JPH07306938A - Image processing method and device therefor - Google Patents

Abstract

PURPOSE:To provide an image processing method/device which can perform the band compression for the dynamic range of image data without losing the useful information nor generating the distortions. CONSTITUTION:An image processor is provided with the differentiating means 2 and 3 which calculate the differential value of image data, a histogram generator means 4 which generates the histogram of the emerging frequency of each signal value for the area where the differential value calculated by the means 2 and 3 are included in a prescribed area, a conversion table generator means 5 which generates a conversion table where the compression rate is increased for the signal value having high emerging frequency of the histogram generated by the means 4, and a band compressing means 6 which performs the band compression of image data based on the conversion table generated by the means 5.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an image processing method and an image processing apparatus, and more particularly to an image processing method and an image processing apparatus capable of band-compressing a dynamic range of image data without losing useful information.

[0002]

2. Description of the Related Art The dynamic range of image data obtained from various medical devices is very wide, and sometimes wider than the dynamic range of a display device such as a CRT display or an image output device such as a film output device.

In such a case, it is extremely difficult to faithfully reproduce image data having a wide dynamic range by a device having a narrow dynamic range. Further, even if the image is reproduced faithfully, the image is displayed in a strong contrast to the naked eye, and it may be difficult to qualitatively recognize the target.

Therefore, in order to perform display or recording for proper recognition, some kind of band compression (dynamic range compression) must be performed in consideration of the characteristics of the cathode ray tube and the photosensitive material.

As a general method for such band compression, a histogram equalization (Histgram Equ
alization) and frequency enhancement processing for increasing or decreasing a specific frequency region.

In histogram equalization,
First, a histogram of signal intensity and appearance frequency shown in FIG. 7A is generated from image data. From this histogram, it can be seen that there is a region of signal strength with a low frequency of appearance. Therefore, a lookup having an input / output characteristic that compresses a low appearance frequency portion (in FIG. 7A) in this histogram and does not compress a portion other than the low appearance frequency portion (FIG. 7,). A table (LUT) is generated (FIG. 7 (b)). Then, by converting the signal value of each pixel of the input image data by this LUT, the image data having the compressed dynamic range is obtained.

By performing such band compression, it is apparent that the original image data of FIG. 7C and the converted image data of FIG. Compressed.

In the frequency emphasis process, high frequency components are emphasized and low frequency components are reduced. Here, a case will be described where band compression is performed by subtracting the low-frequency component of image data from the original image. Here, FIG. 8 (a)
For the original image data of the waveform shown in, smooth (low pass)
By processing, the low frequency component signal shown in FIG. 8B is obtained. At the same time, the high-pass processing is performed to obtain the high-frequency component signal shown in FIG. Then, by adding the high frequency component signal and the low frequency component signal whose amplitude is compressed, a signal having a waveform as shown in FIG. 8D is obtained. By doing so, a signal in which the low frequency components are reduced and the dynamic range is compressed can be obtained. Further, the same effect can be obtained by subtracting only the low frequency component (FIG. 8B) from the original image data (FIG. 8A).

[0009]

According to the above-mentioned histogram equalization, the dynamic range is compressed in the region of low appearance frequency, but the low appearance frequency does not necessarily mean that the data is not so important. Therefore, important data may be compressed in some cases. Further, as shown in FIG. 7 (1), when the low appearance frequency parts are present together, the effect of band compression is conspicuous, but conversely, the effect of band compression is effective for image data with few low appearance frequency parts. It will not appear.

For example, consider the case shown in FIG.
FIG. 9A is an example of a CT image of a cross section of the brain. The histogram in this case is as shown in FIG.
T value = -1000), brain parenchyma (CT value = 0), bone (CT)
Values = + 1000) are distributed almost evenly. And
Since there is no CT value portion between the air, the brain parenchyma, and the bone, the appearance frequency is 0. Therefore, it is possible to compress such a portion having a low appearance frequency.

However, in the image of the cross section as shown in FIG. 9 (c), most of the bone is present and the brain parenchyma is a small area region. Therefore, the histogram is as shown in FIG. 9D, and although the frequency of air and bone is high, the histogram of the brain parenchyma is extremely small. In such a case, according to the histogram equalization, the part of the brain parenchyma having a low appearance frequency is compressed, and it becomes difficult to recognize a subtle change. As described above, although it is an important part, it may be compressed due to its low appearance frequency.

On the other hand, in the waveform (FIG. 8 (d)) obtained by band compression by reducing the low frequency components shown in FIG.
As is clear from comparison with the ideal waveform (FIG. 8E), the image is distorted at the edge portion. In such distortion, the extracted low frequency component (solid line in FIG. 8B) is
Waveform different from ideal faithful low frequency component (broken line in FIG. 8B) and extraction result of high frequency component (FIG. 8C)
The main cause is the distortion. Due to such distortion, there is a problem that it becomes difficult to observe because the distortion occurs in the vicinity of the portion where the edge exists in the image.

The present invention has been made in view of the above points, and an object thereof is an image processing method capable of band-compressing a dynamic range of image data without losing useful information or generating distortion. It is to realize an image processing device.

[0014]

[Means for Solving the Problems] A first means for solving the above-mentioned problems is to obtain a differential value for image data, and to determine the frequency of appearance of each signal value in a region where the obtained differential value is within a predetermined range. Is generated, a conversion table with a high compression rate is generated for signal values having a high appearance frequency in the generated histogram, and band compression of image data is executed based on the generated conversion table. This is an image processing method.

A second means for solving the above problem is to obtain a differential value for image data, and generate a histogram for the frequency of appearance of each signal value in an area where the obtained differential value is within a predetermined range, Generate a normalized histogram by normalizing the generated histogram with a predetermined value, generate an inverse histogram by subtracting the normalized histogram from a predetermined value, generate a conversion table by integrating the inverse histogram, An image processing method is characterized in that band compression of image data is executed based on the generated conversion table.

A third means for solving the above-mentioned problems is a differentiating means for obtaining a differential value of image data, and an appearance frequency of each signal value in an area in which the differential value obtained by the differentiating means is within a predetermined range. With a conversion table generating means for generating a conversion table with a high compression rate for signal values with a high appearance frequency of the histogram generated by the histogram generating means, and with the conversion table generating means. An image processing apparatus comprising: a band compression unit that performs band compression of image data based on the converted table.

A fourth means for solving the above-mentioned problems is a differentiating means for obtaining a differential value of image data, and an appearance frequency of each signal value in an area in which the differential value obtained by the differentiating means is within a predetermined range. And a histogram generating means for generating a normalization histogram by normalizing the generated histogram with a predetermined value, and further generating an inverse histogram by subtracting the normalized histogram from the predetermined value. Conversion table generation means for integrating the histogram to generate a conversion table,
An image processing apparatus comprising: a band compression unit that performs band compression of image data based on the conversion table generated by the conversion table generation unit.

[0018]

In the first means for solving the problem, a histogram of the frequency of appearance of each signal value is generated for an area where the differential value of the image data is within a predetermined range,
A conversion table in which the compression rate increases for signal values having a high appearance frequency of the histogram is generated, and band compression of image data is executed based on this conversion table.

In the second means for solving the problem, a histogram of the appearance frequency of each signal value is generated for an area where the differential value of the image data is within a predetermined range, and the histogram is normalized with the predetermined value. A normalized histogram is generated, an inverse histogram obtained by subtracting the normalized histogram from a predetermined value is generated, and the inverse histogram is integrated to generate a conversion table. Then, the band compression of the image data is executed based on this conversion table.

In a third means for solving the problem, a histogram of the frequency of appearance of each signal value is generated for a region where the differential value of image data is within a predetermined range, and a signal value having a high frequency of appearance of this histogram is generated. A conversion table with a high compression rate is generated, and band compression of image data is executed based on this conversion table.

In a fourth means for solving the problem, a histogram of the appearance frequency of each signal value is generated for an area where the differential value of the image data is within a predetermined range, and is normalized with the predetermined value from this histogram. A normalized histogram is generated, an inverse histogram obtained by subtracting the normalized histogram from a predetermined value is generated, and the inverse histogram is integrated to generate a conversion table. Then, the band compression of the image data is executed based on this conversion table.

[0022]

Embodiments of the present invention will now be described in detail with reference to the drawings. FIG. 1 is a configuration diagram showing an overall configuration when a radiation image is processed as an example of an image processing apparatus according to an embodiment of the present invention. 2 is a flow chart showing the processing procedure of the image processing method according to the embodiment of the present invention, and FIG. 3 is an explanatory diagram showing the states of data, histograms and various waveforms obtained in the process of the image processing method.

In FIG. 1, an image memory 1 stores image data of radiation images collected by various image collecting methods. The differentiating means 2 is the image memory 1
The image data from is differentiated by various differentiating methods to obtain a differential value. The absoluteizing means 3 outputs the absolute value of the differential value obtained by the differentiating means 2. The histogram generating means 4 generates a histogram of the image data in the image memory 1 by referring to the differential value which has been absolute valued by the absolute value converting means 3. The LUT generating means 5 uses a lookup table (LU as a conversion table for band compression from the histogram generated by the histogram generating means).
T) is generated. The band compression means 6 uses the LUT generated by the LUT generation means 5 to generate the image memory 1
The image data within is band-compressed. The image memory 7 stores the image data band-compressed by the band compression means 6. Each of the above means is composed of software, hardware, and firmware such as a computer device or a processing processor in which various processing programs are installed.

The operation of such an image processing apparatus and the image processing method are as follows. Image data collected by an image collecting device not shown in FIG. 1 is stored in the image memory 1. FIG. 3A shows the image memory 1.
Shows an example of the image data inside, the horizontal axis is the pixel number,
The vertical axis represents the signal value.

Differentiating means 2 generates such a differential value of the image data (step 1 in FIG. 2). In this case, the edge portion (the portion where the density changes abruptly in the image) is detected as a high value by applying the differentiation on the image plane by the edge detection processing using the differential operator. This differentiation takes the difference between the data of the adjacent pixels on the digital image.

For example, a two-dimensional differential can be used as this differential, and the differential can be executed by a convolution calculation in which the X axis and the Y axis are weighted. Then, the absolute value conversion means 3 converts the differential values of the X axis and the Y axis into absolute values and outputs them. The differential value in the X-axis direction is P '
When the differential value in the x- and Y-axis directions is P'y, the absolute value P'of the differential value is obtained as P '= √ (P'x ² + P'y ² ).

As a method of obtaining the differential values of the X-axis and the Y-axis, for example, for each 5 pixels on one side in the X or Y direction adjacent to the target pixel (11 pixels including the target pixel), (- It is possible to perform convolution with weights such as 5 −4 −3 −2 −1 0 1 2 3 4 5) / 110. This convolution will be explained in more detail as follows.

For example, the X-axis direction differential value P'x is set to P'x
(X, Y), and the differential value P′y in the Y-axis direction is P′y (X,
If Y),

[0029]

[Equation 1]

It can be expressed as In such a convolution calculation, it is also possible to perform every other pixel, in addition to performing the operation for consecutive pixels. Further, in this way, by performing the two-dimensional differentiation using a plurality of pixels in the vicinity, the influence of fine noise is reduced.

By differentiating and converting the image data into absolute values, the differential result shown in FIG. 3B is obtained. In this differentiation result, the horizontal axis represents the pixel number and the vertical axis represents the differential value. That is, the absolute value of the differential value (hereinafter, also simply referred to as the differential value) is large where the change in the signal value of each pixel is large.

Next, the histogram generating means 4 refers to the differential value and generates a histogram H for the image data of the image memory 1 within a predetermined range (step 2 in FIG. 2). In generating the histogram H, it is conceivable to generate a histogram in an area where the differential value is larger than a predetermined value set in advance, or to generate a histogram in an area where the differential value is smaller than a predetermined value set in advance. Here, a case will be described in which the histogram H is generated in an area where the differential value is larger than the predetermined value. That is, the histogram generation means 4 generates the histogram H for the image data in the area where the differential value is larger than the predetermined value (FIG. 3 (c)). In the histogram H shown in FIG. 3C, the horizontal axis represents the signal value and the vertical axis represents the appearance frequency. Therefore, this histogram H
Is a histogram for a region where the differential value is large (change is large).

Next, the LUT generating means 5 normalizes the maximum value of the histogram H to 1.
Is executed to generate a normalized histogram H '. Further, an inverse histogram H ″ (= 1−H ′, FIG. 3D) obtained by subtracting the normalized histogram H ′ from 1 is obtained. Therefore, the LUT generating means 5 also serves as the inverse histogram generating means. The histogram H ″ has a characteristic such that a signal value having a large differential value and a high appearance frequency approaches 0, and other signal values approach 1. Then, the LUT generation means 5 uses the inverse histogram H ″ (see FIG.
(D)) is integrated to generate an LUT (step 2 in FIG. 2, FIG. 3 (e)).

Then, the LUT thus generated
The band compression means 6 executes the band compression processing (step 4 in FIG. 2). In the LUT shown in FIG. 3E, the horizontal axis represents the input signal value and the vertical axis represents the output signal value. Regarding the signal value region in which the value becomes small in the above-described inverse histogram H ″ (FIG. 3D) The characteristics are such that the inclination is small, that is, the compression rate is high.

FIGS. 4 and 5 are explanatory views showing the principle concept of band compression in the present invention. FIG. 4 shows the image memory 1
Shows an example of image data (horizontal axis is pixel number, vertical axis is signal value). In regions such as the regions A and B in FIG. 4 which are flat or nearly flat, even minute changes (small peaks and valleys) can be clearly recognized, and therefore it is desirable not to compress.

Here, the relationship between the degree of change in image density and the area of slight change will be described with reference to FIG. In FIG. 5, the horizontal axis represents the pixel number and the vertical axis represents the signal value. Then, FIG. 5A shows only a minute change, FIG. 5B shows a change in the image density of the low frequency component, and FIG. 5C shows a small region in the image density of the low frequency component. Are superimposed.

That is, in the case of the image data as shown in FIG. 5C in which FIG. 5A and FIG. 5B are superposed, the change in density is small and flat as shown in FIG. 5C. It can be seen that minute changes are easy to recognize in the region. However, when such a flat area is a relatively narrow area as shown in FIG. 5 (the appearance frequency of the same density value is small), the area is compressed by the conventional histogram method. . That is, the area where it should have been easy to recognize a minute change has been conventionally compressed so that it is difficult to recognize a minute change.

On the other hand, it is difficult to recognize a minute change in a region where the change is large as shown in FIGS. 4A, 4B, 4C and 4D. For example, in (c) of FIG. 5, since it is a region where the density change of the low frequency component is large, it is difficult to recognize a minute change. When such an area is a relatively wide range, it has been expanded by the conventional histogram method. Ie. Although it is a region where it is difficult to recognize minute changes, it has not been compressed in the past.

Therefore, in the present embodiment, a region in which such a minute change is difficult to recognize is detected, and a conversion table that increases the compression rate in such a region is generated to perform band compression. .

Therefore, as described with reference to FIG. 3, a differential value of the image data is generated, a histogram is generated for a partial area of the image data by referring to the differential value, and the inverse histogram H "of the histogram is calculated. The conversion table for band compression as described above is realized by generating the LUT.

FIG. 6 is an explanatory diagram showing the states of data, histograms and various waveforms obtained in the process of another embodiment of the image processing method of the present invention, and corresponds to the explanatory diagram of FIG. . FIG. 6A shows an example of image data in the image memory 1, where the horizontal axis represents the pixel number and the vertical axis represents the signal value. The differentiating means 2 generates such a differential value of the image data (step 1 in FIG. 2). Then, the absolute value conversion means 3 converts the differential value into an absolute value and outputs it. By differentiating and converting the image data into absolute values, the differential result shown in FIG. 6B is obtained. In this differentiation result, the horizontal axis represents the pixel number and the vertical axis represents the differential value.

Next, the histogram generating means 4 refers to the differential value and generates a histogram H for the image data of the image memory 1 within a predetermined range (step 2 in FIG. 2). Here, a case where the histogram H is generated in an area where the differential value is smaller than the predetermined value will be described. That is, the histogram generation means 4 generates the histogram H for the image data in the area where the differential value is smaller than the predetermined value (FIG. 6 (c)). In the histogram H shown in FIG. 6C, the horizontal axis represents the signal value and the vertical axis represents the appearance frequency. Therefore, this histogram H is a histogram for a region with a small differential value (small change).

Then, the LUT generating means 5 executes normalization in which the maximum value of the histogram H is set to 1, and generates a normalized histogram H '. Further, an inverse histogram H ″ obtained by subtracting this normalized histogram H ′ from 1
(= 1−H ′, FIG. 6 (d)) is obtained. And the LUT
The generating means 5 integrates the inverse histogram H ″ (FIG. 6 (d)) to generate an LUT (step 3 in FIG. 2, FIG. 6).
(E)). In the LUT shown in FIG. 6E, the input signal value is shown on the horizontal axis and the output signal value is shown on the vertical axis, and an LUT having characteristics almost similar to those of FIG. 3E is obtained. The band compression means 6 executes the band compression processing by the LUT thus generated (step 4 in FIG. 2).

In the above embodiment, the histogram H is generated by dividing the differential value into the predetermined value or more and the predetermined value or less (less than), but in addition to this, the differential value is the first predetermined value. The same effect can be obtained by generating the histogram H for the area in the range above the second predetermined value.

As described in detail above, the histogram H of the area where the image signal value is changing rapidly (or the change is large) is taken, and the signal value of the area where the change is large is referred by referring to this histogram H. On the other hand, by increasing the compression rate, high dynamic range compression can be realized according to each image, and the visibility of subtle changes is not impaired.

That is, the differential value of the image data is obtained, a histogram of the appearance frequency of each signal value is generated for the area where the obtained differential value is within a predetermined range, and the appearance frequency of the generated histogram H is high. Generate a conversion table that increases the compression rate for signal values,
According to the image processing method of performing the band compression of the image data based on the generated conversion table, the appearance frequency of the histogram regarding the appearance frequency of each signal value is determined for the area where the differential value of the image data is within the predetermined range. A conversion table in which the compression rate becomes large for a large signal value is generated, and band compression of image data is executed based on this conversion table without impairing the visibility of subtle changes.

Further, the differential value of the image data is calculated, the histogram H of the appearance frequency of each signal value is generated for the area where the calculated differential value is within the predetermined range, and the generated histogram H is generated, Generate a normalized histogram by normalizing the generated histogram with a predetermined value, generate an inverse histogram by subtracting the normalized histogram from a predetermined value, generate a conversion table by integrating the inverse histogram, According to the image processing method, which is characterized by executing the band compression of the image data based on the generated conversion table, regarding the appearance frequency of each signal value with respect to the area where the differential value of the image data is within the predetermined range, From the histogram, the normalized histogram is generated, which is normalized by the given value. Conversely histogram calculated is generated, the conversion table reverse histogram is integrated is generated, the band compression of the image data is performed without impairing the visibility of subtle changes on the basis of the conversion table.

Further, a differentiating means for obtaining a differential value of the image data, and a histogram generating means for generating a histogram H about the appearance frequency of each signal value in the area where the differential value obtained by the differentiating means is within a predetermined range. And a conversion table generating means for generating a conversion table having a high compression rate for a signal value having a high appearance frequency of the histogram H generated by the histogram generating means, and an image based on the conversion table generated by the conversion table generating means. According to the image processing device characterized by including the band compression means for performing the band compression of the data, the histogram of the appearance frequency of each signal value for the area where the differential value of the image data is within the predetermined range is displayed. Generates a conversion table that increases the compression rate for signal values that appear frequently Is, the band compression of the image data is performed without impairing the visibility of subtle changes on the basis of the conversion table.

Then, a differentiating means for obtaining a differential value of the image data, and a histogram generating means for generating a histogram H for the appearance frequency of each signal value in the region where the differential value obtained by the differentiating means is within a predetermined range. Then, the generated histogram is normalized with a predetermined value to generate a normalized histogram, and an inverse histogram obtained by subtracting the normalized histogram from the predetermined value is generated,
A conversion table generating unit that integrates the inverse histogram to generate a conversion table, and a band compression unit that executes band compression of the image data based on the conversion table generated by the conversion table generating unit are provided. According to the image processing apparatus, the normalization histogram normalized by a predetermined value is generated from the histogram of the appearance frequency of each signal value in the region where the differential value of the image data is within the predetermined range, and the normalization histogram is generated. An inverse histogram is generated by subtracting the histogram from a predetermined value, and the inverse histogram is integrated to generate a conversion table. Based on this conversion table, band compression of image data is executed without impairing the visibility of subtle changes.

[0050]

As described in detail above, according to the present invention,
Bandwidth compression processing is performed by referring to the histogram of the area where the signal value of the image is changing rapidly (or large change) and preparing a conversion table that increases the compression rate for the signal value of the area where the change is large. By executing it, high dynamic range compression can be realized according to each image, the dynamic range of the image data can be banded without loss of useful information or distortion without impairing the visibility of subtle changes. An image processing method and an image processing apparatus capable of being compressed can be realized.

[Brief description of drawings]

FIG. 1 is a configuration block diagram showing a configuration using an image processing apparatus according to an embodiment of the present invention.

FIG. 2 is a flowchart showing a processing example of an image processing method according to an embodiment of the present invention.

FIG. 3 is an explanatory diagram showing states of data, a histogram, and various waveforms obtained in the process of the image processing method according to the embodiment of the present invention.

FIG. 4 is an explanatory diagram showing a principle concept of band compression in the present invention.

FIG. 5 is an explanatory diagram showing a principle concept of band compression in the present invention.

FIG. 6 is an explanatory diagram showing the states of data, a histogram, and various waveforms obtained in the process of the image processing method according to the embodiment of the present invention.

FIG. 7 is an explanatory diagram showing a state of conventional band compression.

FIG. 8 is an explanatory diagram showing a state of conventional band compression.

FIG. 9 is an explanatory diagram showing a specific example of a state of conventional band compression.

[Explanation of symbols]

 DESCRIPTION OF SYMBOLS 1 image memory 2 differentiating means 3 absolute value converting means 4 histogram generating means 5 LUT generating means 6 band compressing means 7 image memory

Claims (4)

[Claims] 1. A differential value of image data is obtained, a histogram of the frequency of appearance of each signal value is generated for a region in which the obtained differential value is within a predetermined range, and a signal with a high frequency of appearance of the generated histogram is generated. An image processing method characterized by generating a conversion table having a higher compression ratio for a value and performing band compression of image data based on the generated conversion table. 2. A differential value for image data is calculated, a histogram is generated for the frequency of appearance of each signal value in a region where the calculated differential value is within a predetermined range, and the generated histogram is normalized with a predetermined value. To generate a normalized histogram, and then subtract this normalized histogram from a predetermined value to generate an inverse histogram, integrate the inverse histogram to generate a conversion table, and generate an image data based on the generated conversion table. An image processing method, characterized in that band compression is performed. 3. Differentiating means for obtaining a differential value of image data, and histogram generating means for producing a histogram of the frequency of appearance of each signal value in a region where the differential value obtained by the differentiating means is within a predetermined range. , A conversion table generating means for generating a conversion table having a high compression rate for signal values having a high appearance frequency of the histogram generated by the histogram generating means, and image data of the image data based on the conversion table generated by the conversion table generating means. An image processing apparatus comprising: a band compression unit that performs band compression. 4. Differentiating means for obtaining a differential value of image data, and histogram generating means for producing a histogram of the frequency of appearance of each signal value in a region where the differential value obtained by the differentiating means is within a predetermined range. , The generated histogram is normalized with a predetermined value to generate a normalized histogram, and the normalized histogram is subtracted from the predetermined value to generate an inverse histogram, and the inverse histogram is integrated to generate a conversion table. An image processing apparatus comprising: a conversion table generating means; and a band compression means for performing a band compression of image data based on the conversion table generated by the conversion table generating means.