JP2004173078A - Image processor - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide an image processor for performing stable multi-frequency processing which is not affected by an influence of the enlargement or reduction of image data. <P>SOLUTION: Since coefficients for individual frequency bands at the time of decomposing image data into a plurality of frequency components are converted or restored on the basis of compression or enlargement rates of the image data, the frequency processing of the image data is performed in an image processing circuit 112 in accordance with the compression or enlargement rates. Therefore, variances of pixel values of a subject image after compression/enlargement processing, density values after gray-scale transformation, etc. are reduced, and a stable image free from image variations and variations by subjects is generated. <P>COPYRIGHT: (C)2004,JPO

Description

[0001]
TECHNICAL FIELD OF THE INVENTION
The present invention relates to an image processing apparatus that performs frequency processing for each of a plurality of frequency bands, and more particularly, to an image processing apparatus that performs frequency processing based on an enlargement / reduction ratio of a subject.
[0002]
[Prior art]
Due to recent advances in digital technology, a radiation image is converted into a digital image signal, and the digital image signal is subjected to image processing such as frequency processing and noise elimination, and is displayed on a CRT or the like or output as a film to a printer. That is being done.
[0003]
By the way, in such frequency processing, image components of a plurality of frequency bands are separated and the image components of each frequency band are increased or decreased. Further, when the image is displayed on a CRT or the like, a process of enlarging or reducing the width of the image to match the range width of the CRT may be performed. In this case, the frequency processing and the like are performed irrespective of the reduction / enlargement ratio of the image (for example, Patent Document 1).
[0004]
[Patent Document 1]
JP 2000-155838 A
[Problems to be solved by the invention]
However, when an image subjected to frequency processing is enlarged or reduced, there is a problem that the frequency processing effect and the noise removal processing effect after the enlargement / reduction appear to be different.
SUMMARY An advantage of some aspects of the invention is to provide an image processing apparatus that makes an image processing effect effective on the basis of an image scaling ratio.
[0006]
[Means for Solving the Problems]
In the image processing apparatus of the present invention, the frequency component decomposing means decomposes the image data into a plurality of frequency components, and the coefficient converting means compresses the coefficients for each frequency band of the image data decomposed by the frequency component decomposing means. , Based on the enlargement ratio, and the restoration means inversely transforms the coefficients converted by the coefficient conversion means.
[0007]
Further, the analysis means calculates the maximum value and the minimum value of the subject, thereby calculating the compression and enlargement ratio of the image data, the frequency component decomposition means decomposes the image data into a plurality of frequency components, and the coefficient conversion means The coefficients for each frequency band decomposed by the frequency component decomposing means are converted based on the compression and expansion ratios calculated by the analysis means, and the restoration means reversely converts the coefficients converted by the coefficient conversion means.
[0008]
Further, the coefficient conversion means performs conversion so that the magnification of the coefficient decreases as the enlargement ratio of the image data increases.
[0009]
Further, the coefficient conversion means converts the coefficient so that the frequency response differs as the compression and enlargement ratio of the image data is changed.
[0010]
Further, the coefficient conversion means changes the cutoff width as the compression and enlargement ratios of the image data are changed to perform coefficient conversion.
[0011]
BEST MODE FOR CARRYING OUT THE INVENTION
(First Embodiment)
FIG. 1 shows an X-ray imaging apparatus 100 according to a first embodiment of the present invention. That is, the X-ray imaging apparatus 100 is an X-ray imaging apparatus having a function of performing an effective image processing effect on a film or a CRT of a captured image, and includes a preprocessing circuit 106, a CPU 108, a main memory 109, It includes a panel 110, an image display 111, and an image processing circuit 112, and data is exchanged with each other via the CPU bus 107.
[0012]
Further, the X-ray imaging apparatus 100 includes a data collection circuit 105 connected to the pre-processing circuit 106, and a two-dimensional X-ray sensor 104 and an X-ray generation circuit 101 connected to the data collection circuit 105. Are also connected to the CPU bus 107.
[0013]
FIG. 2 is a flowchart showing a flow of processing of the X-ray imaging apparatus 100 according to Embodiment 1 of the present invention. FIG. 3 is an image histogram in the irradiation field extracted by a commonly used irradiation field recognition circuit (not shown), in which the horizontal axis indicates the pixel value and the vertical axis indicates the frequency of appearance of the pixel.
[0014]
Here, in the drawing, reference numeral 301 denotes a peak of a pixel in a pass-through area, reference numeral 302 denotes a peak of the distribution of a subject, and reference numeral 303 denotes a point indicating the lowest frequency between the peaks 302 and 302. Is calculated as
[0015]
FIG. 4 is a diagram showing the coefficient magnification with respect to the image enlargement / reduction ratio. The horizontal axis represents the compression and enlargement ratio, and the vertical axis represents the coefficient conversion pattern representing the coefficient magnification.
In the figure, reference numeral 401 denotes an example of a coefficient conversion pattern. For example, if it is 1.0, nothing is changed, and if it is 0.5, the high frequency coefficient is multiplied by 0.5.
[0016]
FIG. 5 is a diagram showing a frequency balance of the frequency processing, and shows a change rate of the frequency component after the frequency processing. The horizontal axis indicates frequency, and the vertical axis indicates frequency balance for each frequency band. For example, 501 is an example of the frequency balance when the width of the image is doubled, and 502 is an example of the frequency balance when the width of the image is not changed. The larger the value of the frequency balance, the greater the frequency processing effect.
[0017]
FIG. 6 is an example of a coefficient conversion curve having a noise removing effect. Reference numeral 601 denotes a cutoff width. The larger the cutoff width, the greater the noise removing effect. The vertical axis indicates the input coefficient, and the vertical axis indicates the output coefficient. FIG. 7 is a diagram showing the relationship between the block width 601 shown in FIG. 6 and the image compression / enlargement ratio, with the horizontal axis representing the compression ratio and the vertical axis representing the block width.
[0018]
FIG. 8 shows an example of a gradation conversion curve 801. The horizontal axis indicating the relationship between the pixel value and the density value after the gradation conversion is the pixel value, and the vertical axis is the density value.
FIG. 9A is a diagram showing a configuration of the frequency component decomposition circuit 114, and FIG. 9B is a diagram showing a configuration example of a two-level conversion coefficient group obtained by a two-dimensional conversion process, and FIG. FIG. 3 is a diagram showing a configuration of a restoration circuit 117.
[0019]
In the X-ray imaging apparatus 100 as described above, first, the main memory 109 stores various data necessary for processing by the CPU 108 and includes a work memory for the CPU 108 to work.
[0020]
The CPU 108 uses the main memory 109 to control the operation of the entire apparatus in accordance with the operation from the operation panel 110. As a result, the X-ray imaging apparatus 100 operates as follows.
[0021]
X First, the X-ray generation circuit 101 emits an X-ray beam 102 to the inspection object 103. The X-ray beam 102 emitted from the X-ray generation circuit 101 passes through the subject 103 while attenuating, reaches the two-dimensional X-ray sensor 104, and is output as an X-ray image by the two-dimensional X-ray sensor 104. You. Here, the X-ray image output from the two-dimensional X-ray sensor 104 is, for example, a human body image.
[0022]
The data collection circuit 105 converts the X-ray image output from the two-dimensional X-ray sensor 104 into an electric signal and supplies the electric signal to the preprocessing circuit 106. The preprocessing circuit 106 performs preprocessing such as offset correction processing and gain correction processing on the signal (X-ray image signal) from the data collection circuit 105.
[0023]
The X-ray image signal preprocessed by the preprocessing circuit 106 is transferred as an original image to the main memory 109 and the image processing circuit 112 via the CPU bus 107 under the control of the CPU 108.
[0024]
Reference numeral 112 is a block diagram showing the configuration of the image processing circuit. In FIG. 112, reference numeral 113 denotes an analysis circuit, which is an analysis circuit that calculates a maximum value and a minimum value of a subject from a histogram of an image or the like.
[0025]
Reference numeral 114 denotes a first gradation conversion circuit 114 for enlarging an image in accordance with the enlargement ratio calculated by the analysis circuit 113. Reference numeral 115 denotes a discrete wavelet transform (hereinafter, referred to as an original image) enlarged by the first gradation conversion circuit 114. , DWT transform) to obtain a high-frequency coefficient (wavelet transform coefficient) of each frequency band.
[0026]
116 converts the high frequency coefficient obtained by the frequency component decomposition circuit 115 by the coefficient magnification shown in FIG. 4 so that the frequency characteristic shown in FIG. 5 can be obtained. For example, as shown in FIG. This is a coefficient conversion circuit that performs a coefficient conversion with a coefficient conversion curve having a large cutoff region to drop a coefficient corresponding to a noise component.
In this case, it is desirable to convert the coefficients using the coefficient conversion curve having the cutoff region shown in FIG. 6 first, and then perform the coefficient conversion to obtain the frequency balance shown in FIG.
[0027]
A restoration circuit 117 performs an inverse discrete wavelet transform (hereinafter, inverse DWT) based on the high-frequency coefficients converted by the coefficient conversion circuit 116.
Reference numeral 118 denotes a gradation conversion circuit that performs gradation conversion of the image restored by the restoration circuit using, for example, a gradation conversion curve as shown in FIG.
[0028]
The multi-frequency processing is not limited to the DWT, and a Laplacian pyramid may be used, or a filter that does not subband may be used.
[0029]
The first embodiment will be described below in accordance with the processing flow of FIG.
The original image preprocessed by the preprocessing circuit 106 is transferred to the image processing device 112 via the CPU bus 107. The image processing device 112 creates a histogram of the image of the irradiation field area recognized by the irradiation field recognition circuit (not shown) (step S201).
[0030]
First, the analysis circuit 113 extracts the number of peaks in the histogram. If the number of peaks is only one, it is determined that there is no passing through, and the maximum value of the image in the irradiation field is set to the maximum value of the subject. I do.
On the other hand, when there are two or more peaks, for example, the pixel value point 303 indicating the minimum value between the first peak 301 and the second peak 302 from the maximum value side of the histogram in FIG. (Step S202).
[0031]
The minimum value of the subject is set to the minimum value of the image in the irradiation field area (step S203). The example shown here is an example of a simple method of histogram analysis, but the minimum and maximum values of the subject are not limited to these methods, and may be calculated using any method.
[0032]
Next, an enlargement ratio is calculated from the pixel value width determined in advance by the imaging part information and the difference between the maximum value and the minimum value of the subject calculated above. For example, when the difference between the maximum value and the minimum value of the subject calculated above is 1000 and the predetermined pixel value width is 2000, the enlargement ratio is set to 2. Conversely, when the predetermined pixel value width is 2000 and the subject width is 4000, the enlargement ratio is 0.5 (step S204).
[0033]
Then, according to the enlargement ratio calculated by the analysis function 113, the first gradation conversion circuit enlarges the original image. In this case, for example, the subject width is increased centering on the intermediate value between the maximum and minimum pixel values of the subject.
[0034]
Next, the frequency component decomposition circuit 115 performs a two-dimensional discrete wavelet transform process on the original image for f (x, y), and calculates and outputs a high-frequency coefficient for each frequency band. The input image signal is separated into an even address signal and an odd address signal by a combination of a delay element and a down sampler, and is subjected to filter processing by two filters p and u.
[0035]
S and d shown in FIG. 9A represent low-pass coefficients and high-pass coefficients when one-level decomposition is performed on a one-dimensional image signal, and are calculated by the following equations.
d (n) = x (2 * n + 1) -floor ((x (2 * n) + x (2 * n + 2)) / 2) (1)
s (n) = x (2 * n) + floor ((d (n-1) + d (n)) / 4) (2)
Here, x (n) is an image signal to be converted.
[0036]
With the above processing, one-dimensional discrete wavelet transform processing is performed on the image signal. In the two-dimensional discrete wavelet transform, one-dimensional transform is sequentially performed in the horizontal and vertical directions of an image, and details thereof are known, and thus description thereof is omitted here.
[0037]
FIG. 9B is a configuration example of a two-level conversion coefficient group obtained by a two-dimensional conversion process, and the image signal includes high-frequency coefficients HH1, HL1, LH1,. . . , LL (step S206). In FIG. 9B, HH1, HL1, LH1,. . . , LL, etc. (hereinafter referred to as sub-bands) indicate high-frequency coefficients for each frequency band.
The frequency processing effect is obtained by increasing or decreasing this coefficient for each sub-band.
[0038]
First, the coefficient conversion circuit 116 changes the values of all the coefficients except for the LL component according to the magnification determined from the relationship between the enlargement ratio and the coefficient magnification in FIG. Performing such processing is not necessary to apply strong frequency processing when enlarging an image because contrast is increased. Conversely, when compressing an image, a strong frequency processing effect is required. It is because it becomes.
[0039]
Therefore, by changing the frequency processing effect according to the enlargement ratio, there is an effect of eliminating the difference between the processing effect for the image with the changed enlargement ratio and the processing effect when the enlargement ratio is not changed. This is important when performing a diagnosis or the like, and has an effect of reducing the influence on the diagnosis.
[0040]
Here, the enlargement ratio and the processing effect are shown in a linear diagram, but may be in a non-linear relationship. In human subjective evaluation, there are cases where the enlargement ratio and the processing effect (coefficient of coefficient) cannot be said to be linear.
Therefore, even if the enlargement ratio and the processing effect (factor of the coefficient) do not have a linear relationship, there is an effect that can be dealt with.
[0041]
It is also possible to change the coefficient magnification for each subband. This is because, in human subjective evaluation, the relationship between the enlargement ratio and the processing effect (coefficient magnification) may differ for each frequency band, and a corresponding case may be required.
Therefore, there is an effect that the processing effect for each frequency band can be changed with respect to the enlargement ratio by changing the enlargement ratio and the magnification of the coefficient for each subband.
[0042]
Further, the value of the coefficient for each subband is changed in consideration of the frequency balance of the restored image (step S206). For example, when the value of the coefficient of the first level is increased, the frequency component corresponding to the first level is emphasized.
[0043]
Therefore, when the rate of changing the value of the coefficient at each level is changed, the frequency balance of the restored image is adjusted. Here, the first level corresponds to the component with the highest frequency, and the level shifts to a lower frequency component as the level sequentially decreases to the second and third.
[0044]
Also in this case, the frequency balance is changed in consideration of the enlargement ratio.
For example, as shown in FIG. 5, when the enlargement ratio is small, 501 is selected, and when it is large, 502 is selected. This frequency balance also shifts smoothly according to the magnification. This is because the frequency band after the enlargement shifts according to the enlargement ratio of the image.
[0045]
Even in this case, it is possible to make a non-linear shift with respect to the magnification. This is because, as described above, the relationship between the enlargement ratio and the processing effect (the coefficient magnification) is not always linear.
[0046]
Here, the frequency balance refers to changing the processing effect of each frequency band on the restored image by adjusting the increase or decrease of the coefficient for each subband. In general, one of the advantages of multi-frequency processing is that it is easy to adjust the frequency balance. Therefore, by changing the frequency balance linearly or non-linearly according to the enlargement ratio, there is an effect of enabling a frequency processing effect suitable for human subjective evaluation.
[0047]
If it is desired to have a noise reduction processing effect, the coefficient values can be converted, for example, using a coefficient conversion curve shown in FIG. 6 before performing the above-described coefficient conversion.
In this case, generally, a value in a range where the coefficient value corresponding to the noise component is small (the area of the cutoff width 601 in the figure) is set to 0.
[0048]
Further, this area 601 is changed according to the enlargement ratio. As the enlargement ratio increases, the range of the noise region also increases accordingly. Therefore, the cutoff width 601 is increased according to the enlargement ratio. Therefore, there is an effect of stably reducing noise even when the magnification is changed.
[0049]
Then, the restoration circuit 117 performs an inverse discrete wavelet transform process based on the high-frequency coefficients changed by the coefficient transformation circuit 116. The input coefficients are subjected to two filter processes of u and p, are upsampled and then superimposed to output an image signal x ′. These processes are performed by the following equations.
[0050]
x ′ (2 * n) = s ′ (n) −floor ((d ′ (n−1) + d ′ (n)) / 4) (3)
x '(2 * n + 1) = d' (n) + floor ((x '(2 * n) + x' (2 * n + 2)) / 2) (4)
[0051]
With the above processing, one-dimensional inverse discrete wavelet transform processing is performed on the transform coefficients (step S207). In the two-dimensional inverse discrete wavelet transform, one-dimensional inverse transform is sequentially performed in the horizontal and vertical directions of an image, and details thereof are publicly known, and thus description thereof is omitted here.
[0052]
Here, when the coefficient is increased, the effect of the sharpening process is generally obtained, and when the coefficient is reduced, the effect of the smoothing is obtained. Then, the restored image is subjected to gradation conversion using, for example, a gradation conversion curve 801 in FIG.
[0053]
Here, since the pixel value width of the subject is normalized in advance, the density of the subject of the image after the gradation conversion falls within a certain range of the density width (a region where the density becomes black or white). The problem does not arise.
[0054]
As described above, in the first embodiment, since the frequency processing is performed so that the frequency processing effect is different according to the enlargement ratio of the image, the same frequency processing effect is obtained even when the image is enlarged or reduced. There is an effect that can be obtained.
Further, since the enlargement and reduction ratios of the image are determined based on the width of the subject, there is an effect that the pixel value width of the subject image after the enlargement / reduction processing is the same, and the image after gradation conversion has almost the same density value width. Therefore, there is an effect that the variation of the image is reduced.
[0055]
At the same time, since the frequency processing effect is also changed according to the subject width, the frequency processing effect is also applied irrespective of the subject width, and the processed image has no variation between subjects due to a synergistic effect with the above-described normalization effect. The effect is obtained.
[0056]
Further, by changing the frequency processing effect according to the enlargement ratio, there is an effect of eliminating the difference between the processing effect for the image whose magnification ratio has been changed and the processing effect when the enlargement ratio is not changed. This is important when performing a diagnosis or the like, and has an effect of reducing the influence on the diagnosis.
[0057]
Further, by changing the magnification of the coefficient and the magnification of the coefficient for each subband, there is an effect that the processing effect for each frequency band can be changed with respect to the magnification.
[0058]
In addition, by changing the frequency balance linearly or non-linearly according to the enlargement ratio, there is an effect of enabling a frequency processing effect suitable for human subjective evaluation.
Also, there is an effect of stably reducing noise even when the magnification is changed.
[0059]
【The invention's effect】
As described above, according to the present invention, since the frequency processing is performed so that the frequency processing effect is different according to the enlargement ratio of the image data, the same frequency processing effect can be obtained even if the image is enlarged or reduced. There is.
[0060]
According to another feature of the present invention, since the image data compression and enlargement ratio are determined based on the width of the subject, there is an effect that the pixel value width of the subject image after the compression / enlargement processing is the same, Subsequent images have substantially the same density value width, and thus have the effect of reducing image variations.
[0061]
At the same time, since the frequency processing effect is also changed according to the subject width, the frequency processing effect is also applied irrespective of the subject width, and the processed image has no variation between subjects due to a synergistic effect with the above-described normalization effect. The effect is obtained. This is important when performing a diagnosis or the like, and has an effect of reducing the influence on the diagnosis.
[0062]
According to another feature of the present invention, the same frequency processing effect can be obtained irrespective of the image enlargement ratio by changing the coefficient magnification as the enlargement ratio increases.
[0063]
Further, according to another feature of the present invention, there is an effect of enabling a frequency processing effect suitable for human subjective evaluation by changing a frequency balance linearly or non-linearly according to an enlargement ratio.
[0064]
Further, according to another feature of the present invention, there is an effect that noise is stably reduced even when the magnification is changed.
[Brief description of the drawings]
FIG. 1 is a block diagram illustrating a configuration of an image processing apparatus according to a first embodiment of the present invention.
FIG. 2 is a flowchart illustrating a processing procedure of the image processing apparatus according to the first embodiment of the present invention.
FIG. 3 is a characteristic diagram illustrating an example of histogram analysis.
FIG. 4 is a characteristic diagram showing a relationship between an enlargement ratio and a coefficient magnification.
FIG. 5 is a characteristic diagram showing a relationship between an enlargement ratio and a frequency balance.
FIG. 6 is a characteristic diagram for explaining a cutoff width.
FIG. 7 is a characteristic diagram illustrating a relationship between an enlargement ratio and a cutoff width.
FIG. 8 is a characteristic diagram illustrating an example of a gradation conversion curve.
FIG. 9 is an explanatory diagram of a discrete wavelet transform and its inverse transform.
[Explanation of symbols]
113 analysis circuit 114 first gradation conversion circuit 115 frequency component decomposition circuit 116 coefficient conversion circuit 117 restoration circuit 118 second gradation conversion circuit

Claims (5)

Frequency component decomposition means for decomposing the image data into a plurality of frequency components,
Coefficient conversion means for converting a coefficient for each frequency band of the image data decomposed by the frequency component decomposition means, based on the compression ratio of the image data, an enlargement ratio,
An image processing apparatus comprising: a restoration unit that inversely transforms a coefficient converted by the coefficient conversion unit. Analysis means for calculating image data compression and magnification by calculating the maximum and minimum values of the subject;
Frequency component decomposition means for decomposing the image data into a plurality of frequency components,
A coefficient conversion unit that converts the coefficient for each frequency band decomposed by the frequency component decomposition unit based on the compression and expansion ratio calculated by the analysis unit,
An image processing apparatus comprising: a restoration unit that inversely transforms a coefficient converted by the coefficient conversion unit. The image processing apparatus according to claim 1, wherein the coefficient conversion unit performs conversion such that a magnification of the coefficient decreases as an enlargement ratio of the image data increases. The image processing apparatus according to any one of claims 1 to 3, wherein the coefficient conversion unit performs coefficient conversion such that a frequency response differs as the compression and enlargement ratio of the image data is changed. . The image processing apparatus according to claim 1, wherein the coefficient conversion unit changes the cutoff width as the compression and enlargement ratio of the image data is changed to perform coefficient conversion.

Images