JPH01237887A - Method for reducing granular noise in radiograph - Google Patents

Abstract

PURPOSE:To obtain a natural image with high image quality by adding random noise to smoothed image data. CONSTITUTION:An accumulating phosphor sheet 1 is scanned by excitation light 5, and stimulated phosphorescent light 9 is received by a photomultiplier 11. The output of the photomultiplier 11 is inputted to an arithmetic part 19 via an amplifier 16, and an A/D converter 17, and a memory 18. At the arithmetic part 19, the smoothed image data is found by applying smoothing processing so as to reduce granular noise on a radiograph, and thereafter, the random noise is added to the smoothed data, then, image processing data can be obtained. The processing image data is sent to an image display means 20.

Description

Detailed Description of the Invention (Field of Industrial Application) The present invention is a method for reducing granular noise, which reduces granular noise in radiographic images by performing image processing on image data obtained by reading recorded radiographic images. It is about the method.

(Prior Art) It is practiced in various fields to read a recorded radiation image to obtain image data, perform appropriate image processing on this image data, and then reproduce and record the image. For example, an X-ray image is recorded using an X-ray film with a low gamma value designed to be compatible with later image processing, and the X-ray image is read from the film on which it is recorded and converted into an electrical signal. After converting and performing image processing on this electrical signal (image data), contrast and sharpness can be improved by reproducing it as a visible image in photocopies, etc.

Efforts have been made to obtain reproduced images with good image quality performance such as graininess (see Japanese Patent Publication No. 5193/1983).

The applicant has also proposed radiation (X-rays, α-rays).

When irradiated with β rays, γ rays, electron beams, ultraviolet rays, etc., a part of this radiation energy is accumulated, and then when irradiated with excitation light such as visible light, stimulable fluorescence exhibits stimulated luminescence depending on the accumulated energy. Radiographic image information of a subject such as a human body is partially recorded on a sheet of stimulable phosphor using a stimulable phosphor, and this stimulable phosphor sheet is scanned with excitation light such as a laser beam. The resulting stimulated luminescence light is read photoelectrically to obtain image data, this image data is subjected to image processing, and a radiation image of the subject is photographed based on the image data after image processing. Recording materials such as photosensitive materials, C
A radiation image recording and reproducing system that outputs a visible image to RT, etc. has already been proposed (Japanese Patent Laid-Open No. 55-1242
No. 9, No. 56-0395, No. 55-163472,
5Ei-104645, 55-116340, etc. ).

Due to the fluctuation of the radiation when recording the radiation image, the recorded radiation image usually contains a lot of granular noise. When an image is reproduced based on image data obtained by reading this radiographic image, if the reproduced image contains a lot of granular noise, the reproduced image will appear coarse and grainy. On the other hand, when image processing is performed to reduce granular noise, other image quality performance (contrast, sharpness, etc.) usually tends to deteriorate.

In the above-mentioned systems, for example, Japanese Patent Application Laid-Open No. 55-163472 and Japanese Patent Application Laid-open No. 55-87 proposed by the present applicant.
No. 953, Japanese Patent Application No. 12-265017, Japanese Patent Application No. 1987
Sharpness is achieved by using the method described in Japanese Patent No. 285018, that is, an image processing method that emphasizes low spatial frequency components and high spatial frequency components of an image based on read image data and combines them.

Grain noise is reduced while maintaining balance with contrast, etc.

(Problem to be Solved by the Invention) Each part of an image has a different spatial frequency component, and there are parts that change rapidly and parts that change little. When this image is subjected to the image processing described above to reduce the grainy noise, even though the grainy noise is appropriately reduced for the entire image, some parts of the image may have too much grainy noise removed, and when you look at that part, it is rather unpleasant. There have been problems in that the image may appear unnatural or a false image may occur in that area, resulting in the overall image quality not improving much or even deteriorating in some cases.

In view of the above-mentioned problems, the present invention eliminates the occurrence of unnaturalness and false images due to the removal of granular noise, gives a more natural feeling, and provides reproduced images with high image quality performance. The purpose of this invention is to provide a noise reduction method.

(Means for Solving the Problems) A method for reducing granular noise in a radiographic image according to the present invention reads a recorded radiographic image to obtain a large number of initial image data carrying information representing this radiographic image. The present invention is characterized in that after obtaining smoothed image data by subjecting the radiographic image to smoothing processing to reduce granular noise, irregularity noise is added to the smoothed image data.

(Function) The method for reducing grainy noise in a radiographic image of the present invention is to obtain smoothed image data by subjecting the initial image data to smoothing processing to reduce grainy noise in the radiographic image, and then to obtain smoothed image data. Since irregular noise (hereinafter referred to as random noise) is added to the data, the added random noise eliminates the occurrence of unnaturalness and false images in the reproduced image due to the removal of granular noise. , it is possible to obtain a reproduced image that gives a natural feel and has high image quality performance.

(Example) Examples of the present invention will be described below with reference to the accompanying drawings.

First, with reference to FIG. 2, an embodiment of an apparatus using the present invention will be described.

FIG. 2 is a perspective view showing an example of a radiation image reading and displaying device using the granular noise reduction method of the present invention. This device accumulates a portion of this radiation energy when it is irradiated with radiation (X-rays, alpha rays, beta rays, gamma rays, electron beams, ultraviolet rays, etc.), and then when it is irradiated with excitation light such as visible light. This device uses a stimulable phosphor (stimulable phosphor) that exhibits stimulated luminescence depending on energy.

In a photographing device (not shown), a subject such as a human body is irradiated with radiation and photographed, and a radiation image of the subject is stored and recorded on a stimulable phosphor sheet.

The photographed stimulable phosphor sheet 1 is set at a predetermined position in the radiation image reading and reproducing apparatus shown in FIG.

The stimulable phosphor sheet 1 set in this manner is conveyed (sub-scanning) in the direction of arrow Y by a sheet conveying means 3 such as an endless belt driven by a motor 2.

On the other hand, the excitation light 5 emitted from the laser light source 4 is transmitted to the motor 1.
A rotating polygon mirror 6 that is driven by 3 and rotates at high speed in the direction of the arrow.
After passing through a focusing lens 7 such as an fθ lens, the optical path is changed by a mirror 8 and the sheet 1
, and main scanning is performed in the direction of arrow X, which is substantially perpendicular to the sub-scanning direction (direction of arrow Y). From the location of the sheet 1 irradiated with this excitation light 5, stimulated luminescence light 9 is emitted in an amount corresponding to the radiation image accumulated and recorded at that location, and this stimulated luminescence light 9 is transmitted to the light collector IO. The light is collected by a photomultiplier (photomultiplier tube) 11 as a photodetector and photoelectrically detected. The light collector 10 is made by molding a light-guiding material such as an acrylic plate, and has a linear incident end surface 10a extending along the main scanning line on the stimulable phosphor sheet 1. The light receiving surface of the photomultiplier 11 is coupled to the output end surface 10b formed in an annular shape. The stimulated luminescent light 9 entering the light condenser 10 from the input end face 10a travels through the light condenser 10 through repeated total reflection, exits from the output end face fob, and is received by the photomultiplier 11, A photomultiplier 11 detects the amount of stimulated luminescence light 9 carrying information representing the radiation image. The analog output signal S output from the photomultiplier 11 is amplified by the amplifier 16, sampled and digitized by the A/D converter 17. The initial image data S org obtained in this way is input to the memory 18 and stored. The initial image data S org stored in the memory 18 is then read out and input to the calculation unit 19 . As will be described later, the calculation unit 19 performs smoothing processing to reduce grainy noise in the radiation image based on this initial image data S org to obtain smoothed image data S', and then further smoothes the image data S'. Image data S
Random noise is added to ' to obtain processed image data S'. The processed image data S' obtained in this manner is sent to an image display means 20 such as a CRT display, and the processed radiation image P is reproduced and displayed based on the processed image data S'.

FIG. 3 is a block diagram showing an example of a method in which the calculation unit 19 shown in FIG. 2 performs smoothing processing on the initial image data S org to obtain smoothed image data S', and then adds 2 random noise. It is a diagram.

This example uses the method described in Japanese Patent Application Laid-Open No. 55-16347, and uses the image state (
For example, a so-called adaptive smoothing process is performed in which smoothing is performed to suit each portion (eg, flat portions, rapidly changing portions, etc.).

The initial image data S org read out from the memory 18 in FIG. High frequency emphasis calculation section 23. The signal is sent to a low-pass filter 24.

In the low-pass filter 21, each input initial image data Sorg is subjected to smoothing processing that averages surrounding initial image data Sorg on the radiation image, and a blur mask Sus is obtained. Note that the two low-pass filters may be a V-filter or a median filter, which will be described later. Bypass filter 2
In step 2, the difference S org-3us between the input initial image data S org and the blur mask Sus is calculated, thereby obtaining bypass image data S org-Sus in which the high frequency components of the initial image data S org are emphasized. This bypass image data Song-5us is sent to the high-frequency emphasis calculation section 23 and the binarization processing section 25. Based on the input initial image data S org and bypass image data Sorg-3us, the high frequency enhancement calculation unit 23 calculates the following equation using the following formula: %(1) is performed, and the high-frequency emphasized image data Shl
is required.

On the other hand, the low-pass filter 24 obtains low-frequency emphasized image data Slow that emphasizes the low spatial frequency components of the radiation image based on the input initial image data S org. As the low-pass filter 24, a ■-filter, median filter, etc. are used ("Medical Electronics and Bioengineering", Vol. 15, No. 5 (Sept, 1977), 2
(See pages 5 to 31 <Image quality improvement of R1 image using nonlinear digital filter>).

■-Filter is four neighboring small regions RL of size kXk around each point of interest (Lj) on the image;
2, 3, and 4, and the average value μ of image data within each R2
, and the output value is determined according to the variance value σL2. In other words, when the input image data to this filter is expressed as la''l, and the output image data as (b,,l), it is a filter that performs the calculation of j. is μL−−Σ ai・j7 ・・・・・・(3
) k 2 (にJ ) al 纺.

σLz,,,□Σ (a=, aL) 2=・・・(
4) R2 (1°, p insect IJ, WL is the weight determined by the value of the variance σt2 in RL, e.g., Refers to image data.

By using the above-mentioned V filter, it is possible to perform smoothing (low-pass) on a portion of an image where there is a sudden step-like change (edge) while leaving the sudden change intact.

A median filter is a filter that outputs the median value (median) of image data values within a local small area of an image. When this median filter is used, spike-like noise can be removed without blurring the edges too much.

In this way, the low-frequency emphasized image data Slow obtained by the low-pass filter 24 and the high-frequency emphasized image data Shi obtained by the high-frequency enhancement calculation section 23 are manually input to the processing section 27. In the processing unit 27, when the weighting of the high-frequency emphasized image data Shi is set to 1 and the weighting of the low-frequency emphasized image data is set to 1-1 and the smoothed image data S' after processing, S' - (1-K )XSlow +KXShi・
...The calculation (6) is performed. Here, K is a variable depending on each part of the image. Therefore, each part of the image is appropriately reduced and emphasized or high-frequency emphasized.

This weighting factor is determined by the binarized image processing section 25 and the smoothing processing section 26 as follows.

FIG. 4 is a diagram showing an example of how to obtain this weighting factor. Although the weighting coefficients are obtained two-dimensionally over the entire area of the image, only the X direction (one arbitrary direction) is shown here for the sake of simplicity.

(A) in FIG. 4 is a bypass image signal S in the X direction.
org-8us is shown. The binarization processing unit 25 binarizes the input bypass image data Sorg -5us using an appropriate threshold value and outputs binarized data as shown in (B). The smoothing processing unit 26 smoothes this binarized data to obtain a weighting coefficient K that does not have sudden changes, as shown in (C). By performing the calculation shown in equation (6) using the weighting coefficient K obtained in this way, smoothed image data S' that has been appropriately image-processed for each part of the image is obtained. Also, as a weighting coefficient, for example, Fig. 4 (A
) or the binarized image data shown in (B), the image processing method will change rapidly every 5 minutes in each adjacent part of the image, resulting in a change in the image data after processing. When an image is played back based on the image, it becomes a very hard to read image, but since the weighting coefficients are obtained by smoothing the binarized data as described above, there are differences between adjacent parts of the image. The image processing method changes gradually, making it possible to obtain a more natural reproduced image.

In addition, as shown by the broken line in FIG. 3, initial image data S org is also input to the processing unit 27, and Shi, Sorg,
The smoothed image data S' may be obtained based on the three Slow image data. Furthermore, the smoothed image data S' may be obtained based on the two image data of Sorg and Shi without inputting Slow, and S! without inputting Shi! Smoothed image data may be obtained based on the two image data of ow and S org.

The smoothed image data S' obtained by the processing section 27 is
The signal is input to the random noise adding section 28 shown in the figure. The random noise addition unit 28 adds random noise to the smoothed image data S' input over the entire image to obtain final processed image data S'.
The image is sent to an image display means 20 (see FIG. 2) such as a CRT display. The image display means 20 reproduces and displays the image-processed radiation image P based on the processed image data S'.

In the random noise adding section 28 shown in FIG. 3, by adding random noise to the smoothed image data S', the smoothed image data S without adding random noise is
It is possible to eliminate unnaturalness due to the removal of granular noise and generation of false images, which are observed when an image is reproduced and displayed based on .

FIG. 1 is a graph showing an example of the relationship between mf of added random noise and variance σ2 for each small region of an image.

Graph D shows that a constant amount of random noise is added to the entire image regardless of the value of the variance σ2 for each small region of the image. Graph E shows that as the variance σ2 value for each small area of the image increases, the random noise f added to the smoothed image data S' corresponding to that small area increases.
This indicates that f1f is increased. It goes without saying that the standard deviation σ may be used instead of the variance σ2,
Here, the standard deviation σ is referred to as the variance σ2.

Even if an equal amount of random noise is added to the entire area of the image as shown in graph D, it is possible to prevent the above-mentioned unnaturalness and false images from occurring in the reproduced image, but areas where the value of variance σ2 is large are smoothed. Since more granular noise is generally removed from the image data S', the unnatural feeling increases. Therefore, as the value of the variance σ2 increases, more fine-grained image processing will be performed by increasing the random noise ff1f added to the smoothed image data S' corresponding to that small area. .

However, if the Qf of the added random noise increases too much, the effect of removing granular noise will decrease, so as shown in graph E, no matter how large the value of variance σ2 is (no matter how thick the Qf of the random noise is It is preferable to do so.

The variance σ2 may be obtained based on the smoothed image data S', and as shown in FIG.
The initial image data s org may be input to 8 and the image data may be calculated based on this manually generated bright period image data s org.

In addition, the difference σ12−σ22 between the variance σ12 calculated based on the initial image data S org and the variance σ22 calculated based on the smoothed image data S′ is calculated, and the larger the value of this difference, the more random noise will be superimposed. It is also possible to increase the Qf of . This is because the larger the difference, the greater the amount of granular noise removed.

In addition, in order to increase the added random noise f and f as the amount of granular noise removed increases, the difference between the initial image data s org and the smoothed image data S' is calculated, and based on this difference data, The variance σ2 for each small region of the image may be determined, and the larger the value of the variance σ2, the greater the Etf of the random noise to be added.

Note that if the spatial frequency carried by the random noise includes more spatial frequency components higher than the spatial frequency components contained in the removed granular noise, in the reproduced image,
This method is more effective in eliminating unnaturalness and preventing false images without reducing the effect of removing granular noise.

Note that the image processing method shown in FIG. 3 applies not only to the radiation image reading and display device shown in FIG. This device can be generally used when performing image processing on image data read from a radiographic image, such as a device that performs image processing on image data read from a radiation image, such as a device that reproduces it as a visible image on a photocopy or the like.

Furthermore, the granular noise reduction method of the present invention includes not only the smoothed image data S' obtained by the method shown in FIG. The present invention can be applied to any smoothed image data in which granular noise has been reduced by smoothing processing, regardless of the type of smoothing process.

(Effects of the Invention) The method for reducing granular noise in radiographic images of the present invention is to obtain smoothed image data by performing smoothing processing on initial image data to reduce granular noise in radiographic images, and then to obtain smoothed image data. Because we added random noise to the data,
The added random noise eliminates the unnaturalness and false images caused by the removal of granular noise in the reproduced image, making it possible to obtain a more natural and high-quality reproduced image.

[Brief explanation of the drawing]

FIG. 1 is a graph showing an example of the relationship between i:Lf of added random noise and variance σ2 for each small region of an image. FIG. 2 is a graph showing a radiation image reading using the granular noise reduction method of the present invention. A perspective view showing an example of a display device, FIG.
In the calculation unit 19 shown in the figure, initial image data Sor
A block diagram showing an example of a method of performing smoothing processing on g to obtain a smoothed image S' and then adding random noise. It is a figure showing an example of how to obtain a coefficient. 1... stimulable phosphor sheet 2.13... motor 3.
... Sheet transport means 4 ... Laser 6 ... Rotating polygon mirror 9 ... Stimulated luminescent light 1t ... Concentrator 11 ... Photo multiplier 1B ... Amplifier 17 ... A/ D converter 18...Memory 19...Arithmetic unit 2
0... Image display means 21... Low pass filter 22... Bypass filter 23... High frequency emphasis processing section 24... Low pass filter 25... Binarization processing section 26... Smoothing processing section 27...Processing section 28...Random noise addition section Fig. 1 Fig. 2 Fig. 3

Claims (1)

[Claims] (1) reading a recorded radiation image to obtain a large number of initial image data carrying information representing this radiation image;
The radiation radiation method is characterized in that after smoothing the initial image data to obtain smoothed image data by performing a smoothing process to reduce granular noise in the radiation image, irregularity noise is added to the smoothed image data. Image grain noise reduction method.