JPH01219967A - Image processing method - Google Patents

Abstract

PURPOSE:To decrease the gradation without causing a sense of incongruity by dividing the gradation at every prescribed cumulative frequency, based on the number of divisions of the gradation from the gradation histogram of input image data and outputting the representative value of the divided gradation areas as output image data. CONSTITUTION:An original image is inputted from an image input part 1, and the histogram of an input image is generated by a histogram operating part 2. By a dividing part 3, the histogram is divided at every prescribed frequency, and from within the range of values of each gradation of this divided histograms, a representative value is calculated by an output gradation operating part 5. By how much output gradation an image is outputted against an input gradation is designated by an output gradation designating part 4. The processed image is outputted from an image output part 7.

Description

DETAILED DESCRIPTION OF THE INVENTION [Field of Industrial Application] The present invention relates to an image processing method, and particularly to an image processing method for reducing the @ tone of an original image and converting it into an image with a special effect.

[Prior Art] Regarding digital image processing, various devices and methods have been developed in various fields, and natural images such as photographs can be input as digital image data and processed to create creative images. It is possible to create. One of these processes is gradation conversion.

Conventionally, this tone conversion has a method (generally called bossarization) in which the tone of the original image is divided into a certain ratio and all data in each divided tone is replaced with one representative value.

In addition, there is a known method for normalizing output gradations in which, in response to the frequency of input gradations, the gradation areas with high input frequency are increased and the gradation areas with low input frequency are reduced in output gradation. It is known that it is possible to express rich gradations.

[Problems to be Solved by the Invention] However, since the conventional bossarization method described above divides the gradations evenly, there are a large number of parts with input gradations.

In other words, there are many parts with similar colors or similar brightness. Therefore, there is a problem in that the area in which the gradation changes is small for the input image, and the special effect of reducing the gradation, which is the original purpose, is lost. That is, if the frequency of a predetermined gradation is high, the portion where the gradation changes will be small, and the effect of bossarization will not be seen.

In addition, when the heterogeneous frequency value corresponding to each input gradation is normalized using the heterogeneous frequency distribution, the output gradation for the part with high input frequency is divided into small parts, and conversely, the part with low frequency is divided into large parts. By doing so, the above-mentioned problem does not occur, but data whose input gradation is concentrated is spread out in the output gradation, which changes the original color and brightness of the input image. A problem arises in that the difference between the output image and the output image is too large.

The present invention provides an image processing method capable of outputting a natural bossarized image for any original image of any scene, which has not been possible with conventional techniques.

The present invention also provides an image processing method that can output a bossarized image that preserves the characteristics of the original image.

Furthermore, the present invention provides an image processing method capable of outputting a bossarized image that does not depend on the original image.

[Means and effects for solving the problem] In order to solve this problem, the image processing method of the present invention is an image processing method for reducing the gradation of image data. The gradation is divided at every predetermined cumulative frequency based on the number of divisions, and the representative value of the divided gradation area is output as output image data.

Further, in an image processing method for reducing the gradation of image data, continuous gradation areas whose frequency is less than or equal to a predetermined value are divided as one gradation from the gradation histogram of input image data, and the divided gradation areas are The representative value of is output as output image data.

In addition, in an image processing method for reducing the gradation of image data, from the gradation histogram of input image data, continuous gradation areas whose frequency is more than a predetermined value are divided as one gradation, and the divided gradation areas are Among them, tone areas whose cumulative frequency is less than a predetermined value are ignored, and representative values of the divided tone areas are output as output image data.

[Example] First, in order to clarify the characteristics of this example, bossarization according to a conventional example will be described in detail.

FIGS. 6 and 7 illustrate an example of a conventional bossarization method. Input gradation is shown on the horizontal axis of the figure.

The vertical axis shows the output gradation. FIG. 6 shows the normal case where the input image is output as is, and FIG. 7 shows the case where the input is 32 gradations (0 to 31) and the output is 4 gradations.

The output value is 09 input value (8 to 1) for the input value (0 to 7).
This means that the output value for 5) is replaced with a representative value such as 8. This is the method of - intramembrane bossarization.

For example, consider a case where an image is input with 256 gradations (0 to 255) and outputted with 8 gradations. FIG. 12 shows a histogram of the input image, in which the input gradation is equally divided into eight parts. The area ratio of each divided histogram is equal to the area ratio for each gradation of the output image. Therefore, if the input gradation values 160 to 191 are set to a typical value of 160, for example, most of the output image will have a gradation value of 160, and the areas where the gradation changes will be small, resulting in the effect of bossarization. can no longer be seen.

FIG. 8, FIG. 9, FIG. 10, and FIG. 11 show an example of a method for normalizing the output gradation in accordance with the frequency of the input gradation. In FIG. 8, the horizontal axis represents gradation and the vertical axis represents frequency, and shows the frequency of each gradation of certain data. FIG. 9 is a diagram showing the input and output gradations of the same data as in FIG. 8. From the histogram in FIG. 8, it can be seen that the data is concentrated between input gradation values a and C2. . This means that the data is concentrated between the output gradation values a and b2 in FIG. 9, and the gradation is not expressed richly. Therefore, by calculating the cross-product frequency as shown in FIG. 10 and directly replacing the cross-product frequency value corresponding to each input gradation value with the output gradation value as shown in FIG.
The input gradation part between C2 and C2 is converted into the output gradation part from C8 to C3.

In addition, using the heterogeneous frequency distribution, the heterogeneous frequency value corresponding to each input gradation is normalized to 0 to 255, the value is divided into 8 equal parts, and the minimum gradation within the divided area is used as the representative value of each divided area. FIG. 13 is a diagram showing the relationship between input gradation and output gradation when using values. The output gradation for parts with high input frequency is subdivided,
On the contrary, the low-frequency parts are divided into large parts, so there is no problem mentioned above, but the data that was concentrated around the input gradation range of 160 to 191 is now concentrated around the output gradation range of 64 to 223.
As a result, the inherent color and brightness of the input image change, resulting in a problem that the difference between the input image and the output image becomes too large.

<First example> A first embodiment of the present invention will be described in detail below.

FIG. 1 is a block diagram of an image processing apparatus of this embodiment that implements the image processing method of the present invention.

Further, FIG. 2 is a hardware configuration diagram of the image processing apparatus of this embodiment.

1 is an image input unit for inputting an original image, and in FIG. 2, an image input device 11 (for example, a color scanner,
V camera, etc.) and the A/D converter 12. 2 is a histogram calculation unit that creates a histogram of the input image. 3 is a division unit that divides the histogram into parts according to a certain cumulative frequency. 4 is the output gradation specification section, which specifies how much output gradation to output the image with respect to the input gradation;
This corresponds to character/numeric input method 16 in the figure. 5 calculates a representative value from within the range of values of each gradation of the divided histogram in an output gradation calculation section. Reference numeral 7 denotes an image output unit that outputs a processed image, which in FIG. 2 corresponds to the D/A converter 17 and an image output device (eg, color CRT, photo printer, etc.). 8, 9, and 10 mean image data signal lines, indicating that the input image data is treated as RGB luminance data.

Also, the histogram calculation unit 2. Division part 3. In the specific power key calculation section 5, the CPU 14 in FIG.
5b11:r4 The stored program is executed and the input image data is stored in the image memory 13. C
Inside the PU 14, there is a memory 15 consisting of a RAM 15a and a ROM 15b.The RAM 15a stores information input from a character/numeric input device 16 (e.g., keyboard, digitizer, etc.), and the ROM 15b stores information input from the CPU 14.
processing programs are stored.

Next, the contents of the process will be specifically explained.

RGB image data input from the image input device 11 to the image memory 13 through the A/D converter 12 is used to create RGB histograms by the histogram calculation unit 2. Examples of created histograms are shown in FIGS. 3(a) to 3(C).

The horizontal axis shows the gradation, and the vertical axis shows the frequency.

Here, the desired number of gradations of the output image is input from the output gradation specifying section 4, and the dividing section 3 divides the histogram by (total intensity)/(number of gradations). By expressing high-frequency parts using many brightness (gradation) values, we aim to create an effect that smoothly resembles the color and brightness of the original image, so that there are no unpleasant color changes on the output image. It is something that FIGS. 4(a) to 4(C) show the state of equal division.

Here, the output gradation calculating section 5 calculates a representative value of the output gradation within the range of each divided input gradation value. In this embodiment, the average value is used as the representative value of the divided regions, but similar results can be obtained by using the minimum value, maximum value, etc. For example, in FIG. 4(a), the representative value in a certain divided area (shaded area in the figure) is (a+b)÷2. In the manner described above, an output image with a desired gradation can be obtained for the input image. FIGS. 5(a) to 5(C) illustrate the gradation relationship between the input image and the output image when the histogram of the input image is shown in FIG. 3 and the output image has 8 gradations.

<Second Example> Next, a second example will be shown. As in the embodiment shown in FIG. Then, divide the histogram by (total number of pixels) ÷ (specified gradation).

Here, a threshold value is obtained from a threshold value generation unit (not shown). In this embodiment, a preset value is written in the memory 15 in the CPU 14 and the value is called up. It is also possible to determine. This threshold value is used to extract a portion of the histogram that indicates a feature, and is expressed as a frequency value.

As shown in FIG. 14, the representative value of the output gradation is calculated by the output gradation calculation unit 5 as shown below, focusing on the portion of the histogram that is equal to or higher than the threshold value, within the range of each divided value. do.

Input gradation value Output gradation value 0~(d+ IL (d+ 1)/2d1~(d
2-1) (d1+d2-1)/2d5~(da
1) (ds +da-1)/2d6~255
Below (d, +255-1)/2, the representative value of the output gradation of each divided area is similarly determined. As can be seen from FIG. 14, the number of regions divided by the divider 3 is eight, but by using the threshold, the representative values of the output gradations are reduced to seven. Further, when a histogram as shown in FIG. 15 is calculated, the representative value is 10.

In this way, it is possible to automatically change the output gradation depending on the scene of the input image to perform more preferable bossarization processing. Although the average value between the input gradation values is used as the representative value, the same effect can be obtained by using the maximum value, minimum value, etc. as in the first embodiment.

<Third Example> Next, a third example will be shown. As in the first embodiment, an image is input from the image input section 1, a desired number of gradations is input from the output gradation specifying section 4, and a histogram is created by the histogram calculation section 2. Here, a threshold value is obtained from a threshold value generation unit (not shown). In this embodiment, a preset value is called from the memory as in the second embodiment, but it is also possible to determine the threshold value interactively via the character/numeric input device 16 by looking at the situation of the output image. It is. The threshold value is a frequency value as in the second embodiment.

FIG. 16 is a diagram illustrating this embodiment, in which the desired number of output gradations is eight. The horizontal axis shows the input gradation, and the vertical axis shows the frequency. Focusing on the part of the histogram that is above the threshold,
First, the number of consecutive values below the threshold is counted. 1st
In Figure 6, (φ~e l) + (64~ea) + (eo
There are four sections: ~es)+(er~255). Next, count the total number of frequencies greater than or equal to the threshold. The number of divisions for the portion exceeding the threshold value is the specified number of desired output gradations minus the count number of the portions where the values below the threshold value are continuous, and the number of divisions is four.

Next, from the value obtained by dividing the total head degree above the threshold by the number of divisions, each part (et~e4), (ea~8
9), among (e5 to ey), remove those whose sum of frequencies is smaller than a predetermined value. This is to ignore a faint portion such as the portion X surrounded by a dotted circle in FIG.

By repeating the above steps, the histogram will have several rough peaks where the areas above the threshold are divided into 5 peaks, which are calculated by subtracting the number of valleys (3) from the desired output gradation level (8). ), the result will be as shown in Figure 17. At this time, parts that should originally be one area may become separated, such as % (eI 0-84 ) and (es~e'a).

In this embodiment, the part with the highest total frequency of the divided parts was retained, and the part with the lowest frequency was ignored. Therefore, the divided input tone values are (φ~e1) and (e+~e2).

(e 2 ””' e 3 ) + (83~el O
), (el O~e!i) + (es~ea)
, (ea~ey), (et~255), and the output gradation calculation unit 5 calculates the representative value of the output gradation within the range of each input gradation value.
Calculated by In this way, it is possible to obtain a bossarized output image with a desired output gradation and independent of the scene of the input image.

In addition, although the input image and the output image are brightness data in the first, second, and third embodiments, density data may also be used.

As explained above, by using the histogram to change the number of gradations in the original image, it is possible to obtain the desired gradation image without being affected by the scene of the original image, and it is possible to obtain a desired gradation image without being affected by the scene of the original image. It is possible to obtain a tone-converted image without any gradation conversion.

[Effects of the Invention] According to the present invention, it is possible to provide an image processing method that can output a normalized bossarized image for any original image of any scene.

Furthermore, it is possible to provide an image processing method that can output a bossarized image that preserves the characteristics of the original image.

Furthermore, it is possible to provide an image processing method that can output a bossarized image that does not depend on the original image.

[Brief explanation of the drawing]

FIG. 1 is a block diagram of an image processing apparatus according to this embodiment that implements the image processing method of the present invention, FIG. 2 is a hardware configuration diagram of the image processing apparatus according to this embodiment, and FIGS. c) is a diagram showing an example of the histogram of the input image, Figures 4(a) to (c) are diagrams in which the gradation is divided into eight parts depending on the frequency, and Figures 5(a) to (e) are diagrams showing the histogram of the input image. Figure 3 shows the gradation relationship of the human output when processing the data used, Figure 6 shows the gradation relationship of the normal image appearance, and Figure 7 shows the gradation in the conventional example. Diagram showing the conversion, Figure 8~
Fig. 11 is a diagram showing a conventional method for normalizing output gradations in accordance with the frequency of input gradations, and Figs. 14 and 15 are diagrams illustrating the second embodiment, and FIGS. 16 and 17 are diagrams illustrating the third embodiment. In the figure, 1... image input section, 2... histogram calculation section, 3... division section, 4... output gradation specification section, 5...
- Output gradation calculation unit, 7... Image output unit, 11... Image input device, 12... A/D converter, 13... Image memory, 14... CPU, 15... Memory , 15
a-RAM, 15b... ROM, 16... character/numeric input device, 17... D/A converter, 18... image output device. Figure 1 Figure 6 Figure 7 Figure 8 Figure 9 Figure 10 Figure 11 Figure 12 Figure 14 Figure 15 Figure 16

Claims (3)

[Claims] (1) In an image processing method for reducing the gradation of image data, from the gradation histogram of input image data, the gradation is divided at each predetermined cumulative frequency based on the number of divisions of the gradation, and the divided gradation is An image processing method characterized by outputting a representative value of a region as output image data. (2) In an image processing method for reducing the gradation of image data, from the gradation histogram of input image data, continuous gradation areas whose frequency is less than a predetermined value are divided as one gradation, and the divided gradation is An image processing method characterized by outputting a representative value of a region as output image data. (3) In an image processing method for reducing the gradation of image data, from the gradation histogram of the input image data, continuous gradation areas whose frequency is equal to or higher than a predetermined value are divided into one gradation, and the divided gradation is An image processing method characterized in that, among the regions, tone regions whose cumulative frequency is less than or equal to a predetermined value are ignored, and representative values of the divided tone regions are output as output image data.