JP2892677B2 - Image processing method - Google Patents

Description

Description: BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an image density conversion system and an image background removal system, and more particularly to an image processing system suitable for medical field image diagnosis.

[Conventional technology]

Conventionally, an image enhancement method using histogram matching is described in "Procedures of the Eye".
EE, VOL. 69, No. 5, 1981/5, p. 655 (Pro
ceedings of the IEEE, VOL.69, No.5, May 1981, pp643-6
55) ”.

The background removal method is discussed in "Digital Image Processing, Modern Science Co., Ltd., 1978, pp. 266 to 276".

[Problems to be solved by the invention]

In the above-described related art, an input histogram and a reference histogram necessary for creating a density conversion function are calculated from the entire image. Here, the input histogram is a density value histogram of an image to be processed (hereinafter, referred to as an input image), and the reference histogram is an emphasized image having a target density distribution (hereinafter, referred to as a reference image). 3) is a density value histogram.

Here, although the backgrounds of the input image and the reference image and the size of the observation target area are different from each other, the density distribution is almost completely matched, so that an artifact is generated in the output image. There was a problem that the contrast and brightness did not match the target area of the reference image.

Further, as a conventional technique of background removal, there is a method of setting a threshold on a histogram and performing threshold processing on the entire image to perform background removal on the assumption that there is a density difference between the background and the object. In this method, the histogram has a background,
If there is a peak corresponding to the target area and a clear valley between the peaks, the valley can be automatically detected and a threshold can be set. However, in practice, there are few cases where there are clear valleys,
In addition, there is no guarantee that the density corresponding to the valley is the boundary density between the background and the target area, and the threshold value is often set by trial and error, resulting in a problem that the operability and reliability are poor.

In view of the above, an object of the present invention is to reduce artifacts caused by inappropriate density conversion due to the difference in the area of the background area between the input image and the reference image, and to match the contrast and brightness between the input image and the reference image. To make it happen.

Another object of the present invention is to improve operability of background removal, and
It is to perform the automation.

It is still another object of the present invention to prevent the influence of the noise component of the histogram from affecting the conversion function, and to reduce the artifact by a method of roughly adjusting the density distribution.

[Means for solving the problem]

In order to achieve the above object, the present invention divides an image into a background and a target region, obtains an input / standard histogram from the target region, and performs density conversion such that the input histogram matches the reference histogram. .

In order to achieve the above and other objects, the present invention sets an arbitrary frame that does not include a background in an image, and sets a minimum frame within the frame.
Find the maximum density, lower than the minimum density in the image, or
Any image having a density higher than the maximum density is considered as a background and is removed from the image, and the background is automatically removed by automatically setting the frame size.

In order to achieve the still another object, a smoothing process is performed on the histogram, a density conversion function is created by reducing the number of density levels, and interpolation is performed on the density conversion function.
A density conversion function is created by restoring the number of density levels.

[Action]

By the above means, it is possible to perform density conversion using the density distribution of only the target area, realize appropriate density conversion processing, and match the input image with the contrast and brightness of the reference image without generating artifacts. Can be.

In addition, the background is located around the image by the other means, and the background region in the image can be appropriately removed from the fact that the background is brighter or darker than the target region.

〔Example〕

Hereinafter, an embodiment of the present invention will be described with reference to FIGS.

FIG. 1 is a diagram for explaining the flow of the entire process. Background-removed input image excluding the background part in the image from the input image 10
11 is obtained, and an input histogram 12 is calculated from the target area. On the other hand, the same type as the input image, that is, an image in which the imaging device, the part and the like match, the image emphasized for diagnosis, that is, the background removal image 14 is obtained from the reference image 13,
Further, a reference histogram 15 is calculated. Then, a density conversion function 16 such that the input histogram matches the reference histogram is obtained, and density conversion is performed on all the pixels of the input image using the conversion function 16. This converted image is the output image 17. Next, an example of creating the density conversion function in FIG. 1 will be described with reference to FIG.

In FIG. 2, the statistic 2 concerning the density is obtained from the input to the background-removed image and the reference histograms 21 and 23, respectively.
Ask for 2,24. Here, Min and σin are the average and standard deviation of the density of the background-removed input image, respectively. Mref,
σref is the average / standard deviation of the density of the background-removed reference image. A density conversion function 27 for obtaining an output density y26 with respect to an input density x25 from these statistics 22 and 24 is given by the following equation.

However, if y is larger than the density range, it is clipped to its maximum density, and if y is negative, it is clipped to 0.

Here, Equation (1) means that piecewise linear density conversion is performed so that the statistic of the background-removed input image matches the statistic of the background-removed reference image. In addition, the fact that the average and the density deviation of the densities of the images match corresponds to the matching of the brightness and contrast, which are the visual feature amounts of the images. In other words, when the statistic is obtained from the entire image as in the above example, the background area in the image is reflected in the statistic, so that the brightness and contrast of the area to be observed (the area excluding the background) match. If not, there will be problems. However, if the statistics after the background removal is used, it can be said that the brightness and the contrast are the same regardless of the difference in the size of the background region. As described above, the greatest feature of the present invention is that density conversion is performed without using unnecessary information in an image. Next, another example of creating a density conversion function will be described with reference to FIG.

The histograms Hin31 and Href32 are obtained from the input and the reference image after the background removal. Here, the minimum and maximum densities of the input histogram are set as imin 311 and imax 312, and the minimum and maximum densities of the reference histogram are set as rmin321 and rmax322. In addition, since the total frequency number is different between Hin and Href, it is necessary to make the total frequency number equal to either one. Here, the size is changed so that the total frequency of Hin matches the total frequency of Href. Further, Hin and Href are shifted with respect to the density axis so that the histogram has frequency from the density 0.
in33 and Href34 are obtained. Then, a density conversion function 36 for converting Hin33 into Href34 is obtained by the histogram matching method 35. This density conversion function 36 corresponds only to the density range of the observation area. Therefore, the linear conversion 38 is performed on the density corresponding to the background. The density of the observation area is calculated by the conversion function
The amount of shift of the histogram shifted to obtain 36 is obtained by 39 conversion functions shifted in the reverse direction with respect to the conversion function. Finally, the entire density conversion function 37 is obtained.

When the density conversion is performed on the input image using the conversion function 37, the brightness of the reference image is obtained for both the background and the observation area.
Close to contrast. Also, originally, the low density area in the observation area is changed to the background density, and an unnatural density difference (artifact) does not occur in the observation area.

Next, the details of the background removal method in FIG. 1 will be described with reference to FIGS.

FIG. 4 shows a background removal method using a chest X-ray image as an input image 41.

The chest X-ray image includes a background region 411, a lung field 412 corresponding to a lung, and a mediastinum 413 as another high density region.

As a feature of the input image 41, the regions 412, 41 other than the background
The feature 3 is that the variation in density is almost constant irrespective of the location, so the following background removal method can be applied. That is, a rectangular frame 42 as large as possible that does not include the background in the input image 41 is set, and the minimum and maximum densities in the frame are calculated (4
3) Yes. Then, the minimum / maximum density is set as a threshold value 44 and threshold processing (45) is performed on an area outside the frame of the input image 41.
I do. Specifically, for all pixels in the area outside the frame,
Pixels having a density smaller than the minimum density or larger than the maximum density are removed as a background. Thus, the background-removed input image 46 is obtained.

Next, the automatic setting method of the frame in FIG. 4 will be described with reference to FIG.

On the input image 41, the frame size 511 is continuously changed 51 from a size as large as possible to a size as small as possible. Change in minimum density 5 when frame size 511 is changed in this way 5
When observing No. 2, when the frame size is large, that is, when the background area is included in the frame, the minimum density is low and the change 53 is a gradual change. Then, at a position 541 where the frame size does not include the background, the minimum density changes abruptly 54. In a state in which the frame size is further reduced, that is, in a frame size that does not include the background at all, a gentle change 55 occurs again.

Here, when changing the frame size from large to small, the first abrupt change 54 is detected, and the frame size of a portion having a gradual change thereafter is set as an area for threshold calculation. Then, if the threshold processing is performed in the same manner as in FIG. 4, the background removal can be automated.

Next, another automatic method of setting the frame in FIG. 4 will be described with reference to FIG.

On the input image 61, the frame size 61 is continuously changed from the maximum size that does not protrude the image to a small size, and a histogram of the entire image is obtained within the range of the minimum and maximum densities in the frame. At the same time, a correlation value 64 between the closest histograms 63 obtained continuously is obtained. Here, when changing the frame size from large to small, the smallest large valley 65 is detected,
Thereafter, the maximum frame size having a gradual change is set as an area for threshold calculation. Then, if threshold processing is performed in the same manner as in FIG. 4, the background removal can be automated. In the above automatic frame setting method, the frame size is changed from large to small, but it goes without saying that the reverse is also possible.

Next, referring to FIG. 7, after performing a smoothing process on the histograms of both the input image and the reference image, an image processing method for creating a density conversion function and a histogram smoothing process,
Image processing that reduces the number of density levels, creates a density conversion function using the histogram at the reduced level, and then interpolates the density conversion function to restore the number of density levels and create a density conversion function An embodiment of the system will be described.

The input histogram 71 is smoothed to reduce the number of density levels to a fraction (72). In this example, 25
It has decreased from 5 levels to 8 levels. The same operation 74 is performed on the reference histogram 73. Histogram matching 75 is performed from the histograms 72 and 74 in which the range has been reduced in this way, and the range compression input histogram 72
Is converted into a range reduction reference histogram 74 to obtain a density conversion function 76. Furthermore, a final density conversion function 77 is obtained by performing linear interpolation to return the number of density levels to the original 255 levels.

According to this method, it is possible to realize piecewise linear density conversion with less nonlinearity without affecting the density conversion function by the minute change of the histogram.

〔The invention's effect〕

According to the present invention, by performing density conversion using the image after the background removal, it is possible to make the contrast and brightness of the region to be observed coincide with the reference image while suppressing the artifact of the processed image.

In addition, the background area with good operability can be removed by an appropriate frame pre-setting for realizing the background removal processing or an automatic setting method of the optimum frame.

Further, by smoothing the histogram and piecewise linear transformation, it is possible to obtain an image having a smooth change and reduced artifact.

[Brief description of the drawings]

FIG. 1 is a flowchart showing the overall processing of the present invention, FIG. 2 and FIG. 3 are explanatory diagrams of an embodiment relating to the creation of a density conversion function in the present invention, and FIG. FIG. 5 and FIG. 6 are explanatory diagrams of an embodiment of an automatic system of a background removal system, and FIG. 7 is an embodiment of a histogram smoothing and piecewise linear density conversion. FIG.

Continuation of the front page (72) Inventor Tetsuo Yokoyama 1099 Ozenji Temple, Aso-ku, Kawasaki-shi, Kanagawa Prefecture System Development Laboratory, Hitachi, Ltd. (58) Fields investigated (Int. Cl. ⁶ , DB name) G06T 1/00 G06T 5/00 A61B 6/00 H04N 1/387

Claims (8)

(57) [Claims] 1. An image processing method, wherein a density distribution after removing a background portion in an image is used in a density conversion process for matching a density distribution of an input image with a density distribution of a preset reference image. 2. An arbitrary frame that does not include a background in an image is set, and the minimum and maximum densities in the frame are obtained. Pixels having a density lower than the minimum density or higher than the maximum density in the image are set as background. As an image processing method to remove from the image. 3. In the frame setting according to claim 2, when the size of the frame is continuously reduced from the maximum size which does not protrude from the image, the change of the minimum density in the frame increases, and thereafter the steady state changes. An image processing method that automatically detects the status of a part and automatically sets the frame size. 4. In the frame setting according to claim 2, when the size of the frame is continuously enlarged from the minimum size, the change in the minimum density in the frame increases, and thereafter, a portion where the steady state is reached is determined. An image processing method that detects and automatically sets the frame size. 5. In the frame setting according to claim 2, when the size of the frame is continuously reduced from the maximum size which does not protrude from the image, the histogram of the range of the minimum / maximum density in the frame is adjacent to the histogram. An image processing method that continuously checks the correlation with the histogram in the range of the minimum and maximum densities in the frame, detects a valley where the correlation is small, and automatically sets the frame size. 6. In the frame setting according to claim 2, when the size of the frame is continuously enlarged from the minimum size, the histogram of the range of the minimum / maximum density in the frame and the minimum / maximum density in the adjacent frame are set. An image processing method that continuously checks the correlation with the histogram in the maximum density range, detects valleys where the correlation is small, and automatically sets the frame size. 7. The method according to claim 1, wherein the input image,
An image processing method that creates a density conversion function after smoothing the two-sided image histogram of a reference image. 8. After the histogram smoothing process according to claim 7, the number of density levels is reduced, a density conversion function is created using the histogram at the reduced level, and then the density conversion function is interpolated. An image processing method wherein the number of density levels is restored to create a density conversion function.