DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0050] Now, a description will be given of preferred embodiments of the present invention with reference to the accompanying drawings.

[0051] In a first embodiment, the present invention is applied to, for example, an x-ray sensing apparatus 100 shown in FIG. 1 .

[0052] Prior to the specific description of the x-ray sensing apparatus 100 in this embodiment, a characteristic function implemented by the x-ray sensing apparatus 100 will be described.

[0053] The x-ray sensing apparatus 100 is an apparatus that is so designed as to acquire a single image from the assembly of a plurality of segment images that are acquired by divisionally driving a sheet of sensor panel, and particularly has a function of selecting two adjacent segment images from the plurality of segment images, adjusting the offset between those segment images to form one new portion, and repeatedly executing such a process to unify these segment images into one final image in which the boundary is not visible.

[0054] Hereinafter, the characteristic function of the x-ray sensing apparatus 100 will be specifically described.

[0055] (1) First, a Substantial Step Value of the Boundary of the Segment Images is Acquired as Follows:

[0056] Paying attention to the one-dimensional direction of the image, it is assumed as shown in FIG. 2 that there exists a gap of g pixels between a segment image 1 and a segment image 2 adjacent to the segment image 1, and it is also assumed that the pixel value of the segment image 1 is x(0), x(1), . . . , x(n), and the pixel values of the segment image 2 is x(n+g), x(n+g+1), . . .

[0057] Also, because the segment image 1 and the segment image 2 are both segment images resulting from x-ray-sensing of the object to be sensed by divisionally driving a sheet of sensor panel, an error exists in the offset value between the pixel n of the segment image 1 and the pixel (n+g) of the segment image 2, and there is the possibility that the error is recognized as a step.

[0058] However, because both of the segment images are obtained by sensing the same object to be sensed, the pixel n of the segment image 1 and the pixel (n+g) of the segment image 2 have a relationship that the inclination is continuous. Therefore, the average of an inclination obtained from the pixel values x(0), x(1), . . . , x(n) of the segment image 1 and an inclination obtained from the pixel values x(n+g), x(n+2), . . . of the segment image 2 is now defined as an inclination “K” of the object image information.

[0059] A difference across the boundary of the pixel value x(i+p) of the segment image 1 and the pixel value x (i+p) of the segment image 2 is x(i+p)−(i). The inclination is the above-mentioned “K”, and an expected difference is “pK”, and a difference between the pK value and the value of x(i+p)−x(i) becomes an expected value (difference value) d of the substantial step. That is, the substantial step value d of the boundary of each of the segment images 1 and 2 is obtained by the following expression (1):

d=PK−x ( i+P )+ x ( i ) (1)

[0060] As a method of obtaining the step value d, there are various methods dependent on manners of obtaining a differential value besides the above-mentioned method. One example of those methods is described below.

[0061] For example, the inclination of the segment image 1 is obtained using the pixel value x(n) as a reference, and the inclination of the segment image 2 is obtained using the pixel value x(n+g) as a reference. The respective inclinations are averaged by using the pixel value of a point m to obtain the inclination K.

[0062] In this case, the inclination K is represented by the following expression (2). 1 $\begin{array}{cc}\begin{array}{c}K=\frac{1}{2m}[\sum _{p=1}^m\frac{1}{p}\left(x\left(n\right)-x\left(n-p\right)\right)+\\ \sum _{p=1}^m\frac{1}{p}\left(x\left(n+g+p\right)-x\left(n+g\right)\right)]\end{array}& \left(2\right)\end{array}$

[0063] Also, differences across the boundary of the pixel value x(i+p) of the segment image 1 and the pixel value x(i) of the segment image 2 are considered as symmetry with respect to the boundary so as to be x(n+g)−x(n), x(n+g+1)−x(n−1), x(n+g+2)−x(n−2), . . . , x(n+g+m)−x(n−m).

[0064] In this case, the substantial step value d of the boundary of each of the segment images is represented by the following expression (3). 2 $\begin{array}{cc}d=\frac{1}{m}\sum _{q=0}^m\left\{\left(g+2q\right)K-x\left(n+g+q\right)+x\left(n+q\right)\right\}& \left(3\right)\end{array}$

[0065] The substantial step value d on a line that is cross one boundary can be obtained through the above-mentioned calculation process. Also, the calculation process is intensive to the calculation process of accumulative addition that makes accumulation by substantially multiplying the pixel value by a predetermined coefficient.

[0066] Here, because a plurality of lines (0 to L) exist in the image, there is actually obtained the substantial step value line d(i) (i=0 to L) which is continuous on one boundary. Also, because a plurality of boundaries between one segment image and other adjacent segment images exist in the segment image, a plurality of step value lines d(i) exist for one segment image.

[0067] In this example, a change in the pixel value with respect to the segment image (object segment image) to be processed is statistically obtained from such a step value line d(i).

[0068] Various methods are applicable to this processing method depending on the modes of the object segment image and should be used case by case.

[0069] As one example thereof, for example, in the case where a single sensor panel is divided into four segments and driven, as shown in FIG. 3, four segment images A, B, C and D are connected in the form of a matrix. In this example, the boundaries are represented by four boundaries a, b, c and d shown in FIG. 3. As to the respective boundaries a, b, c and d, it is assumed that the step values obtained with the above-mentioned method are d0(0 to L−1), d1(0 to (L−1)), d2(−1 to −(L−1)) and d3(−1 to −(L−1)).

[0070] In this example, the image center is an origin, and the position is represented by “±”.

[0071] In this case, if the data of the segment images A, B, C and D is in proportion to the x-ray intensity, the step values d0(0 to L−1), d1(0 to (L−1)), d2(−1 to −(L−1)) and d3(−1 to −(L−1)) correspond to the offset value. Therefore, in this case, with the execution of the logarithmically converting process, the step values d0(0 to L−1), d1(0 to (L−1)), d2(−1 to −(L−1)) and d3(−1 to −(L−1)) correspond to a variation in the gain.

[0072] As a correcting method using the step values d0(0 to L−1), d1(0 to (L−1)), d2(−1 to −(L−1)) and d3(−1 to −(L−1)), that is, the variation in the gain, there are proposed various methods depending on the situation, and as one example thereof, a correcting method which conducts the correction to the partial images A, B, C and D with addition/subtraction of a constant value, will be described below.

[0073] In the case of using the constant value, the step values d0(0 to L−1), d1(0 to (L−1)), d2(−1 to −(L−1)) and d3(−1 to −(L−1)) at the boundaries a, b, c and d are consolidated into one certain value. For that reason, values D0, D1, D2 and D3 are derived from the step values d0(0 to L−1), d1(0 to (L−1)), d2(−1 to −(L−1)) and d3(−1 to −(L−1)) one by one, respectively.

[0074] As a method of deriving the values D0, D1, D2 and D3, there is a method in which the respective average values of the step values d0(0 to L−1), d1(0 to (L−1)), d2(−1 to −(L−1)) and d3(−1 to −(L−1)) are obtained as the values D0, D1, D2 and D3. However, if correction is not conducted averagely but conducted in a range as wide as possible, for example, as shown in FIG. 4A, a maximum frequency value (the peak of histogram) is used as the values D0, D1, D2 and D3.

[0075] It is needless to say that the following expression is satisfied:

D 0 +D 1 +D 2 +D 3=0

[0076] However, due to an error in calculation, an influence of image noises, an unstable offset value and so on, it is necessary to generally apply the following expression:

D 0 +D 1 +D 2 +D 3=ε(≠0)

[0077] This is an inconsistency resulting from obtaining the respective values independently.

[0078] In order to eliminate the above inconsistency, first, correction is made so as to satisfy a correction value F0=0 and a correction value F3=−D0 by using the step value D0 in FIG. 4A . As a result, the partial image A and the partial image D are regarded as one partial image AD without any step. In this situation, there is a case in which a smaller value is referenced so that a negative value is not added.

[0079] Also, correction is made so as to satisfy F2= and F5=−D4 in the same manner, as a result of which a step of the partial image B and the partial image C is eliminated, and the partial image G and the partial image C are regarded as one partial image BC.

[0080] Therefore, as shown in FIG. 4B, a linear boundary is defined between the partial image AD and the partial image BC. Then, in that boundary, a step value line d4(−L to L) is newly calculated, thereby obtaining the correction value D4 as its representative value.

[0081] Then, the correction value F4=0 and the correction value F5=−D4 are newly added to the partial image AD and the partial image BC.

[0082] Then, the entire image is structured by the partial image AD and the partial image BC as shown in FIG. 4C . As a result, the inconsistency caused by only one calculation of a step value is prevented, and the partial images are synthesized without any inconsistency.

[0083] In the case where three or more partial images exist, the steps are corrected with respect to each of two partial images which are regarded as independent partial images, to constitute one new partial image, and the calculation of the step value with respect to the new partial image is repeatedly executed.

[0084] As described above, in the process of putting those two partial images into one partial image, without correcting only the predetermined step value, the step value sequence are fitted by a function with a low degree of freedom, thereby being applicable to a case in which the interior of the partial images are not a constant offset error.

[0085] That is, in FIG. 3 , assuming that the axis of ordinate is “y” and the axis of abscissas is “x”, and the step value sequence d0(0 to L−1) and d2(0 to L−1) are fitted by the function that is small in the degree of freedom, respectively, the statistic step values are obtained. It is assumed that the functions at this time (the function that conducts the fitting) are d0′(y) and d2′(y), and the center of the entire image is an origin.

[0086] In this case, because the offset is constant over the entire surface, and there is no partial image of “0”, that is, there exists no partial image to be referenced, correction is made with respect to all of the partial images. Also, if correction that is not an even plane is added (shirring is conducted) with respect to only any one partial image, because there is a risk of the image density being greatly deformative, the correction is divided to the respective partial images half by half.

[0087] In FIG. 3 , the step values are shared to the respective faces such that the correction value when viewing the partial image A from the boundary a is “d0′(y)/2”, and the correction value when viewing the partial image D is “−d0′(y)/2”.

[0088] As a result, the function F0′(x,y) used when correcting the partial image A becomes a two-dimensional function, which is represented by the following expression:

F 0′( x,y )= d 0′( y )/2 (4)

[0089] Similarly, the functions of other partial images are represented by the following expressions (5) to (7).

F 3′( x,y )=− d 0′( y )/2 (5)

F 1′( x,y )=− d 2′( y )/2 (6)

F 3′( x,y )= d 2′( y )/2 (7)

[0090] The functions for correction is obtained in the above-mentioned manner to correct the partial images by adding the function values to the respective partial images. This correction enables the new partial image AD and the partial image BC to be obtained even if the offset steps of the partial images A, B, C and D are not constant within the partial images.

[0091] Under the above circumstances, as shown in FIG. 4 B, the substantial step value sequence d4(−L to L) are calculated with respect to the boundary e between the new partial image AD and the partial image BC and then subjected to the function fitting, to thereby obtain the statistic step value d4′(x). Then, F4′(x,y) is added to the new partial image AD, and F5′(x,y) is added to the partial image BC, to thereby eliminate the step between the partial image AD and the partial image BC and produce one image.

[0092] In this situation, F4′(x,y) and F5′(x,y) are represented by the following expressions (8) and (9):

F 4′( x,y ) d 4′( y )/2 (8)

F 5′( x,y ) −d 4′( y )/2 (9)

[0093] The above-mentioned process is a correcting process for the two-dimensional image, but in the correcting process, a one-dimensional function is employed.

[0094] Incidentally, generally, when a variation which is not constant within the partial image is corrected in the case where the partial images are coupled together, it is preferable that the boundary portion of the partial image is linear in a certain direction (x-direction or y-direction).

[0095] For example, the boundaries a and b shown in FIG. 3 are the boundaries of the partial image A, and even if other partial images B, C and D are coupled together so as to be regarded as a new partial image BCD as shown in FIG. 5 , it is necessary to process the boundaries a and b as a series of steps. For that reason, it is necessary to fit the steps by one function or to fit the respective steps taking into consideration the bonding condition that makes two functions continuous. In addition, the two-dimensional function is structured by the above fitting and must be converted to a correction value to the partial image A. Also, as shown in FIG. 5 , in the boundaries a and b in the different directions, the step series different in shape is liable to be produced due to the characteristic of the sensor.

[0096] Accordingly, the partial image are sequentially selected so as to prevent the shape shown in FIG. 5 , and the selected partial images thus obtained are sequentially unified. Then, when two partial images are finally obtained, their boundary is made linear.

[0097] Then, a process for not giving a user (observer) a sense of incongruity that the single entire image are the collection of a plurality of partial images A, B, C and D, is executed as follows:

[0098] First, because the above process is a process for the vicinity of the boundaries a, b, c and d of the partial images A, B, C and D, there is applicable a method in which the components having the characteristic that is continuous in the boundary direction in the vicinity of the boundary are extracted with respect to the boundaries a, b, c and d, and the above characteristic components are removed in the vicinity of the boundary to make the boundaries a, b, c and d unvisible as disclosed in, for example, Japanese Patent Application Laid-open No. 2000-132663 or the like. However, there is a case in which a problem arises in a specific image.

[0099] For example, FIG. 6A schematically shows a state where attention is paid to two partial images (partial image 1 and partial image 2) among the partial images A, B, C and D, and the substantial boundary step of those partial images remain.

[0100] In FIG. 6 A, reference numeral 51 denotes a section of the partial image 1, and 52 denotes a section of the partial image 2. Those lines 51 and 52 is provided with the moving average of some degree in the boundary direction and show the shapes of the steps.

[0101] FIG. 6B shows a state of the boundary in the case where the components having the characteristic continuous in the boundary direction in the vicinity of the boundary are extracted with respect to the boundary of the partial image 1 and the partial image 2 to remove the characteristic components in the vicinity of the boundary so as to make the boundary unvisible through a method disclosed in Japanese Patent Application Laid-open No. 2000-132663 or the like.

[0102] In FIG. 6 B, assuming that there is no fine image variation in parallel with the boundary direction, a repair is made as indicated by 53 . However, as shown in FIG. 6 B, because the fine variation remains in the moving-averaged data in the boundary direction, the fine image variation occurs in parallel with the boundary direction, and information on that portion is lost.

[0103] FIG. 6C shows a state where a smooth (linear in the figure) correction component is added to only W front and rear pixels of steps, as shown by “54” of FIG. 6 C, and the partial image 1 and the partial image 2 are connected to each other so as to hold the fine image variation information in a direction parallel with the boundary direction.

[0104] This method will be described in more detail below.

[0105] First, a process of obtaining the substantial step value d of the boundary by using the above-mentioned expression (3) is again executed with respect to the respective boundaries that have been subjected to the correcting process once. As a result, the step series indicated by d0(0 to L−1), d1(0 to (L−1)), d2(−1 to −(L−1)) and d3(−1 to −(L−1)) are obtained again.

[0106] This step series indicated by d0(0 to L−1), d1(0 to (L−1)), d2(−1 to −(L−1)) and d3(−1 to −(L−1)) should be “0” on the average, but the steps partially remains, and an image noise is added thereto.

[0107] Therefore, the moving averages of the step series d0(0 to L−1), d1(0 to (L−1)), d2(−1 to −(L−1)) and d3(−1 to −(L−1)) are calculated, and its results are set as dm0(0 to L−1), dm1(0 to (L−1)), dm2(−1 to −(L−1)) and dm3(−1 to −(L−1)).

[0108] Note that the same results are obtained even if the difference operation is conducted after the moving averages are calculated. That is, the process does not depend on the calculation order.

[0109] In the boundary a shown in FIG. 4A, A 0(x,y) indicated by the following expression (10) is added to data of the partial image A:

A 0( x,y )= dm 0( y )(− x/W+ 1)/2 (10)

[0110] where x is 0 to W

[0111] Also, D0(x,y) represented by the following expression (11) is added to the partial image D:

D 0( x,y )= dm 0( y )( x/W+ 1)/2 (11)

[0112] where x is −1 to W

[0113] Also, similarly, in the boundary b shown in FIG. 4A, A 1(x,y) indicated by the following expression (12) is added to data of the partial image A:

A 1( x,y ) 32 dm 1( x )(− y/W+ 1)/2 (12)

[0114] where y is 0 to W

[0115] Also, B1(x,y) represented by the following expression (13) is added to the partial image B:

B 1( x,y )= dm 1( x ) ( x/W+ 1)/2 (13)

[0116] where y is −1 to W

[0117] Also, similarly, in the boundary c shown in FIG. 4A, B 2(x,y) indicated by the following expression (14) is added to data of the partial image B:

B 2( x,y )= dm 2( y ) (− x/W+ 1)/2 (14)

[0118] where x is 0 to W

[0119] Also, C2(x,y) indicated by the following expression (15) is added to the partial image C:

C 2( x,y )= dm 2 ( y ) (x/W+1)/2 (15)

[0120] where x is −1 to W

[0121] Also, similarly, in the boundary d shown in FIG. 4A, C 3(x,y) indicated by the following expression (16) is added to data of the partial image C:

C 3( x,y )= dm 3( x )(− y/W+ 1)/2 (16)

[0122] where x is 0 to W

[0123] Also, B3(x,y) indicated by the following expression (17) is added to the partial image B:

B 3( x,y )= dm 3( x )( y/W+ 1)/2 (17)

[0124] where x is −1 to W

[0125] As described above, the respective functions of A0(x,y), A1(x,y), B1(x,y), B2(x,y), C2(x,y), C3(x,y) and D3(x,y) are added respectively to the partial images A, B, C and D to correct local steps to linear shapes, thereby being capable of eliminating the steps on the boundaries.

[0126] Note that the substantial step hardly damages the image per se because the inclinations of data of the partial images A, B, C and D are taken into consideration, but if the substantial step is an extremely large value, a device may be made so as not to correct the difference.

[0127] Further, the above-mentioned correction of the local steps may be conducted each time when a new partial image is structured by the above-mentioned coupling of the partial images.

[0128] The characteristic functions implemented by the x-ray sensing apparatus 100 according to this embodiment are described above. Hereinafter, the structure and operation of the x-ray sensing apparatus 100 will be described in more detail below.

[0129] The entire structure of the x-ray sensing apparatus 100 :

[0130] As shown in FIG. 1 , the x-ray sensing apparatus 100 includes partial panels 101 a to 101 d , amplifiers 102 a to 102 d , A/D converters 103 a to 103 d , DMA controllers 104 a to 104 d , a CPU 109 , memories 105 to 108 , 110 and 116 , a memory unit 111 and an exposure control unit 112 .

[0131] A CPU 109 , the memories 105 to 108 , 110 and 116 , the high-capacity memory unit 111 , and the exposure control unit 112 are so connected as to communicate with each other through a bus 115 .

[0132] The sequential operation of the x-ray sensing apparatus 100 :

[0133] The CPU (central processing unit) 109 controls the operation of the entire x-ray sensing apparatus 100 .

[0134] For example, the memory 116 (program memory) stores processing program according to a flowchart of FIG. 7 therein in advance. The CPU 109 reads and executes the processing program from the program memory 116 to control the operation of the entire x-ray sensing apparatus 100 . As a result, the operation of the x-ray sensing apparatus 100 is executed as described below.

[0135] Step S 201 :

[0136] Upon radiation of x-rays onto an object 114 (a human body in this example), the exposure control unit 112 first controls an x-ray vessel so that the x-rays radiated from the x-ray vessel 113 pass through the object (human body) 114 and are inputted to each of the partial panels 101 a to 101 d.

[0137] Step S 202 :

[0138] The partial panels 101 a to 101 d are panels divided from a sheet of sensor panel into four segments, and are driven on the basis of a timing at which the x-ray vessel 113 generates the x-rays independently (sequential driving of switching transistors not shown).

[0139] As a result, electric signals corresponding to pixels are outputted from the respective partial panels 101 a to 101 d.

[0140] The respective amplifiers 102 a to 102 d receive the electric signals outputted from the corresponding partial panels 101 a to 101 d , amplify and output the electric signals.

[0141] The respective A/D converters 103 a to 103 d receive the electric signals outputted from the corresponding amplifiers 102 a to 102 d , digitalize and output the electric signals.

[0142] Step S 203 :

[0143] The respective DMA controllers 104 a to 104 d store the data (partial image data) outputted from the corresponding A/D converters 103 a to 103 d in the memory 105 (dual port memory).

[0144] The dual port memory 105 is so structured as to read corresponding partial image data with a string of addresses through the bus 115 .

[0145] Step S 204 :

[0146] The partial image data stored in the dual port memory 105 is stored in the memory 106 (frame memory) by the CPU 109 or a DMA controller (not shown) on the bus 115 .

[0147] Steps S 205 and S 206 :

[0148] Then, the same operation as the above is conducted in a state where no x-rays are irradiated by the x-ray vessel 113 . The partial image data thus obtained is stored in the memory 107 through the dual port memory 105 as the data of an offset-like fixed-pattern.

[0149] Steps S 207 and S 208 :

[0150] In this situation, information on the gain variation is stored in the memory 108 in advance.

[0151] More specifically, x-rays are exposed toward the respective partial panels 101 a to 101 d by the x-ray vessel 113 in a state where no object 114 exists, and the partial image data resultantly obtained is stored in the memory 108 as the information on the gain variation for each of the pixels in the partial panels 101 a to 101 d.

[0152] Thus, the CPU 109 subtracts the fixed pattern data stored in the memory 107 from the partial image data stored in the frame memory 106 , and a correcting process for each pixel of the entire image using the gain variation information stored in the memory 108 is executed with respect to the subtraction result, and the correction result is stored in the memory 117 .

[0153] The above-mentioned correcting process includes a dividing process, and the dividing process may be dealt with, for example, as a subtracting process that conducts logarithmic transformation using a look up table (LUT).

[0154] Also, the CPU 109 stores in the memory 118 information pertaining to the position of a pixel that cannot be corrected on the basis of the gain variation information (a pixel which is a defective pixel from which no data is obtained, etc.) in the above-mentioned correcting process.

[0155] The CPU 109 determines as a value of the defective pixel a pixel value analogized from the data of the peripheral pixels (pixels except for the defective pixels) of the pixels (defective pixels) indicated by the information stored in the memory 118 in the image data stored in the memory 117 which has been corrected. For example, the CPU 109 determines the average value of the peripheral pixels as a value of the defective pixel.

[0156] Step S 209 :

[0157] The image data that has been subjected to the above-mentioned process is stored in the memory 110 as the image data of the object 114 .

[0158] In this case, the image data may be over-written in the memory 117 without using the memory 110 . As a result, the memory can be effectively employed.

[0159] The image data may be saved in the memory unit 111 (a magnetic memory device, a large-capacity nonvolatile memory device or the like) as a file. Alternatively, the image data may be outputted to an external display device, recording device, memory device or the like by an interface (not shown).

[0160] In this case, the partial image data obtained independently is normalized by a process using the above-mentioned data stored in the memory 107 (offset-like fixed pattern data), data (gain variation information) stored in the memory 108 or the like, and the user cannot recognize that the data is related to the partial image. The characteristic operation of the x-ray sensing apparatus 100 :

[0161] FIG. 8 shows a state of the image data of the object 114 stored in the memory 110 through the above-mentioned operation.

[0162] As shown in FIG. 8 , the data of the respective partial images A, B, C and D corresponding to the partial panels 101 a to 101 d is stored in the memory 110 . Also, a gap of g pixels (hereinafter referred to as “gap g”) is defined between the images of the partial images A, B, C and D. The gap is a defective pixel in a sense, but in the case where the offset-like steps are produced between the respective partial images A, B, C and D, the average pixel correction is nonsense. In this embodiment, after the correction for each of the partial images, which will be described later, is conducted, the pixel is corrected with the average value.

[0163] More specifically, it is assumed that gap g is equal to 1, for example. The substantial step value d is obtained in accordance with the above-mentioned expression (3).

[0164] In the expression (3), assuming that m is equal to 3, the coefficient of the pixel value x(n−2) to x(n+4) is obtained in accordance with a table shown in FIG. 9 . In the table, a coefficient for obtaining the differentiation K at a value q and a coefficient for calculating the step value d are separated from each other, and finally summed.

[0165] Therefore, the substantial step value d is obtained by the following expression (18). 3 $\begin{array}{cc}\begin{array}{c}d\left(i\right)=\sum _{j=-2}^4c\left(j\right)x\left(n+j\right)\\ =\frac{1}{6}\left(x\left(n+4\right)-x\left(n-2\right)\right)+\\ \frac{2}{3}\left(x\left(n+3\right)-x\left(n-1\right)\right)+\frac{11}{6}\left(x\left(n\right)-x\left(n-2\right)\right)\end{array}& \left(18\right)\end{array}$

[0166] The substantial step value sequence d(i) are obtained with taking into consideration the inclination of the image data if the substantial step value d is converted into the one-dimensional data sequence that interpose the boundary x(n+1) of the partial image.

[0167] The “i” in the above-mentioned expression (18) is an index in a direction parallel with the boundary of the partial image.

[0168] FIGS. 10 and 11 show a correcting process using the above-mentioned step value sequence d(i). The correcting process is implemented by executing the processing program in accordance with flowcharts shown in FIGS. 10 and 11 which are stored in a program memory 116 in advance.

[0169] Steps S 301 and S 302 : See FIG. 10

[0170] Step value sequence d0(0) to d0(n) of the boundary a are obtained, and the step value sequence d2(n+2) to d2(M−1) of the boundary c are obtained, through the above-mentioned expression (18).

[0171] More specifically, for example, as shown in FIG. 12 , the pixel values are extracted across the boundary in accordance with the addresses of pixels indicated by “21” and then subjected to calculation process of the above-mentioned expression (12), to thereby obtain the difference value sequence d0(i). The above calculation process is executed with respect to all of the boundaries.

[0172] Steps S 303 and S 304 :

[0173] The step value sequence d0(0) to d0(n) obtained in Steps S 301 and S 302 and the respective maximum frequency values D0, D1 of the step value sequence d2(n+2) to d2(M−1) are obtained.

[0174] As D0 and D1, values that represent the distribution of the average values, the center values or the like may be used.

[0175] Step S 305 to Step S 307 :

[0176] The offset value that is added to the partial image A is set to “0”, and the maximum frequency value D0 obtained in Step S 303 is added to the partial image D that is opposite to the partial image A. In this case, the maximum frequency value D0 means the step value of the partial image D when viewed from the partial image A.

[0177] As a result, the substantial step between the partial image A and the partial image D is eliminated and can be regarded as a new partial image AD.

[0178] Step S 308 to Step S 310 :

[0179] The offset value that is added to the partial image B is set to “0” , and the maximum frequency value D1 obtained in Step S 304 is added to the partial image C that is opposite to the partial image B. In this case, the maximum frequency value D1 means the step value of the partial image C when viewed from the partial image B.

[0180] As a result, the substantial step between the partial image B and the partial image C is eliminated and can be regarded as a new partial image BC.

[0181] Therefore, two new partial images AD and BC shown in FIG. 4B are obtained. Then, those two partial images AD and BC are unified in the following manner.

[0182] Steps S 311 and S 312

[0183] The substantial step sequence d4(0) to d4(M−1) of the boundary e of the partial image AD and the partial image BC are prepared. Then, the maximum frequency value D2 is obtained.

[0184] Steps S 313 and S 314 :

[0185] The offset value that is added to the partial image AD is set to “0”, and the step value D2 obtained in Step S 312 is added to the partial image BC to unify the partial image AD and the partial image BC. Then, this result is stored in the memory 110 .

[0186] The image thus stored in the memory 110 results from conducting only the offset-like correction over the entire partial images A to D. However, in fact, the variations of the partial images A to D are not limited to the uniform offset, and there are many cases in which the above correction is insufficient.

[0187] Under the above circumstances, a process is executed in which only the vicinity of the boundaries of the image stored in the memory 110 is corrected so that a sense of incongruity that the entire image is the collection of a plurality of partial images is not given to an observer through process starting from Step S 322 which will be described below. As one example, the correction is conducted by a method shown in FIG. 6C .

[0188] Step S 322 to Step S 325 : See FIG. 11

[0189] The substantial step value sequence d0(0) to d0(n), a1(n+2) to d1(M−1), d2(n+2) to d2(M−1), d3(0) to d3(n) of the original boundaries a, b, c and d are obtained in the same manner as Step S 301 in FIG. 10 .

[0190] Step S 326 to Step S 329 :

[0191] The step value sequence d0(0) to d0(n), a1(n+2) to d1(M−1), d2(n+2) to d2(M−1), d3(0) to d3(n) obtained in Steps S 322 to S 325 are converted into smooth numeric sequence due to the moving average or the like. This is because the step value sequence d0(0) to d0(n), a1(n+2) to d1(M−1), d2(n+2) to d2(M−1), d3(0) to d3(n) are influenced by various noises.

[0192] Step S 330 to Step S 333 :

[0193] The local correcting process of the step components shown in the above-mentioned expressions (10) to (17) is executed.

[0194] That is, as to the original boundary a, the changed origin of the above-mentioned expressions (10) and (11) is used, and dm0(y)(−x/W+1)/2, wherein x: n+1 to n+1+W, y: 0 to n, is added to the partial image A, and dm0(y)(x/W+1)/2, wherein x: n−W to n, y: 0 to n, is added to the partial image D.

[0195] Similarly, as to the original boundary b, the changed origin of the above-mentioned expressions (12) and (13) is used, and dm1(x)(−y/W+1)/2, wherein x: n+1 to M−1, y: n−W to n, is added to the partial image A, and B1(x,y) =dm1(x)(y/W+1)/2, wherein x: n+1 to M−1, y: n+1 to n+1+W, is added to the partial image B.

[0196] Similarly, as to the original boundary c, the changed origin of the above-mentioned expressions (14) and (15) is used, and B2(x,y) dm2(y)(−x/W+1)/2, wherein x: n+1 to n+1+w, y: n+W to M−1, is added to the partial image B, and C2(x,y) dm2(y)(x/W+1)/ 2 , wherein x: n−W to n, y: n+1 to M−1, is added to the partial image C.

[0197] Similarly, as to the original boundary d, the changed origin of the above-mentioned expressions (16) and (17) is used, and C3(x,y) dm3(x)(−y/W+1)/2, wherein x: 0 to n, y: n−W to n, is added to the partial image C, and B3(x,y) dm3(x)(y/W+1)/2, wherein x: 0 to n, y: n+1 to n+1+W, is added to the partial image B.

[0198] Step S 334 :

[0199] After the local steps are corrected to linear shapes with respect to the partial images A, B, C and D so as to eliminate the steps on the boundaries through Steps S 330 to S 333 , the pixels of the boundary portion that finally remains are interpolated with pixels that are not the defective pixels in the vicinity of the boundary portion. The region of the interpolated pixels includes a region where only one pixel does not exist in the boundary portion, that is, a region that cannot be subjected to interpolation unless correction is conducted for each of the partial images up to now.

[0200] In this embodiment, the above-mentioned operation is realized by programming in a calculator system, but the present invention is not limited to this structure. For example, the above-mentioned operation may be realized by hardware, or may be realized by a structure that mixes hardware and program environments.

[0201] Also, the above-mentioned repair of the vicinity of the boundary may be executed for each of the processes shown in FIG. 10 .

[0202] Subsequently, a second embodiment of the present invention will be described below.

[0203] In this embodiment, in the case where the offset components of the partial image is not uniform within the partial image, the step value sequence is not represented by a fixed value (the maximum frequency value, the average value or the like), but the step value sequence between two partial images is fitted by a function of the low degree of freedom to prepare a step function, and the step function is divided into two and added to both of the partial images, to thereby unify two partial images.

[0204] Note that, in this embodiment, the step function is prepared, and the step function is divided into two and added to both of the partial images. However, for example, it is possible that the step function is added to one partial image, and a function resulting from reversing the sign of the step function is added to the other partial function.

[0205] FIGS. 13A to 13 C schematically show this embodiment.

[0206] First, as shown in FIG. 13 A, an non-uniform inclined offset is superimposed on four partial images A, B, C and D.

[0207] Then, as shown in FIG. 13 B, the substantial step value sequence in the boundary region between the partial image A and the partial image B is obtained. A function resulting from fitting the step value sequence is divided into two, and the two step value sequences whose signs are reversed to each other are added to both of the partial image A and the partial image D so as to eliminate the step value sequence. The same process is conducted on the partial image B and the partial image C.

[0208] Then, as shown in FIG. 13 C, the same process as the above is conducted on the boundary region of the new partial image AD and the partial image BC which are obtained in the process shown in FIG. 13 B, to thereby generate one image without any boundary between the partial image AD and the partial image BC.

[0209] FIG. 14 shows a process of this embodiment.

[0210] This process is implemented by executing the processing program, for example, by a CPU 109 in accordance with a flowchart shown in FIG. 14 which is stored in the program memory 116 in advance.

[0211] Step S 401 :

[0212] The step value sequence d0(0) to d0(n) of the boundary between the partial image A and the partial image D is obtained through the above-mentioned expression (18).

[0213] Step S 402 :

[0214] The step value sequence d2(n+2) to d2(M−1) of the boundary between the partial image B and the partial image C is obtained through the above-mentioned expression (18).

[0215] Step S 403

[0216] The step value sequence d0(0) to d0(n) 25 obtained in Step S 401 is fitted by a function f0(y,p1). The function f0(y,p1) is a function having a parameter p1 assuming that y is the longitudinal axis of the image.

[0217] Step S 404 :

[0218] The step value sequence d2(n+2) to d2(M−1) obtained in Step S 402 is fitted by a function f2(y,p2). The function f2(y,p2) is a function having a parameter p2 assuming that y is the longitudinal axis of the image.

[0219] Step S 405 :

[0220] In order to set the step function f0(y,p1) to substantially “0”, f0(y,p1)/2 is added to the partial image A, and −f0(y,p1)/2 is added to the partial image D.

[0221] Step S 406 :

[0222] In order to set the step function f2(y,p2) to substantially “0”, f2(y,p2)/2 is added to the partial image B, and −f2(y,p2)/2 is added to the partial image C.

[0223] Therefore, the partial image A and the partial image D are unified into a new partial image AD, and similarly the partial image B and the partial image C are unified into a new partial image BC.

[0224] Step S 407 :

[0225] The step value sequence d4(0) to d4(n) of the boundary between the new partial image AD and the new partial image BC is obtained through the above-mentioned expression (18).

[0226] Step S 408

[0227] The step value sequence d4(0) to d4(n) obtained in Step S 407 is fitted by a function f4(y,p3). The function f4(y,p3) is a function having a parameter p3 assuming that y is the longitudinal axis of the image.

[0228] Step S 409 :

[0229] In order to set the step function f4(y,p3) to substantially “0”, f4(y,p3)/2 is added to the partial image AD, and −f4(y,p3)/2 is added to the partial image BC.

[0230] The above-mentioned process enables an appropriate correction of the step between the respective partial images A, B, C and D even if those four partial images A, B, C and D are images on which not uniform but the varied components are superimposed.

[0231] Thereafter, the process shown in FIG. 11 is executed to add the repair at the step position. The repair in the vicinity of the boundary in this situation may be conducted every time the process shown in FIG. 14 is executed.

[0232] Subsequently, a third embodiment of the present invention will be described below.

[0233] In this embodiment, for example, in the x-ray sensing apparatus 100 shown in FIG. 1 , the four partial panels 101 a to 101 d are replaced with nine partial panels to acquire nine partial images (00) to (22) as shown in FIG. 15 A, and two partial images which are independent from each other are sequentially selected from those partial images (00) to (22) and combined together.

[0234] More specifically, the nine partial images (00) to (22) that are in a state shown in FIG. 15A are unified as indicated in the order of FIG. 15B, 15C , 15 D and 15 E. In this situation, as shown in FIG. 15 D, it should be noted that when two partial images are formed, the boundary of both the partial images is linear.

[0235] After the unified state shown in FIG. 15 D, the process shown in FIG. 11 is executed to add the repair at the step position. The repair in the vicinity of the boundary in this situation may be conducted every time the process shown in FIG. 14 is executed.

[0236] Subsequently, a fourth embodiment of the present invention will be described below.

[0237] In this embodiment, it is assumed that an image of the object 114 obtained by the x-ray sensing apparatus 100 shown in FIG. 1 , that is, an image stored in the memory 110 is, for example, an image shown in FIG. 16 .

[0238] In FIG. 16 , “41” denotes a portion onto which no x-rays are radiated, “42” denotes a portion onto which the x-rays radiated from the x-ray vessel 113 is focused through a collimator, and “43” denotes a portion onto which the object 114 is projected.

[0239] That is, a pixel onto which the x-rays are hardly radiated or not radiated at all exists in the portion 41 . On the contrary, a pixel onto which very intensive x-rays that have not passed through the object 114 are radiated exists in the portion 42 (a portion where there exists no object 114 ). Then, a pixel onto which intermediate intensive x-rays are radiated exists in the portion 43 (a portion of the object 114 ).

[0240] In this example, the output (electric signal) of the x-ray sensor panel is generally in proportion to the incident x-ray intensity, and in the energy transformation or the transforming process of the electric signal (the amplification, the impedance transformation or the like), the x-ray intensity and the linearity of the output electric signal are not always constant in the sensitive region of the sensor panel.

[0241] In particular, in the portion where the x-ray intensity is very high, the non-linearity gradually appears as approaching to the output saturation region of the electric system. Also, in the portion where the x-ray radiation is weak, the non-linearity appears due to the influence of noises, the non-stability of action at a super-low voltage of the electric circuit or the like.

[0242] The image information on the portion where the above-mentioned non-linearity appears is nonsense information in almost all cases, and there arises no problem even if such image information normally exists.

[0243] However, in the case where the offset component of the entire partial image (for example, an offset in case of the image data that is in proportion to the x-rays, and a gain in case of the image data that is in proportion to the logarithm) is derived from the statistical action of a part of the partial image, when the above-mentioned portion where an error caused by the non-linearity is large occupies a very large area, the error adversely affects the entire image, thereby failing to extract an accurate offset component.

[0244] Under the above circumstances, in this embodiment, a section where the linearity is reliable is determined in the sensitive region of the sensor panels 101 a to 101 d in advance, and only in the case where a representative value such as the average value in the vicinity is in the reliable section before or at the same time when the substantial step value d is calculated, the step value d is used to provide the step value d in a necessary region.

[0245] For example, in the case of medical x-ray sensing, at the time when it is detected that the amount of x-rays that passes through the object 114 and reaches the sensor panels 101 a to 101 d becomes an appropriate dose value by radiating the x-ray amount onto the object 114 in advance or measuring the x-ray amount that is called “photo timer”, the x-rays are blocked. As a result, the x-ray amount is always adjusted to an optimum value, and the linearity at the portion where there exists the object information is always optimum in many cases.

[0246] For this reason, in this embodiment, the average value m(i) of the data that interposes the boundary is obtained by the following expression (19). 4 $\begin{array}{cc}m\left(i\right)=\frac{1}{6}\left(\sum _{j=-2}^0x\left(n+j\right)+\sum _{j=-1}^4x\left(n+j\right)\right)& \left(19\right)\end{array}$

[0247] If the average value m(i) satisfies the following condition assuming that the minimum value is V0 and the maximum value is V1 in the reliable section, the step value d(i) obtained by the above-mentioned expression (18) is used and added to the histogram preparation.

[0248] V0≦m(i)≦V1

[0249] In this embodiment, two partial images are unified by using only the above-mentioned step value d(i). That is, in the generation of the histogram in Steps S 303 , S 304 and S 312 shown in FIG. 10 , only a value obtained from the average value m(i) is used.

[0250] Subsequently, a fifth embodiment of the present invention will be described below.

[0251] In this embodiment, only the step value is used for the reliable section in the fourth embodiment to conduct the function fitting corresponding to the second embodiment even if the step value at the necessary region is obtained.

[0252] That is, only the step value in the reliable section is used for the function fitting. As a result, the fitting of the data sequence of the irregular intervals is made.

[0253] In the above-mentioned first to fifth embodiments, in the case where one image is made up of a plurality of partial images that do not overlap each other, an offset between two adjacent partial images selected from the plurality of partial images is corrected (adjusted) to generate one partial image where the boundary is not visible, as one new partial image into which the two partial images are unified. In this situation, those two partial images may be corrected on the basis of the statistic property of the pixel value in the boundary portion region between the unified two partial images. This unifying process is repeatedly executed until all of the plural partial images have been unified into one new image with the result that one final image becomes an image where the boundary is not visible.

[0254] With the above structure, because the variations of the individual partial images can be stably corrected, one final image in the excellent state into which the partial images have been unified can be obtained.

[0255] A sixth embodiment of the present invention will be described below, in which the description of the same parts as those in the first embodiment will be omitted.

[0256] Differences of the sixth embodiment from the first embodiment reside in a correcting method where the correction to the partial images A, B, C and D is conducted by addition/subtraction of a constant value, and the operation related to Steps S 301 to S 314 in the operation of the x-ray sensing apparatus 100 . First, the correcting method according to the sixth embodiment will be described.

[0257] In the case of using the constant value, the step values d0(0 to L−1), d1(0 to (L−1)), d2(−1 to −(L−1)) and d3(−1 to −(L−1)) at the boundaries a, b, c and d are consolidated into one certain value. For that reason, the values D0, D1, D2 and D3 are derived from the step values d0(0 to L−1), d1(0 to (L−1)), d2(−1 to −(L−1)) and d3(−1 to −(L−1)) one by one, respectively.

[0258] As a method of deriving the values D0, D1, D2 and D3, there may be applied a method of obtaining the respective average values of the step values d0(0 to L−1), d1(0 to (L−1)), d2(−1 to −(L−1)) and d3(−1 to −(L−1)) as the values D0, D1, D2 and D3. However, for example, if the step values are not averagely corrected, but a region as large as possible is corrected, for example, as shown in FIG. 17 , the maximum frequency value (a peak of the histogram) is used as the values D0, D1, D2 and D3.

[0259] It is needless to say the following expression is satisfied.

D 0 +D 1 +D 2 +D 3=0

[0260] However, the following expression needs to be applied due to an error in calculation, an influence of the image noise, unconstant offset value and so on.

D 0 +D 1 +D 2 +D 3=ε(≠0)

[0261] The correction values F0, F1, F2 and F3 are added to the data of the respective partial images A, B, C and D to correct the step on the boundary.

[0262] For example, it is assumed that F0 is equal to 0 with the partial image A as a reference. In this case, if D0+D1+D2+D3=0 is satisfied, the same effect is obtained even if clockwise or counterclockwise order is applied, and the correction values F0, F1, F2 and F3 are represented by the following expressions:

F 0=0

F 1= D 1 =−D 0 −D 3 −D 2

F 2= D 1 +D 2 =−D 0 −D 3

F 3= D 1 +D 2 +D 3 =−D 0

[0263] On the other hand, if the following expression is satisfied, the inconsistency is prevented with the average of the clockwise and counterclockwise.

D 0 +D 1 +D 2 +D 3=ε(≠0)

[0264] That is, the correction values F0, F1, F2 and F3 in this case are represented by the following expressions:

F 0=0

F 1=( D 1 −D 0 −D 3 −D 2)/2

F 2=( D 1 +D 2 −D 0 −D 3)/2

F 3=( D 1 +D 2 +D 3 −D 0)/2

[0265] This corresponds to the uniformly distribution of the error all over.

[0266] As a method of preventing the above-mentioned inconsistency, there is applicable a method in which the minimum absolute value among the values D0 to D3 is replaced with the sum of other values, whose sign is reversed.

[0267] Also, because there is resultantly no fear that the pixel value of the image becomes negative if all of the correction values F0 to F3 are set to positive values, the minimum value among the correction values F0 to F3 may be added to all the values.

[0268] Subsequently, Steps S 301 to S 321 which are the operation different from Steps S 301 to S 314 of the operation of the x-ray sensing apparatus 100 according to the first embodiment will be described below.

[0269] FIGS. 18 and 10 show a correcting process using the above-mentioned step value sequence d(i). The correcting process is implemented by executing the processing program by the CPU 109 in accordance with the flowcharts shown in FIGS. 18 and 10 which are stored in the program memory 116 in advance.

[0270] Step S 301 to Step S 304 : See FIG. 18

[0271] Four sequences of d0(0) to d0(n), a1(n+2) to d1(M−1), d2(n+2) to d2(M−1), d3(0) to d3(n) on the boundaries of the partial images A, B, C and D are prepared through the above-mentioned expression (18).

[0272] More specifically, for example, as shown in FIG. 12 , the pixel value is extracted across the boundaries in accordance with the address of a pixel indicated by “21” , and the calculation process of the above-mentioned expression (12) is conducted, to thereby obtain the step value sequence d0(i). This calculation process is executed on all of the boundaries.

[0273] Step S 305 to Step S 308 :

[0274] The respective maximum frequency values of d0(0) to d0(n), a1(n+2) to d1(M−1), d2(n+2) to d2(M−1), d3(0) to d3(n) obtained in Step S 301 to Step S 304 are obtained as D0, D1, D2 and D3.

[0275] As D0, D1, D2 and D3, for example, values that represent the distribution of the average value, the center value or the like may be employed.

[0276] Step S 309 to Step S 312 :

[0277] It is assumed that the offset value F0 added to the partial image A is “0”, and the offset values F1, F2 and F3 added to other partial images B, C and D are (D1−D0−D3−D2)/2.

[0278] Step S 313 to Step S 317 :

[0279] It is assumed that all of the offset values F0 to F 3 in Step S 309 to Step S 312 are positive values. This is because since the pixel value of the image is basically a positive value, it is ensured that the pixel value is a positive value even after it has been subjected to calculation process.

[0280] More specifically, for example, assuming that the minimum value of the offset values F0 to F 3 is “MV”, “MV” is added to the respective values.

[0281] Step S 318 to Step S 321 :

[0282] The offset values F to F3 obtained in Step S 313 to Step S 317 are added to the partial images A to D, respectively.

[0283] This result is stored in the memory 110 . The image thus stored in the memory 110 results from conducting only the offset-like correction over the partial images A to D. However, in fact, the variations of the partial images A to D are not limited to the uniform offset, but there are many cases in which the above-mentioned correction is insufficient.

[0284] Under the above circumstances, there is executed a process in which only the vicinity of the boundaries of the image stored in the memory 110 is corrected so that a sense of incongruity that the entire image is the collection of a plurality of partial images is not given to an observer through processing starting from Step S 322 which will be described below. As one example, a method shown in FIG. 6C is employed.

[0285] Next, a seventh embodiment of the present invention will be described below.

[0286] In the sixth embodiment, four partial images obtained by the partial panels 101 a to 101 d resulting from dividing a single sensor panel into four pieces are processed by independent systems in parallel. On the other hand, in this embodiment, more partial images are processed by independent systems in parallel.

[0287] For example, as shown in FIG. 19 A, in the case where a single image is made up of the collection of nine partial images (00), (01), (02), (10), (11), (12), (20), (21) and (22), 12 boundaries exist in partial images (00), (01), (02), (10), (11), (12), (20), (21) and (22).

[0288] In this case, the substantial step value d is obtained in the same manner as that in the first embodiment in a direction indicated by an arrow in FIG. 19B with respect to the partial images (00), (01), (02), (10), (11), (12), (20), (21) and (22).

[0289] The step value d is obtained, independently, and thus all of the step values d are not inconsistent. Accordingly, the correction offset values to be added to the respective partial images need to be obtained from the average value of the sum of the step values at every routes.

[0290] More specifically, the joint of the partial images (00), (01), (02), (10), (11), (12), (20), (21) and (22) shown in FIG. 19A leads to a problem on circulation between cells (partial images) as shown in FIG. 19B .

[0291] In this example, if an offset value starts from a certain cell and is added with the step value on the connection line, and a value at the time of reaching the original cell is not 0 even if the system passes through any route, this system is inconsistent. In this example, a method of solving this inconsistency is realized.

[0292] For example, it is assumed that F00=“0” with reference to the offset value that is added to the partial image (00). The offset value F01 of a succeeding partial image (01) is obtained by averaging the offset values that pass through all of paths from the partial image (00) to the partial image (01).

[0293] In this case, four routes indicated by a broken line in FIG. 19B are considered to cover all with the least routes. The offset value F01 in this case can be obtained by the following calculation process. 5 $\begin{array}{c}\begin{array}{c}\mathrm{F01}=(d001+3·d0010+2·d1020+\\ d2122-d1222-\end{array}\end{array}$