JPS60171573A - Picture emphasis system in picture processing system - Google Patents

Abstract

PURPOSE:To enable overall, effective picture emphasis processing by dividing original picture data into a partial picture, by selecting an optimum filter for every divided area and making it work, and by adding a processed picture at optional ratio. CONSTITUTION:Original picture data 21 is divided into a part picture, and characteristic extraction processing 22 of respective part areas whose characteristics as variance values are extracted from a small area for every part is performed. Next, the optimum processing among a filtering treatment 23, original picture adding treatment 25 and a contrast conversion treatment 27 emphasizing a specific frequency component is performed by using a transcendental known information 201, for example, a diagnosed part location, diagnosis object, etc. Further, by using the information 201, a filter function selection treatment 24 of a filter to be used for the treatment 23, a processed picture addition coefficient setting treatment 26 for the treatment 25, and a contrast conversion function selection treatment 28 for the treatment 27 is performed respectively. In this way, by using various processing means and adding the processed pictures at optional ratio, the system can overall, effective picture emphasis processing.

Description

[Detailed description of the invention] [Field of application of the invention] Concerning the image enhancement method of Bobobo in the ring system.

[Background of the invention]

Conventionally, since it is difficult to identify a lesion in a plain image (original image) taken with an X-ray device or the like, the original image is sometimes subjected to filter processing. If the original image is G and the image after filter processing is G', it can be expressed as G' = F*G in real space. Here, F is a filter function, *
represents convolution.

Different filter functions F are used depending on whether to emphasize high frequencies, band emphasis, or low frequencies.

Conventionally, when processing one image, a filter function F having the same characteristics has been used. If the entire image is processed using a filter function F having the same characteristics, the following problem will occur.

FIG. 1 is a sketch of an X-ray image of the foot. In FIG. 1, 11 to 13 represent bones.

Furthermore, the area 14 has a smooth pattern, and the area 15 has a complex pattern with variations. Therefore, for example, area 15
When a high-frequency emphasis filter is used to emphasize the pattern, noise generally becomes noticeable in the region 14. Conversely, if a band emphasis filter is used to emphasize region 14, the image in region 15 will be blurred. Therefore, the information contained in the image cannot be effectively displayed over the entire image.

On the other hand, Japanese Patent Application Laid-Open No. 55-87953 discloses a non-sharp camera for improving the diagnostic performance of X-ray photographs.

A soap treatment has been disclosed.

This means that the reproduced image is D′ and the original image is D er1.
The low-frequency emphasized image is D□, and β is a constant, and both reproduced images 1)'
is, D'=D. It is expressed as □+β(D6□-ri,,,). At this time, the constant β is changed depending on the concentration value of Du, for example. According to this method, depending on the density value of one image,
The original image D0,1 and the highly enhanced image (D , , -Du
, ) can be changed. This method has a problem in that although it is possible to change the degree of emphasis according to the density value of the image, it is not possible to select a filter function.

[Purpose of the invention]

SUMMARY OF THE INVENTION An object of the present invention is to provide an image enhancement method in an image processing system that can eliminate the above-mentioned similarities and perform effective image enhancement processing over the entire image.

[Summary of the invention]

In order to achieve the above object, one aspect of the present invention includes an image input section;
In an image processing system that has an image data processing section and a two-image display section, one image is divided into partial images, characteristics are extracted for each partial image, and a priori known information is used according to the characteristics. It is characterized by selecting filter functions, processing image addition ratios, contrast conversion functions, etc., and obtaining images with different emphasis methods depending on the diagnosis site, diagnosis purpose, etc.

[Embodiments of the invention]

Hereinafter, one embodiment of the present invention will be described in detail using the drawings.

FIG. 8 is a configuration diagram showing an example of an image processing system that is an embodiment of the present invention.

Image processing units 1 to 800 process image data. 8
01 is a film reader as an image input unit; 802 is a display having a keyboard and the like as an image display unit;
803 is a film writer serving as an image output unit, and 804 is an optical disk unit.

Further, FIG. 9 is a functional block diagram of the image processing system of FIG. 8.

Input X-ray image 900 is input to image processing unit 800 by film reader 801 . In the image processing unit 800, a feature amount extraction unit 9 for knowing the characteristics
02. Processing is performed by a similarity calculation unit 903 from the obtained feature amount, a selective filtering unit 904 that emphasizes specific frequency components in interactive mode, and a contrast conversion unit 905 to further emphasize differences in density changes. It will be done. The processing result of the image processing unit 800 is output as a processed X-ray image 920 by the film writer 803.

FIG. 2 is an explanatory diagram showing the processing concept of the present invention.

Original image data 21 input from film reader 801
A small region is set for each large portion, and characteristic extraction processing 22 for each partial region is performed to extract characteristics such as a variance value and an average value from the small region.

Next, according to this extraction characteristic, a priori known information 201,
For example, the most suitable treatment is selected from the following treatments using the diagnosis site, diagnosis purpose, etc. The processing includes a filtering process 23 that emphasizes a specific peripheral number component, an original image addition process 25, and a contrast conversion process 27.

Further, using the a priori known information 201, for the filtering process 23, a filter function selection process 24 of the filter to be used, for the original image addition process 25, a processed image addition coefficient setting process 26, and a contrast conversion process. For processing 27, contrast conversion function selection processing 28 is performed.

FIG. 3 shows a method for extracting characteristic values for each partial area. In the 6-original image 30, partial areas 31, 32, etc. are set for each pixel, and respective partial characteristics are extracted. For example, when the size of the partial region 31 is Ω×β and the density value of the pixel is glJ.

The variance value σ and the average value can be determined by the following formula: '66 g=-2ΣΣ g+J Here, 1+] represents the separation value of the coordinates (x, y) in the X-axis direction and the X-axis direction.

In addition to a rectangular area, it is also possible to set a meaningful area such as an organ.

Furthermore, as characteristic values, it goes without saying that differential values, co-occurrence matrices, power spectra, etc. can be used in the statistical analysis described above. In the explanation below.

To simplify the discussion, we will take variance and mean value as characteristic values.

First, regarding the filter function selection process 24, FIG. 4(a)
, (b), and (C).

In FIG. 4, (a) shows a filter function, (b) shows a correspondence function between the filter function and a composite characteristic value, and (C) shows a sketch image 49 with a foot. In the filter functions shown in FIG. 4(a), the horizontal axis represents the cutoff frequency fc and the vertical axis represents the intensity S, and a plurality of filter functions are prepared in advance. Although any characteristic can be set, a high-frequency emphasis characteristic is shown here as an example. As the filter function moves from 44 to 41, the cutoff frequency fc increases, and therefore the strength S increases, resulting in a characteristic that emphasizes higher frequencies more.

In addition, the correspondence function between the Buoy 3 router function and the characteristic value in Fig. 4(b) is that the horizontal axis is the composite characteristic value Q, and the vertical axis is the filter function 41 to 44 shown in the filter function in Fig. 4(a). The number is F.

The composite characteristic value Q can be determined, for example, as follows. Now, the dividing area value of each partial area obtained by dividing the original image 30 into meshes with the same size β×Q as the partial area 31 is 07, the maximum value of the variance value of each partial axis of the original image 30 is σ, and the minimum value Similarly, the average value of each partial area of the original image 30 is iA, and the maximum value of the average value is 1.
, when the maximum tJ\ value of the average value is 1□7, σ sentence −σ n+ln [, ′. Next, a normalized weight w+ t Ww (W+ + W2 '' 1 ) is given to each characteristic, and its cumulative value is I) (coalescence ('! (l?+, Q
is Q=w, σ force'10w, g north'.

The tower characteristic value Q obtained here takes a value between 0 and 1. The correspondence relationship between this composite characteristic value Q and the filter function number F can be set arbitrarily, but in general, it is an increasing function as shown in the time function between the filter relationship and the characteristic value in Fig. 4(b). Become.

The reason for this will be explained with reference to the sketch image 49 with feet. In the foot sketch image 49, in an area with a complicated pattern such as the bone culvert part 401, the composite characteristic value Q has a high value, so the filter function 41 is applied to emphasize the edges. On the other hand, in an area with a flat pattern such as the area outside the bone area (low density value) 402, the composite characteristic value Q is a small value, so 44 filter functions are applied, resulting in an image close to the original image.

As a result of the above processing, in the image of the foot attachment shown in the sketch image 49 of the foot attachment, for example, the trabecular bone part 401 in the bone part, willow ■1;
Areas with distinct patterns are most emphasized.

A region with a flat pattern like outside the bone 402\ can create an image that is almost the same as a surface.

Further, the composite characteristic value Q can be made into a single characteristic value by, for example, setting the sag w2 to O.

Next, process Ii! The image addition coefficient setting process 26 will be explained. The processed image addition coefficient setting process 26 is to compensate for the fact that it is generally said that the image after the filtering process 23 lacks volume.

Now, when the original image is G, the filter function is F (Q), and the image after filtering processing 23 is G', G' = G
There is a relationship of *F (Q,).

Further, if the image after the original image addition process 25 is G N, then a' = G + β (Q) · G' is calculated as G'. Here, β(Q) is an emphasis coefficient using the above-mentioned composite characteristic value Q as a variable.For example, as shown in FIG. 5, it is set by taking Q on the horizontal axis and β(Q) on the vertical axis.6 In Figure 5, the case where the value of β(Q) increases the same as the value of Q increases is 53, and the case where it increases gradually is 52.
, it shows 54 when it increases rapidly. The image after the original image addition process 25 has a large G' component in areas with a large composite characteristic value Q, that is, edge-enhanced areas, and a large G component in flat areas, resulting in an image that is sharp and has a sense of volume. It becomes possible to obtain a certain image.

Next, the contrast conversion function selection process 28 will be explained. In the contrast conversion function selection process 28, the contrast conversion process 27 is performed on the processed image G# obtained above.

, specifically, the input concentration N is plotted on the horizontal axis as shown in FIG.
A conversion function is used in which the output density K is plotted on the vertical axis. For example, by using the characteristics of 5G, it is possible to further emphasize differences in density changes in highlighted areas (large areas of the density field).

In the processes described above, there are parameters to be determined for each process, such as the filter function in Figure 4(a) and the correspondence function between the composite characteristic value Q and the filter function number F in Figure 4(b). . Further, these parameters should be changed depending on the diagnosis site, diagnosis purpose, etc.

FIG. 7 shows an example of the a priori known information 201. In FIG. 7, 61 indicates that the site is the coccyx and the purpose of diagnosis is periostitis. fc indicates the cutoff frequency of the filter. In this example, io is between 0.05 and 0.5 (cycles/
Bypass filter (characteristics in Figure 4 (,)) in the range of (field)
The characteristics of the emphasis coefficient β(Q) are expressed as 53 in Figure 5.
The contrast conversion is set as shown in FIG. Also,
62 shows an example of four parts, and fc is 0.01 to 0.
．． Using a plurality of bypass filters in the range of 6 (recycle/+nm), the characteristic of 1 emphasis coefficient β (Q> is set as 52 or 54 in FIG. 5, and the contrast conversion is performed using the 6th filter.
This indicates that the highlighted portion shown in 56 in the figure is emphasized.

Note that if such a priori known information 201 is registered in advance or set interactively by the operator, it is possible to reduce the amount of parameter input by the operator.

[Benefits of invention]

According to the present invention C.2, the optimal frame for each partial region of the image,
Since it is possible to select a filter, apply 1, and add the processed image with any white, it is possible to create an image with a sharp and cool feel, which improves the degree of ``ra'' in diagnosis. effective.

[Brief explanation of drawings]

Figure 1 is a sketch of an X-ray image of a human foot bone, Figure 2 is an explanatory diagram showing the processing of the present invention (four thoughts), Figure 3 is a diagram showing a method for extracting characteristic values of partial regions, and Figure 4 is Figure 5 is a diagram showing the filter function, Figure 5 is a diagram showing the processed image addition coefficient, Figure 6 is a diagram showing the contrast, l- & conversion rxg number, and Figure 7 is a diagram showing an example of specifying a priori known information. , FIG. 8 is a configuration diagram showing an example of an image processing system that is an embodiment of the present invention,
FIG. 9 is a functional diagram of the image processing system of FIG. 8. 24...Filter function selection process;! 6-... Processed image addition coefficient setting process, 28... Contrast conversion function selection process +1. 2 (11... A priori known information. Agent Me1-Ri 1: High (;6 Akio 1 Figure Il Figure 2 Tomi 3rd Eye 4th Figure ridge) (Right '5'1 ■ 5 Mouth fJ l Fig. 7 Fig. ■ 8 Fig.

Claims (1)

[Claims] In an image processing system consisting of an image input section, an image data processing section, and an image display section, an image is divided into partial images, characteristics are extracted for each partial image, and characteristics are extracted using a priori known information. An image enhancement method in an image processing system, characterized in that a relational expression according to characteristics is selected, and enhancement processing is performed for each partial image using the selected relational expression.