &! f 6 &! f 7 ♯f 0 &! f 2 & f 3 & f 4 &! f 5 &! f ₆ ♯! f ₇ ) …… In the above formula,! Is NOT operation, & is AND operation, ♯ is OR
The operations are shown respectively.

Each time the detection logic unit 50 detects an illegal pattern and outputs a flag of “1”, the illegal pattern histogram generation unit 13 counts the frequency of occurrence of this flag and generates a histogram of the illegal pattern.

FIG. 3 shows a specific example of the illegal pattern histogram generation unit 13, which includes a counter 51, a memory 52, and a timing control unit 53. The memory 52 is provided with an address input pins A brightness data d ₁ from the latch circuit 27 of the window scanning unit 11 is provided, and a data input pin D of the count data is provided in the counter 51, the illegal pattern each time the detection flag is generated, the data of the address area corresponding to the brightness data d ₁ (occurrence frequency of illegal pattern) is adapted to be rewritten by the counter 51 at that time. Thus the counter 51 when the detection flag illegal pattern occurs, after adding 1 to the value read brightness current value than the area of the memory 52 at the address corresponding to the data d ₁ (power) at that time, the addition The contents of the address area are rewritten with the later value. In the figure, the access control signal is a signal for clearing each area of the memory 52 to zero at the initial stage. Also, R / W of the memory 52 means read / write, CS means chip select, UP of the counter 51 means up count, OE means output enable, LD means load, and the timing control unit 53 is illegal. A predetermined signal is applied to these signal input pins based on a pattern detection flag and an access control signal to control the timing of the illegal histogram generation operation and the like.

FIG. 4 shows a specific example of the brightness histogram generation unit 14, which comprises a counter 54, a memory 55, and a timing generation unit 56, like the illegal histogram generation unit 13. This counter 54 is a latch circuit of the window scanning unit 11.
Each brightness data d ₁ than 27 is given, after adding 1 to the value read current value (brightness of the power) than the area of the memory 55 at the address corresponding to the luminance data d _1,
The contents of the address area are rewritten with the value after the addition. The functions of the memory 55, the counter 54, the timing control unit 56, and the like are the same as those of the illegal pattern histogram generation unit 13, and a description thereof will be omitted.

In this way, the illegal pattern histogram generation unit 13 generates the illegal pattern histogram f _hist (k) on the memory 52. Here, f _hist (k) means the frequency of occurrence of illegal patterns at the threshold value k. In the case of this embodiment, the brightness data of a specific pixel is
Brightness data d ₀ and d ₂ of other pixels with d ₁ as a threshold
To evaluate the pattern by ~d _7, the result is an expression flag
The illegal pattern histogram f
_hist (k) is formed according to the following equation.

f _hist (k) = f _hist (k) + flag However, it is assumed that each element of the illegal pattern histogram is initialized to zero.

The brightness histogram generator 14 generates a brightness histogram hist (k) on the memory 55. Here hist
(K) means the frequency of occurrence of a pixel of brightness k, and this brightness histogram hist (k) is similarly generated according to the following equation.

hist (k) = hist (k) +1 Returning to FIG. 1, the image quality histogram 15 generates an image quality histogram q _hist (k) after the generation of the illegal pattern histogram and the brightness histogram is completed. The image quality histogram q _hist (k) provides an evaluation standard for image quality, in this embodiment, a standard indicating poor image quality, and is generated according to the following equation.

(1) When hist (k) = 0 q _hist (k) = Q ₀ (2) When hist (k) ≠ 0 In the above equation, Q ₀ is a constant (maximum value), and q
_The larger the value of _hist (k), the worse the image quality.

Next, the histogram smoothing processing section 16 in FIG. 1 is for smoothing the image quality histogram q _hist (k) as necessary, and as a result, as shown in FIG. sq _hist (k) is generated.

In FIG. 11, the horizontal axis is k (threshold), and the vertical axis is sq
_hist (k) (bad image quality), and the bad image quality is minimum when the threshold value k _opt is used.

The optimum threshold value searching unit 17 is means for searching for the minimum value of the smoothed image quality histogram sq _hist (k),
This threshold k _opt is searched as the optimum threshold, and the search result is notified to the microcomputer 10 through the CPU bus 18.

The configuration of the above embodiment is such that a multi-valued image such as a grayscale image can be clearly separated into a black and white object and background.
That is, it is based on the idea that the image quality is better as the binary image having a smoother contour is obtained, and the contour portion of the binary image obtained by binarization with the optimum threshold value is determined by another threshold value. The number of illegal patterns is smaller than the outline portion of the binarized binary image. This embodiment is based on the principle that the brightness of an outline pixel of a binary image before binarization is equal to the threshold value used to obtain the binary image. if the contour pixel value image, the image in the window 60 to around a specific pixel a ₁ of the on grayscale images was based generation of the binary image, with the brightness of the pixel a ₁ 2 In the binary image obtained by binarization,
Fewer illegal patterns.

In other words, the fact that that the image in the window 60 centered on the pixel A ₁ on the gray image, illegal pattern is reduced in the binary image obtained by binarizing the brightness of the pixel A ₁ is window brightness of the pixel a ₁
When the image in the window 60 is binarized, it means that the image in the window 60 becomes an optimal binary image. The image quality histogram q _hist (k) is based on such a concept, and the brightness of the pixel A ₁ in the window 60 with few legal patterns, that is, with many illegal patterns, is determined by the window
Even if 60 is spread over the entire image, binarizing with the brightness means that illegal patterns increase. Since the brightness of the pixel of interest in the window 60 does not always have a uniform distribution in the entire image, the illegal pattern histogram f _hist (k) is represented by the brightness histogram h
The image quality histogram q _hist (k) is calculated by normalizing using the data of ist (k).

Note that, in the above embodiment, the extracted brightness data d ₁ of the pixel A ₁ of the specific position per image in the window 60 produces a brightness histogram, the window 60 in the luminance data d ₁ as a threshold value Is binarized, but the invention is not limited to this. For example, a brightness histogram is generated by calculating an average value of brightness data for a plurality of pixels in the window 60, and the average value is thresholded. The image in the window 60 may be binarized as a value.

Next, the operation of the above embodiment will be described.
Stores a shaded image composed of an object and a background. The grayscale image is captured by an imaging device (not shown), and before the binary image is binarized, the binarization threshold value is calculated by the device of this embodiment.

First, the memory 52 of the illegal pattern histogram generation unit 13
After clearing the memory 55 of the brightness histogram generation unit 14 to zero, the window scanning unit 11 performs a raster scan by setting the window 60 on the grayscale image. That is, in the window scanning unit 11, the address generator 19 sequentially generates the address of each pixel in the image memory 24, and outputs the brightness data of the pixel corresponding to the address to the first shift register 25 via the image bus 20. Let it. These brightness data are stored in a first shift register 25 and a first group of latch circuits 26 to 2.
9, the second shift register 30, the second group of latch circuits 31 to 34
To the latch circuits 26 to 26 of the first and second groups.
From window 28, 31 to 34 and second shift register 30, window 6
Brightness data d ₀ to d ₇ of the pixel A ₀ to A ₇ in the 0 is to be retrieved.

Illegal pattern detector 12 fetches these brightness d ₀ to d _7, after binarizing the pixels in the window 60 brightness data d ₁ specific pixel A ₁ as a threshold value, the binary It is determined whether the pattern corresponds to an illegal pattern. If the pattern is an illegal pattern, a detection flag “1” is output, and the frequency of occurrence of the illegal pattern for each brightness (threshold) is counted in the illegal pattern histogram generation unit 13 to generate an illegal pattern histogram. Is done. Similarly, the brightness histogram generation unit 14 counts the frequency of each brightness and generates a brightness histogram. When the generation of these histograms is completed, the next image quality histogram generation unit 15 calculates an image quality histogram using both histograms. This image quality histogram is smoothed by a histogram smoothing processing unit 16 as needed, and then an optimum threshold value searching unit 17 searches for an optimum threshold value from the smoothed image quality histogram, and the search result is micro-scaled. Output to computer 10.

In the above embodiment, the optimum threshold value is calculated by hardware, but the present invention is not limited to this, and the optimum threshold value can be calculated by software.

FIG. 12 shows a control procedure by a microcomputer (not shown) in this case. This microcomputer has a CPU as a control main body, and the CPU decodes and executes a program stored in a ROM, and executes a program stored in a RAM. Of course, the procedure shown in FIG. 12 is sequentially executed while reading and writing data from and to.

In step 1 (indicated by “ST1” in the figure) and step 2 in FIG. 12, the CPU clears the brightness histogram and the illegal pattern histogram generated in the RAM or the like to zero, and then sets the window in the next step 3 Set to the initial position of the grayscale image. Next, the CPU accesses each pixel in the window and binarizes the image in the window using the brightness of a specific pixel in the window as a threshold (steps 4 and 5).

Next, the CPU reads the frequency of the brightness histogram corresponding to the brightness of the threshold from the RAM, adds 1 to the frequency, and rewrites the contents of the RAM with the addition result (step 6).

The next step 7 is to determine whether the binary pattern in the window is a legal pattern or an illegal pattern.
Then, the frequency of the illegal pattern histogram corresponding to the brightness of the threshold value is read and added by 1, and the contents of the RAM are rewritten with the addition result (step 8).

In the next step 9, it is determined whether or not scanning of the grayscale image by the window has been completed.
In the case of "O", the CPU changes the position of the window in a raster scan manner in step 10 and then returns to step 4 to repeatedly execute the same procedure.

When the scanning by the window is completed, "YES" is determined in step 9 and the CPU calculates an image quality histogram from the brightness histogram and the illegal pattern histogram in step 11, and then, in step 12, an optimum threshold value based on the image quality histogram. Is calculated.

<Effects of the Invention> As described above, the present invention sets and scans a window of a predetermined shape on a multi-valued image, and scans each multi-valued image based on the brightness level extracted for the image in the window at each scanning position. Information indicating the number of windows corresponding to the brightness level is generated, and for each of the extracted brightness levels, a binary pattern obtained when this level is set as a threshold matches a legal pattern or an illegal pattern. Since the information relating to the degree of the image quality is generated, and the third information for giving the evaluation value of the image quality is generated using the information, even if the shape and size of the object are unknown,
The image quality of a multi-valued image is evaluated by one scan using a window, and illumination and other imaging conditions can be optimally controlled based on the evaluation value, and a multi-valued image suitable for the evaluation object can be obtained. . Further, in the second invention, the optimum threshold value that gives the best image quality is determined based on the third information, so that the threshold value can be determined even if the size of the object is unknown. is there. Furthermore, a single scan with a window enables evaluation of a multi-valued image and determination of a threshold value, and has a remarkable effect of attaining the object of the invention, such as being able to quickly and stably determine an optimum threshold value.

[Brief description of the drawings]

FIG. 1 is a circuit block diagram of a threshold value determination device using the multi-valued image evaluation device according to the present invention, FIG. 2 is a circuit block diagram showing a specific example of an illegal pattern detection unit, and FIG. FIG. 4 is a circuit block diagram illustrating a specific example of a legal pattern histogram generation unit, FIG. 4 is a circuit block diagram illustrating a specific example of a brightness histogram generation unit, and FIG. 5 is an explanatory diagram illustrating a concept of a window set on a grayscale image. , FIG. 6 is an explanatory diagram showing another embodiment of the window, FIG. 7 is an explanatory diagram showing a binarization process of an image in the window, FIG. 8 is an explanatory diagram showing an example of a legal pattern, FIG. Is an explanatory diagram showing an example of an illegal pattern, FIG. 10 is an explanatory diagram showing a legal pattern assumed in this embodiment, FIG. 11 is an explanatory diagram showing a smoothed image quality histogram, and FIG. 12 is a threshold diagram. The procedure for calculating the value by software To the flowchart, FIG. 13, second 16
FIG. 1 is an explanatory diagram showing the principle of a conventional binarization threshold value calculation method. 11 window scanning unit 12 illegal pattern detection unit 13 illegal pattern histogram generation unit 14 brightness histogram generation unit 15 image quality histogram generation unit 17 optimal threshold value search unit

Claims (7)

(57) [Claims] 1. A multi-valued image evaluation apparatus for evaluating the image quality of a multi-valued image, comprising: a window scanning means for setting and scanning a window of a predetermined shape on the multi-valued image; First information generation means for extracting a brightness level using predetermined conditions for an image, and generating first information indicating the number of windows corresponding to each brightness level in the multi-valued image based on the extraction result; For the image in the window at each scanning position, the binary pattern obtained using the brightness level extracted under the predetermined condition as a threshold value matches a legal pattern or an illegal pattern for determining the smoothness of the image. After the determination as to whether or not each of the brightness levels is a threshold value, the result of the determination is associated with the threshold value. When the second information generation means over down to generate a second information indicating a degree of matching the legal pattern or illegal pattern, the scanning by the window scanning means has been completed, the
1. A multi-valued image evaluation device comprising: a third information generation unit that generates third information that gives an image quality evaluation value using information generated by each of the second information generation units. 2. A multi-value image evaluation apparatus according to claim 1, wherein said window scanning means is means for setting and scanning a rectangular window comprising a plurality of pixels in a vertical and horizontal direction on a multi-value image. . 3. The first information generating means extracts a brightness of a pixel at a specific position in a window for each scanning position, and generates a histogram of brightness in a multi-valued image based on each extraction result. The multi-value image evaluation device according to claim 1. 4. The first information generating means extracts an average value of brightness of a plurality of pixels in a window for each scanning position, and generates a brightness histogram of a multi-valued image based on each extraction result. 2. The multivalued image evaluation device according to claim 1, wherein the multivalued image evaluation device is a means for performing the operation. 5. The information generating means according to claim 1, wherein said third information generating means uses the information generated by said first and second information generating means for said multi-valued image to calculate each level of brightness and an evaluation value of image quality. 2. The multi-valued image evaluation device according to claim 1, wherein said multi-valued image evaluation device is means for generating an image quality histogram in which the image quality histogram is associated. 6. A threshold value determining device for determining a threshold value for processing a multi-valued image, comprising: a window scanning means for setting and scanning a window of a predetermined shape on the multi-valued image; First information for extracting a brightness level using a predetermined condition for an image in a window, and generating first information indicating the number of windows corresponding to each brightness level in the multi-valued image based on the extraction result. Generating means, for the image in the window at each scanning position, a binary pattern obtained by using the brightness level extracted under the predetermined condition as a threshold value to determine the smoothness of the image; After determining whether or not the pattern matches the pattern, the result of the determination is associated with the threshold value to obtain the level as a threshold value for each brightness level. Binary pattern and a second information generating means for generating a second information indicating a degree of matching the legal pattern or illegal pattern, when scanned by the window scanning means is completed, the
A third information generating means for generating third information for giving an image quality evaluation value by using information generated by each of the second information generating means; and providing the best image quality based on the third information. A threshold value determining device comprising: an optimum threshold value determining means for determining an optimum threshold value. 7. The third information generating means, for the multi-valued image, uses information generated by the first and second information generating means to calculate each level of brightness and an evaluation value of image quality. The optimum threshold value determining means determines a threshold value corresponding to an evaluation value indicating that the image quality is the best from the image quality histogram as an optimum threshold value. 7. The threshold value determination device according to claim 6, wherein the threshold value determination device is means for performing the threshold value determination.