(19) Japan Patent Office (IP)
`
`(11) Japanese Unexamined Patent
`Application Publication Number
`
`(12) Japanese Unexamined Patent
`Application Publication (4)
`Identification codes
`JPO file numbers
`(43) Publication date September5, 1985
`6619-5B
`7033-4C
`7313-5B
`7923-5C
`
`$60-171573
`
`Request for examination Not yet requested Number of inventions 1
`
`(Total of 6 pages)
`
`Ot Int. cl
`GO06F
`A6IB
`GOOF
`GO09G
`
`15/68
`6/00
`15/42
`1/00
`
`(54) Title of the invention
`
`(72) Inventor
`
`(21) Japanese Patent Application
`(22) Date ofapplication
`Koichi MorisHItTa
`
`IMAGE HIGHLIGHTING METHOD FOR IMAGE PROCESSING SYSTEM
`$59-27169
`
`February 17, 1984
`Hitachi, Ltd. Systems Development Laboratory
`1099 Ozenji, Asao-ku, Kawasaki-shi
`Hitachi, Ltd. Systems Development Laboratory
`1099 Ozenji, Asao-ku, Kawasaki-shi
`Hitachi, Ltd. Systems Development Laboratory
`1099 Ozenji, Asao-ku, Kawasaki-shi
`Hitachi Medical Corporation Research and Development
`Laboratory, 2-1 Shin-Toyofuta, Kashiwa-shi
`4-6 Kanda-Surugadai, Chiyoda-ku, Toky6-to
`1-1-14 Uchi-Kanda, Chiyoda-ku, Toky6-to
`and | other
`
`Conventionally, since it is difficult to identify le-
`sions in images taken using X-ray devices and the
`like (original images), filter processing is performed
`on the original image. If the original image is G and
`the image after filter processing is G‘, this can be
`expressed in real space as
`G=F*G,
`In this expression, F represents the filter function and
`* represents convolution.
`Different filter functions F are used for high-pass
`highlighting, in-band highlighting, and low-pass hig-
`hiighting.
`Conventionally, when processing a single image, a
`filter function F with the same characteristics has
`been used. If processing is performed using a filter
`function F with the same characteristics as the single
`image as a whole, the following problemarises.
`Figure | is a sketch of an X-ray image of the bones
`of the foot. In Figure 1,
`11 through 13 represent
`bones. Moreover, region 14 is a smooth pattern and
`region |5 is a complex pattern with variations. Thus,
`for instance, if a high-pass highlighting filter is used
`to highlight the pattern of region 15, noise will gen-
`erally become noticeable in region 14. Conversely, if
`
`(72) Inventor
`
`Tetsuo YOKOYAMA
`
`(72) Inventor
`
`Nobutake YAMAGATA
`
`(72) Inventor
`
`Yoshihiro GoTo
`
`(71) Apphcant
`(71) Applicant
`(74) Agent
`
`Hitachi, Ltd.
`Hitachi Medical Corporation
`Patent Attorney Akio TAKAHASHT
`
`SPECIFICATION
`TITLE OF THE INVENTION: Image highlighting
`method for image processing system
`SCOPE OF PATENT CLAIMS
`An image highlighting method for an image
`processing system in an image processing system
`consisting of an image input unit, an image data
`processing unit, and an image display unit, characte-
`rized in that an image is divided into partial images,
`the characteristics of each partial image are extracted,
`a relational expression is selected according to said
`extracted characteristics using preexisting known
`information, and highlighting processing is per-
`formed on eachpartial image using said selected rela-
`tional expression.
`DETAILED DESCRIPTION OF THE INVENTION
`
`(FIELD OF USE OF THE INVENTION)
`The present invention relates to image highlighting
`methods for image processing systems, and specifi-
`cally to image highlighting methods fee4+meses—ete-
`in image processing systems for images, etc. taken by
`X-ray devices.
`(BACKGROUND OF THE INVENTION)
`
`— 403 -
`
`

`

`Japanese Unexamined Patent Application Publication S60-171573 (2)
`
`an in-band emphasis filter is used to highlight region
`14, the image will become bhurred in region 15. Thus,
`it is not possible to effectively display the informa-
`tion contained in the image across the image as a
`whole.
`To counteract this problem, Japanese Unexamined
`Patent Application Publication $55-87953 discloses
`non-sharp mask processing in order to improve the
`diagnostic capabilities of X-ray photographs.
`If D’ is the reproduced image, D,,, is the original
`image, D,, is the low-frequency highlighted image,
`and / is a constant, the reproduced image D’ can be
`expressed as
`‘= Dag + B Dag ~ D,.).
`In this expression the constant f is varied, for in-
`stance, according to the density value of D,,. With
`this method, it is possible to vary the additive propor-
`tions of the original image D,,,. and the highlighted
`image (Do — Dys) according to the density value of
`the image. With this method,it is possible to vary the
`degree of highlighting according to the density value
`of the image, but, as shown in the sketch of the X-ray
`image of foot bones in Figure 1, the problem remains
`that it is not possible to perform processing to select a
`filter function for highlighting according to the struc-
`tural characteristics of each partial region.
`(PURPOSE OF THE INVENTION)
`The purpose of the present invention is to provide
`an image highlighting method for
`an
`image
`processing system that eliminates the problems de-
`scribed above and allows for effective image hig-
`hlighting processing across an entire image.
`(SUMMARYOF THE INVENTION)
`The present invention, in order to achieve the pur-
`pose described above, is characterized in that, in an
`image processing system consisting of an image input
`unit, an image data processing unit, and an image
`display unit, a single image is divided into partial
`images, the characteristics of each partial image are
`extracted, and images with highlighting methods that
`differ depending on the diagnostic site, diagnostic
`purpose, etc. are obtained by picking filter functions,
`additive proportions for processed images, contrast
`conversion functions, etc. according to said extracted
`characteristics using preexisting known information.
`(EXAMPLE OF EMBODIMENTOF THE INVENTION)
`Below, an example of embodiment of the present
`invention will be described in detail using the draw-
`ings.
`Figure 8 is a configuration drawing showing an
`example of an image processing system that
`is an
`example of embodiment of the present invention.
`
`800 is an image processing unit that processes im-
`age data, 801 is a film reader that acts as an image
`imput unit, 802 is a display with a keyboard or the
`like that acts as an image display unit, 803 is a film
`reader that acts as an image output unit, and 804 ts an
`optical disk unit.
`Moreover, Figure 9 is a functional block diagram
`of the image processing system of Figure 8.
`The input X-ray image 900 is input into the image
`processing unit 800 by the film reader 801. In the
`image processing unit 800, processing is performed
`by the characteristic quantity extraction unit 902 to
`ascertain the characteristics, processing is performed
`by the similarity calculation unit 903 based on the
`characteristic quantities obtained, processing is per-
`formed by the selective filtering unit 904 to highlight
`specific frequency components based. on the dialogue
`mode, and processing is performed by the contrast
`conversion unit 905 to further highlight density varia-
`tion. The processing results
`from the
`tmnage
`processing unit 800 are output by the film writer 803
`as a processed X-ray image 920.
`Figure 2 is an explanatory diagram showing an
`outline of pressing in the present
`invention. Small
`regions are set for each part of the original image
`data 21 input from the film writer 801 and characte-
`ristic extraction processing 22 is performed on each
`partial region to extract characteristics such as va-
`riance values, mean values, etc.
`Next, according to these extracted characteristics,
`the optimal processing is selected from the following
`types of processing using preexisting knowninforma-
`tion 201, for instance, the diagnostic site, the diag-
`nostic purpose, etc. Types of processing include fil-
`tering processing 23 to highlight specific frequency
`components, source image addition processing 25,
`and contrast conversion processing 27.
`Also, using preexisting known information 201,
`filter function selection processing 24 of the filter to
`be used is performed for filtering processing 23,
`processed image addition factor setting processing 26
`is performed for source image addition processing
`25,
`and contrast
`conversion function selection
`processing 28 is performed for contrast conversion
`processing 27.
`Figure 3 shows the characteristic value extraction
`methods for each partial region. Partial regions 31,
`32, etc. are set for each pixel in source image 30 and
`the partial characteristics are extracted for each. For
`instance, if the size of partial region 31 is / x / and the
`pixel density value is g,, the variance value o and the
`mean value g can be found.
`
`— 404 —
`
`

`

`Japanese Unexamined Patent Application Publication S60-171573 (3)
`
`{see source for formulae]
`
`[see source for formulae}
`
`In these formulae, 7 and 7 denote the distance val-
`ues in the x-axis direction and the y-axis direction,
`respectively, of the coordinates(x, y).
`When setting regions, rather than rectangular re-
`gions, regions can also be set to be meaningful with
`regard to organs or thelike.
`Moreover, in addition to the previously mentioned
`statistical values, differential values, co-occurrence
`matrices, power spectra, etc. can of course also be
`used as characteristic values. In the following expla-
`nation, variance values and mean values are used as
`the characteristic values in order to simplify the dis-
`cussion.
`First, filter function selection processing 24 will be
`described based on Figure 4 (a), (b), and (c).
`In figure 4, (a) shows filter functions, (b) shows
`correspondence functions between the filter functions
`and compound characteristic values, and (c) shows a
`sketch 49 of foot bones. In the filter functions in Fig-
`ure 4 (a), the horizontal axis shows the cutoff fre-
`quencies f, and the vertical axis shows strength S,
`with multiple examples are shown in advance. Cha-
`racteristics can be set as desired, by in this example
`high-pass highlighting characteristics are shownas an
`example. As the filter function shifts from 44 to 41,
`the cutoff frequency f{ rises and so the strength S
`becomes greater, producing greater high-pass hig-
`hlighting characteristics.
`Moreover, in the correspondence function for the
`filter function and the characteristic values in Figure
`4 (b), the horizontal axis shows the compound cha-
`racteristic value Q and the vertical axis shows the
`number F ofthe filter function of Figure 4 (a), from
`41 through 44, indicating the filter function.
`The compound characteristic value Q can be found,
`for imstance, as follows. If the variance value of the
`partial region 31 of the same size / < / into which the
`source image 30 is divided into by a mesh is o, the
`maximumvariance value of each partial sed-example
`region of the source image 30 is o,,,, the minimum
`value iS Onin, the mean value of the partial region of
`the source image 30 is similarly g), the maximum
`mean value is Znex, and the minimum mean value
`iSmin, the normalized variance value o/ and mean
`value g/ can be found as follows:
`
`Next, the normalized weights w, and wz Gv) +42 = 1)
`are added to each characteristic, and the compound
`characteristic value Q becomes the cumulative value
`Q=wya/ + Wy gy.
`The compound characteristic value Q found here
`takes the form of a value from 0 to |. The correspon-
`dence function between this compound characteristic
`value Q and the number F ofthe filter function can
`also be set as desired, but generally it takes the form
`of an increasing function, as shown in the correspon-
`dence function for the filter function and characteris-
`tic values shownin Figure 4 (b).
`The reason for this will be explained using the
`sketch 49 of foot bones. In the sketch 49 of foot
`bones, in regions with complex patterns like the bone
`trabecular area 401, since the compound characteris-
`tic value Q takes a high value, filter function 41 is
`applied, highlighting edges. Conversely,
`in regions
`with even patterns like area outside the bone (with
`lowdensity) 402, the compound characteristic value
`Q takes a low value, so filter function 44 is applied,
`producing an image close to the source image.
`With the processing described above, with the foot
`bone image shown in the sketch 49 of foot bones, it is
`possible to create an image in which the regions with
`the most complex patterns within the bone, such as
`the trabecular area 401 are highlighted most strongly,
`while the regions with even patterns like area outside
`the bone 402 vary hardly at all from the source im-
`age. Moreover, it is also possible to set the compound
`characteristic value Q to represent a single characte-
`ristic, for instance, by setting weight w, to zero.
`Next, the processed image addition factor setting
`processing 26 will be described. Processed image
`addition factor setting processing 26 is generally per-
`formed
`to
`supplement
`images
`after
`filtering
`processing 23, since such images are considered poor
`in volume sensitivity. If the source image is G, the
`filter function is F (Q), and the image after filtering
`processing 23 is G’, the functionis as follows:
`G'=G*F(Q)
`Furthermore, if the image after source image addi-
`tion processing 25 is G”, G” is found using the fol-
`lowing:
`
`— 405 -
`
`

`

`Japanese Unexamined Patent Application Publication S60-171573 (4)
`
`mat by the operator in order to reduce the parameter
`imput workload of the operator.
`(EFFECT OF THE INVENTION)
`With the present invention, since optimal filters are
`selected for each partial region of the image and the
`processed image can be added in the desired propor-
`tion, it has the effect of making it possible to create
`images that are sharp and volume-sensitive, tmprov-
`ing diagnostic precision.
`BRIEF DESCRIPTION OF THE DRAWINGS
`Figure 1 is a sketch of an X-ray image of a human
`foot bone area, Figure 2 is an explanatory diagram
`showing an outline of the processing of the present
`invention, Figure 3 is a diagram showing the charac-
`teristic extraction method for partial regions, Figure 4
`is a diagram showing filter functions, Figure 5 is a
`diagram showing processed image addition factors,
`Figure 6 is a diagram showing contrast conversion
`functions, Figure 7 is a diagramshowing examples of
`the designation of the preexisting known information,
`Figure 8 is a configuration diagram showing an ex-
`ample of an image processing system that is an ex-
`ample of embodiment of the present invention, and
`Figure 9 is a functional block diagram for the image
`processing system of Figure 8.
`24 ... Filter function selection processing, 26 ...
`Processed image addition factor setting processing,
`28
`Contrast
`conversion function
`selection
`processing, 201... Preexisting known information
`Agent: Patent Attorney Akio TAKAHASHI
`[seal iegible]
`
`G"=G+6(QG'
`In this expression, # (Q) ts the highlighting factor,
`taking the previously described compound characte-
`ristic value Q as a variable, so, for instance, it can be
`set as shown in Figure 5, with Q on the horizontal
`axis and # (Q) on the vertical axis.
`Figure 5 shows a case where the value of f (Q)
`increases along with an increase in the value of Q as
`53, a case where it gradually increases as 52, and a
`case where it rapidly increases as 54. In regions with
`large compound characteristic values Q, i.e., i areas
`with edge highlighting, the G’ component is promi-
`nent
`in the image after
`source image addition
`processing 25, while the G component is more prom-
`inent in even areas, making it possible to obtain an
`image with sharp volumesensitivity.
`selection
`Next,
`contrast
`conversion
`function
`processing 28 will be explained. In contrast conver-
`sion function selection processing 28, contrast con-
`version processing 27 is performed on the processed
`image G” obtained as described above. Specifically, a
`conversion function taking the input density N and
`the output density K, as shown onthe horizontal axis
`and vertical axis, respectively, in Figure 6,
`is used.
`For instance, differences in density in highlighted
`areas (areas with high densities) can be further hig-
`hlighted by using the characteristics of 56.
`In the processing described above, there are para-
`meters that should be set for each type of processing,
`including the filter function in Figure 4(a), the cor-
`respondence function between the compound charac-
`teristic value Q and the filter function number F in
`Figure 4 (b), etc. Moreover, these parameters should
`be varied depending on the diagnostic site, the diag-
`nostic purpose,etc.
`These parameters are set using the preexisting
`known information 201 in Figure 2. Figure 7 shows
`an example of preexisting known information 201. In
`Figure 7, 61 shows a case where the site is the foot
`bones and the diagnostic purpose is periostitis. f, in-
`dicates the filter cutoff frequency. In this example
`multiple high-pass filters (characteristics in Figure 4
`(a))
`are used within the f£ range 0.05 to 0.5
`(cycles/mm),
`the characteristics of the highlighting
`factor # (Q) are set as shown in 53 in Figure 5, and
`highlight area highlighting is performed as shown in
`56 in Figure 6 as contrast conversion. Moreover, 62
`shows an example in the stomach area, where mul-
`tiple high-pass filters are used within the ( range 0.01
`to 0.6 (cycles/mm),
`the characteristics of the hig-
`hlighting factor / (Q) are set as shown in 52 or 54 in
`Figure 5, and highlight area highlighting is performed
`as shownin 56 in Figure 6 as contrast conversion.
`This kind of preexisting known information 201
`may be previously recorded or set in a dialogue for-
`
`— 406 -
`
`

`

`Japanese Unexamined Patent Application Publication S60-171573 (5)
`
`FIGURE 2
`
`FIGURE |
`
`toxg
`
`FIGURE 3
`
`FIGURE 4
`
`hagdedededseemk
`
`ET
`
`— 407 -
`
`
`
`
`
`

`

`Japanese Unexamined Patent Application Publication S60-171573 (6)
`
`
`
`
`
`
`
` aprecerrererenssnneseNeeennnnnnnnnnen,rece,lbeneennsennsennnenney~
`
`nnnnmr
`
`FIGURE 5
`
`FIGURE 7
`
`TH
`
`
`
`EA
`
`IGURE
`
`9
`
`98esiAeeoe
`
`FIGURE 6
`
`6eaesweQBy
`
`— 408 -
`
`potwae?eeLoantt
`
`6,gARES8eta|me$oomSLEBOsoe(aEBM;
`
`
`
`
`
`
`
`
`
`