JP2003296730A - Data processor, data processing method, image processor and image processing method - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To quickly calculate a median by a simple method. <P>SOLUTION: Nine pieces of data are divided into three sets with three pieces of data as one set, data are rearranged in order of a numerical value in each set, the data are rearranged in order of the value of second data of each rearranged set, and data that is not included in a target is determined to reduce data and the number of times to be compared. <P>COPYRIGHT: (C)2004,JPO

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a data processing device, a data processing method, an image processing device, and an image processing method for processing data such as image data using a median filter.

[0002]

2. Description of the Related Art Conventionally, a pixel area (window) of a certain size is set around each pixel, and the data (pixel value) of each pixel is replaced with the median value (median) of all the data in the window. Median filters that perform processing are known. The median filter has the feature that it can remove isolated points such as spike noise without impairing the resolution of image data, and has been often used as one function of image processing software on computers.

As a general method of implementing the median filter, there is a method of sorting the data in the window and using the data in the intermediate rank as the median. But,
There is a problem that the sorting process has a large processing amount. Therefore,
Various methods have been proposed for performing median filtering at high speed.

The technique disclosed in Japanese Patent Laid-Open No. 2001-45507 is applied to a window of size N × M. The median value of each row in the window is calculated, and the median value of the median values of the rows is used as an approximate median value. It is what that time,
The processing speed is increased by using the processing result of the adjacent pixels.

The technique disclosed in Japanese Patent Laid-Open No. 2001-167262 is applied to an N × M size window.
The median is calculated by sorting the pixel values in the window. By using the result of sorting when calculating the median of adjacent pixels, and combining the result with unnecessary pixel values removed and the result of sorting with only newly added pixels, the current pixel within the window The result of sorting the pixel values of is obtained. It is assumed to be implemented by hardware.

The method disclosed in Japanese Patent Laid-Open No. 5-300372 is applied to a window of size N × M, in which each row in the window is sorted, and then each column is sorted.
After that, the median of the diagonal rows is obtained and used as the median approximation. In the case of 3x3 size, the median is accurate.

The method disclosed in Japanese Patent Laid-Open No. 5-128253 is applied to a window of size N × M. Each row in the window is sorted, and then the maximum value among the maximum values of each row is taken in order. By going through, the pixel value of an arbitrary rank is obtained not only by the median. It is assumed to be implemented by hardware.

The technique disclosed in Japanese Patent Laid-Open No. 5-233804 is applied to an N × M size window, in which pixel values in the window are sorted to obtain a median. Use the result of sorting when calculating the median of adjacent pixels, remove unnecessary pixels from the result, and add a new pixel value one pixel at a time to obtain the result of sorting the pixel values in the window at the current pixel. There is. It is assumed to be implemented by hardware.

The method disclosed in Japanese Patent Laid-Open No. 6-274617 is applied to a 3 × 3 size window, and the pixel values in the window are sorted to obtain the median. First, sort by each column in the window, 1
By synthesizing columns, the sorting result for all pixels in the window is obtained. At that time, the processing speed is increased by using the processing results of the adjacent pixels.

However, since the median filter is a process performed for each pixel, a faster method is desired. Further, when used in software, parallel processing is often impossible, and even in that case, a method capable of high speed processing is desired. Further, in order to further improve the performance, it is desired to obtain an accurate median value rather than an approximate value.

The method disclosed in Japanese Patent Laid-Open No. 2001-45507 is not desirable because it is a high speed but an approximate value. The device disclosed in Japanese Patent Laid-Open No. 2001-167262 is assumed to be implemented in hardware, and it cannot be said that the operation of the similar method on software is sufficiently high speed.

The device disclosed in Japanese Patent Laid-Open No. 5-300372 cannot be said to be sufficiently high speed. Japanese Patent Laid-Open No. 5-1282
The one shown at 53 is assumed to be implemented in hardware, and it cannot be said that the operation in software by the similar method is sufficiently fast.

The device disclosed in Japanese Patent Laid-Open No. 5-233804 is assumed to be implemented in hardware, and it cannot be said that the operation of the similar method on software is sufficiently high speed. The device disclosed in JP-A-6-274617 cannot be said to be sufficiently high speed. Therefore, there is a demand for a median that can be obtained at high speed by a simple method.

[0014]

SUMMARY OF THE INVENTION It is an object of the present invention to obtain a median value of a plurality of data in a simple method and at high speed in a data processing such as image data using a median filter. I am trying.

[0015]

The data processing apparatus of the present invention obtains the value of the central data when 9 pieces of data are arranged in numerical order, and sets 9 pieces of data and 3 pieces of data as a set. The data is sorted into the first data, the second data, and the third data in descending order of numerical value for each group.
Setting means for setting the first set, the second set, and the third set in the descending order of the value of the second data of each sorted set, and the second set of the second data rather than the value of the first set of the third data. Is large and the value of the first data of the third group is larger than the value of the second data of the second group, the second data of the second group is the central data when the above nine data are arranged in numerical order. And a value of the third data set that is smaller than the value of the second data set of the second group and smaller than the value of the second data set of the first set. Is smaller than the value of the third data of the first set, the second determination means for determining the second data of the second set as the central data when the above nine data are arranged in numerical order.
When the value of the second data of the set is small and the value of the first data of the third set is larger than the value of the second data of the second set, the third data of the first set
A third judging means for judging the data having the largest value among the data, the third data of the second group, and the first data of the third group to be the central data when the nine data are arranged in numerical order; When the value of the second set of second data is larger than the value of the first set of third data and the value of the third set of first data is smaller than the value of the second set of second data, the first set of third data, The data of the largest value in the third data of the second group and the first data of the third group is judged to be the central data when the nine data are arranged in numerical order.
And the judgment means of.

The data processing apparatus of the present invention obtains the median value of nine pieces of data, divides the nine pieces of data into three sets of three pieces, and sets the data value of each set. The first processing means for rearranging in ascending or descending order, the first set having the largest median value, the second set having the next largest median value after the first set, and the third set having the smallest median value are provided. Second processing for checking the first determining means for determining, the magnitude relationship between the minimum value of the first set and the median value of the second set, and the magnitude relationship between the median value of the second set and the maximum value of the third set Means and the second processing means, if the minimum value of the first set is greater than or equal to the median value of the second set and the median value of the second set is greater than or equal to the maximum value of the third set, or
If the minimum value of the set is less than or equal to the median value of the second set and the median value of the second set is less than or equal to the maximum value of the third set, then the median value of the second set is determined to be the median value of the above nine data. If the minimum value of the first set is larger than the median value of the second set and the maximum value of the third set is larger than the median value of the second set by the second determining means and the second processing means, The third judgment means for judging the minimum value of the set, the maximum value of the third set, and the maximum value of the second set as the median value of the nine data, and the second processing means. If the minimum value of one set is less than the median value of the second set and the maximum value of the third set is less than the median value of the second set, then the minimum value of the first set, the maximum value of the third set and the second set of And a fourth judging means for judging the maximum value among the minimum values as the median value of the nine data.

According to the image processing apparatus of the present invention, for each pixel of the input image, the median value of nine pixel values in a 3 × 3 size window centered on each pixel is defined as the pixel value of each pixel in the output image. When the pixels in the window are divided into three groups of three pixels arranged in the row direction or the column direction, the processed pixels that are adjacent in a direction different from the above direction among the three groups of the window. The first processing means for rearranging one set that is not included in the previous window centered on the ascending order or the descending order, and two sets included in the previous window when processing the processed pixels,
A storage means for holding the results sorted in ascending or descending order, a first set having the largest median value, a second set having the next largest median value of the first set, and a second set having the smallest median value. The first determination means for determining three sets, the size relationship between the minimum value of the first set and the median value of the second set, and the size relationship of the median value of the second set and the maximum value of the third set are examined. The second processing means, and when the minimum value of the first set is equal to or larger than the median value of the second set and the median value of the second set is equal to or larger than the maximum value of the third set by the second processing means, or When the minimum value of one set is less than or equal to the median value of the second set and the median value of the second set is less than or equal to the maximum value of the third set, the median value of the second set is the median value of the nine pixels. The minimum value of the first set is larger than the median value of the second set and the maximum value of the third set is larger than the median value of the second set by the second judging means for judging and the second processing means. In this case, a third judging means for judging the minimum value of the minimum value of the first set, the maximum value of the third set and the maximum value of the second set as the median value of the nine pixels, and the second judgment means. When the minimum value of the first set is smaller than the median value of the second set and the maximum value of the third set is smaller than the median value of the second set by the processing means, the minimum value of the first set and the maximum value of the third set And a fourth judgment means for judging the maximum value of the minimum values of the second set as the median value of the nine pixels.

In the image processing apparatus of the present invention, for each pixel of the input image, the median value of nine pixel values in a 3 × 3 size window centered on each pixel is set as the pixel value of each pixel in the output image. In terms of values, storage means for storing pixels in three rows or three columns of the input image and the stored pixels are divided into groups of three pixels arranged in the column direction or the row direction, and the pixel value is set for each group. A first processing means for rearranging in ascending or descending order, a first set having the largest median value, a second set having a next larger median value than the first set, and a third set having a minimum median value. The first determination means for determining, the magnitude relationship between the minimum value of the first set and the median value of the second set, the median value of the second set, and the third value
The second processing means for checking the magnitude relation of the maximum value of the set, and the second processing means for setting the minimum value of the first set to the median value of the second set and the median value of the second set to the third set. If the minimum value of the first set is less than the median value of the second set and the median value of the second set is less than or equal to the maximum value of the third set, the median value of the second set is greater than or equal to the maximum value. Is determined to be the median value of the nine pixels, and the second processing means determines that the minimum value of the first set is greater than the median value of the second set and the maximum value of the third set is the maximum value. If it is larger than the median value of the two sets, the minimum value of the minimum value of the first set, the maximum value of the third set and the maximum value of the second set is set to 9 above.
The third determination means for determining the median value of the pixels and the second processing means make the minimum value of the first set smaller than the median value of the second set and the maximum value of the third set of the second set. And a fourth determination means for determining the maximum value among the minimum value of the first set, the maximum value of the third set, and the minimum value of the second set as the median value of the nine pixels when the value is smaller than the median value. Have.

In the image processing apparatus of the present invention, for each pixel of the force image, the median value of the four pixel values in the 2 × 2 size window including each pixel is set as the pixel value of each pixel in the output image. It has a processing means for selecting any three pixel values out of the four pixels and taking the median value of the three pixel values as the pixel value of the output image.

In the image processing apparatus of the present invention, for each pixel of the input image, the median value of the four pixel values in the 2 × 2 size window including each pixel is set as the pixel value of each pixel in the output image. 2 pixels in the above window
The pixel is divided into two groups, one group is divided into two groups, the first processing means for examining the magnitude relationship between two pixel values for each group, and the magnitude relationship between the two larger pixel values for each group are examined. The magnitude relationship between the first output means for outputting the smaller pixel value of the two pixel values and the two smaller pixel values of each set is checked, and this 2
A second output means for outputting the larger pixel value of the pixel values, and an average of the outputs of the first and second output means as the pixel value of the output image. And processing means.

[0021]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS [First Embodiment] A first embodiment of the present invention will be described below with reference to the drawings.
As the first embodiment, an example will be described in which "median (median value)", which is one of the statistical quantities representing the population, is to be obtained at high speed.

The median is data that is in the central position when all data are arranged in ascending order (or descending order). In addition to medians, there is “average” as a similar statistic that represents a population, but the average is easily affected by outliers such as noise. Since the median is an amount that is not affected by the outlier, it is often used when processing that is not affected by noise or the like is desired. In this embodiment, the median of 9 pieces of data is obtained.

FIG. 1 is a block diagram of a processing apparatus according to the first embodiment of the present invention. In this processing apparatus, in this drawing, each group sorting section 1 that sorts three sets of three data, group central value order determining section 2 that examines the magnitude relationship of the median values of each group, and group central value order determining section 2 The median determination unit 3 and the median determination unit 3 which determine the median using the result of 3 are called, and the minimum value determination unit 4 and the median determination unit 3 which determine the minimum value of the three data are called and the three data The maximum value determining unit 5 determines the maximum value of

The processing of each unit in the above configuration will be described in detail below. First, each group sort section 1 divides nine pieces of data into three groups of three pieces. The situation is shown in FIG. In the figure, circles represent data and rectangles represent sets. Then, each set is sorted in ascending order. This may be in descending order. The situation is shown in FIG. The inequality sign indicates the magnitude relationship of the data.

Here, when comparing the magnitude relation of two data having the same value, it is assumed that one of the data is processed as being larger than the other. Therefore, for example, in FIG.
In the inequality sign, the magnitude relationship is shown, but in reality there are cases where they are equal.

Next, the set median value order determination unit 2 checks the order of the median values of each set. This is shown in FIG. In the figure, the data with double circles is the median value of each set, the set having the largest median value is set A (first set), and the set having the second largest median value is set B. (Second group), the group having the smallest median value is determined as the group C (third group). FIG. 5 shows the rearranged order of A, B, and C.

Actually, it is not necessary to rearrange, and a rearranged diagram is shown for convenience of explanation. For the sake of explanation, the minimum value of the set A is D, the median value of the set B is E, the maximum value of the set C is F, the maximum value of the set B is G, and the minimum value of the set B is H.

Next, the processing of the median decision unit 3 will be described. Here, pay attention to E. As you can see from Fig. 5, there are 3 data above E, 3 data below E, E
There are two data (D and F) whose size relationship with is unknown. Therefore, regarding D and F, the magnitude relation between E and E is examined. The results are classified into the following three cases.

Case 1: When D is E or more and E is F or more, or when D is E or less and E is F or less In this case, there are four data of E or more and four data of E or less. It is confirmed that E is a median.
In this case, the median decision unit outputs E.

Case 2: When D is larger than E and F is larger than E In this case, since there are 5 pieces of data above E and 3 pieces of data below E, E is the fourth data from the smallest. There will be. This is shown in FIG. Then, the median is the smallest data among the five data of E or more. Five pieces of data equal to or more than E are three pieces of data belonging to the set A, G and F, but three pieces belonging to the set A.
Since D is the smallest among the pieces of data, it is possible to exclude other than D from the median candidates. Therefore, there are three median candidates, D, G, and F. Therefore, D, G, F
If you choose the smallest data among the three, it is the median. Therefore, the median determination unit 3 uses the minimum value determination unit 4 to obtain the minimum value of D, G, and F, and outputs the result.

Case 3: When D is smaller than E and F is smaller than E In this case, since there are 3 pieces of data above E and 5 pieces of data below E, E is the fourth largest data. There will be. This is shown in FIG. Then, the median is the largest data out of the five data below E. 5 pieces of data less than E are 3 pieces of data belonging to the set C, H and D, and 3 pieces of data belonging to the set C.
Since F is the largest among the individual data, other than F can be excluded from the median candidates. Therefore, there are three median candidates F, H, and D. Therefore, F, H, D
If you choose the largest of the three, the median. Therefore, the median determination unit 3 uses the maximum value determination unit 5 to obtain the maximum value of F, H, and D, and outputs the result.

Here, the number of comparisons will be examined. First, 3
The number of comparisons for determining the order of the data pieces has a probability of 1
2/3 and 2/3 with a probability of 2/3, so an average of 2 + 2
/ 3 times, up to 3 times. To determine the order of three data,
In the first embodiment, the number of comparisons is 10 + 2/3 times on average, and 12 times at maximum, because it is performed four times at the beginning (sorting each group and determining the order of the median of each group).

Next, for D and F, the comparison is performed twice when checking the magnitude relation between E and E. The total number of comparisons so far is 12 + 2/3 times on average, and 14 times at maximum.
Finally, the probability of becoming Case 1 is 1/2, the probability of becoming Case 2 is 1/4, the probability of becoming Case 3 is 1/4, and when Case 1 results in 0 comparisons and Case 2 If the number of comparisons is 2, the number of comparisons is 2, and if the number of 3 is 3, the number of comparisons is 2. Therefore, the total number of comparisons is 13 + 2/3 on average and 1 at maximum.
Six times.

On the other hand, as a method for obtaining the median of nine pieces of data, Japanese Patent Application Laid-Open No. 5-300372 and Japanese Patent Application Laid-Open No. 5-1
28253, the method disclosed in Japanese Patent Laid-Open No. 5-300372 requires an average of 18 + 2/3 times and a maximum of 21 times, and the method disclosed in Japanese Patent Laid-Open No. 5-128253 has an average of 18 times and a maximum of 21 times. 19 comparisons are required.

In general, as a method for obtaining the median of nine pieces of data, a method of sorting in ascending order and ending when the data of the fifth rank is decided is often used. In this method, the data of the first rank is determined 8 times, the data of the second rank is determined 7 times, 3
Six comparisons are required to determine the data of the fourth rank, five times to confirm the data of the fourth rank, and four times to confirm the data of the fifth rank, so a total of 30 comparisons are required. is there. From the above, it can be seen that the present invention is extremely fast as a method for obtaining the median of 9 pieces of data.

A concrete example of the above processing is shown in FIGS.
12 to 12, and the flowcharts shown in FIGS. 13 and 14 will be described. For example, the 9 pieces of data to be referred to are divided into 3 columns (groups) of 3 pieces each. Then, as shown in FIG. 3, the three in the first column are rearranged in the order of large, medium, and small (first data, second data, and third data), and the three in the second column are large, medium, and small. Rearrange in the order of small, and rearrange the three in the third column in the order of large, middle, and small (ST1, 2,
3).

The center (median value) of each column is compared, and each column is ordered (order is determined) (ST4). As shown in FIG. 8, the array of virtual data strings in this ordered state is A1, A2, A3 (first set, second set, 3rd group), and B1 and B
2, B3, and C1, C2, and C3 in order from the largest data in the column having the smallest center value. At this time, A1> A2
> A3, B1>B2> B3, C1>C2> C3,
It is confirmed that A2>B2> C2.

As a result, there are definitely five (A3, B2, B3, C2, C3) smaller than itself as A1 and A2, and cannot be the fifth. In addition, C2 and C3 are definitely larger than themselves (A1, A2, B1, B2, C
1) Yes, not the fifth. Further, it is known that B1 is larger than B2, A1, A2 and B1 are smaller than B2, and B3, C2 and C3 are smaller than B2. But not
A3 and C1.

In this state, A3 and B2 are compared to obtain a magnitude relationship, and B2 and C1 are compared to obtain a magnitude relationship (ST5). At this time, as shown in FIG. 9, A3 <B2
When B2 <C1 (ST6), four are larger than B2, namely A1, A2, B1, and C1, four are smaller than B2, four such as A3, B3, C2, and C3, and B2 is the fifth. (ST7).

Further, as shown in FIG. 10, when A3> B2 and B2> C1 (ST8), A is larger than B2.
1, A2, A3, B1 are four, less than B2 are B3, C1, C2, C3, and B2 is determined to be the fifth (ST9).

Further, as shown in FIG. 11, when A3> B2 and B2 <C1 (ST10), five are larger than B2: A1, A2, A3, B1 and C1, and B2 is 6 from the top. Judged to be the second, A1, A2, A3, B1, C1
The smallest of the five is judged to be the fifth. At this time, A1 and A2 are excluded because they are known to be larger than A3, and the smallest one is determined to be the fifth by comparing the three A3, B1, and C1 (ST11). For example, the smaller one obtained by comparing A3 and B1 and the smaller one obtained by comparing C1 are the smallest ones, and the comparison may be performed twice.

Further, as shown in FIG. 12, in the case of A3 <B2 and B2> C1 (ST12), five smaller than B2 are A3, B3, C1, C2 and C3, and B2 is 6 from the bottom. Judged to be th, A3, B3, C1, C2, C3
The largest of the five is judged to be the fifth.
At this time, C2 and C3 are excluded because they are known to be smaller than C1, and the largest one is determined to be the fifth by comparing three of A3, B3, and C1 (ST13). For example, the larger one according to the comparison of A3 and B3 and the larger one according to the comparison of C1 are the largest ones, and the comparison may be performed twice. [Second Embodiment] As a second embodiment, an example will be described in which it is desired to apply a median filter of a 3 × 3 size window to an input image at high speed.

In the second embodiment, the case where the cache is not large enough will be described. An embodiment in which the cache is sufficiently large will be described in the third embodiment. The median filter replaces the value of each pixel of the image with the median of the pixel in the window centered on the pixel. FIG. 15 shows a specific example. It appears that the median filter process is applied to the position (i, j). A frame is a window drawn with a solid line. If nine pixel values in the window are arranged in descending order, 255,1
68,130,121,69,66,51,13,2, and the median is the fifth ranking 69.
Therefore, 69 is written at the position (i, j) of the output image.

In the following description of the second embodiment, i, j
Shall start from 0. In addition, the median filter processing is performed by fixing j and applying i from 0 in ascending order. When i becomes maximum, j is incremented by 1 and the same processing is repeated. Regarding j, it is assumed to increase from 0 in order. Of course, the above order is for the purpose of explanation, and the present invention is effective even if i and j are exchanged, or the order from the small is changed to the order from the large.

When obtaining the median, the method described in the first embodiment can be used, but since the windows of adjacent pixels include six common pixels,
In the second embodiment, it will be described that higher speed processing can be performed by utilizing this.

FIG. 16 is a block diagram of a processing device according to the second embodiment of the present invention. In this processing apparatus, in this drawing, a new sorting unit 11 that sorts three newly added pixels, a sorting result that adds a new sorting result to the sorted results of processed adjacent pixels, and deletes unnecessary sorting results. Result processing storage unit 12, median determination unit 1 that determines the median using the results of the median value order determination unit 13 and the median value order determination unit 13 that examines the magnitude relationship of the median values of each group.
4. By the median determination unit 14, the minimum value determination unit 15 that determines the minimum value of the three data, and the median determination unit 14 that is called by the maximum value determination unit 16 that determines the maximum value of the three data It is configured.

The processing of each unit in the above configuration will be described in detail below. In FIG. 17, the position (i-1,
It is assumed that the pixel of j) has been processed. Also, the position (i,
j) the window centered on the pixel of L, and the position (i-
Let M be a window centered on the pixel of (1, j), position (i-2, j-1), position (i-2, j), position (i-
2, j + 1) of three pixels as one set (set 1) and position (i-
1, j-1), position (i-1, j), position (i-1, j)
+1) three pixels as one set (set 2), position (i, j-
1), position (i, j), and position (i, j + 1), three pixels are set as one set (set 3), position (i + 1, j-1), position (i
Three pixels at +1, j) and position (i + 1, j + 1) are set as one set (set 4).

The new sort unit 11 sorts the sets 4 which are not included in the window M but included in the window L in ascending order. This may be in descending order. The sort result processing storage unit 12 stores the result of sorting the set 1, the set 2, and the set 3 at the time of processing the pixel at the position (i-1, j), and adds the sort result of the set 4 to that. A process of deleting the sorting result of the set 1 is performed.

The set median order determining unit 13 checks the order of the median value of each set by using the sort results of the sets 2, 3, and 4 obtained from the sort result processing storage unit 12. This is the same as the set median order determining unit 2 in the first embodiment.
The median determination unit 14 is the same as the median determination unit 3 in the first embodiment.

The minimum value determining unit 15 is the same as the minimum value determining unit 4 in the first embodiment. Maximum value determination unit 1
6 is the same as the maximum value determination unit 5 in the first embodiment. In the second embodiment described above, i
Can be used for two or more pixels. A pixel having i of 1 or less is not specified here, but the method described in the first embodiment can be used as an example.
In the second embodiment, since sorting of the sets 2 and 3 is not required, it is possible to realize higher-speed processing than that of the first embodiment.

[Third Embodiment] As a third embodiment, an example will be described in which it is desired to apply a 3 × 3 size median filter to an input image at high speed. In the third embodiment, the case where the cache is sufficiently large will be described. In the third embodiment, the median filter processing is collectively performed for each row. The situation is shown in FIG. The row direction is j and the column direction is i. FIG. 18 shows a state in which the median filter is collectively applied to the j-th row. In order to apply the median filter to all the pixels in the j-th row, it is necessary to store all the pixel values in the frame written by the bold line.

In the following description of the third embodiment, j is 0.
Shall begin with. In addition, the median filter is applied in ascending order of j. Of course, the above order is for explanation, and the present invention is effective even if i and j, rows and columns are exchanged, or small order is changed to large order.

FIG. 19 is a block diagram of a processing device according to the third embodiment of the present invention. In this processing device, a three-row image reading unit 2 for reading an image for three rows is shown in FIG.
1, a three-row image storage unit 22 for storing the images of three rows read by the three-row image reading unit 21, and an image stored in the three-row image storage unit 22 into groups of three pixels in each column A group median order in which each group is sorted, and each group is sorted and stored in the three-row image storage unit 22 and the magnitude relationship between the median values of three adjacent groups stored in the three-row image storage unit 22 is checked. The median determining unit 25, which determines the median using the results of the determining unit 24 and the set median order determining unit 24, is called by the median determining unit 25, and the minimum value determining unit 26 determines the minimum value of the three data, and the median. Maximum value determination unit 2 which is called by the determination unit 25 and determines the maximum value of three data.
7

The processing of each unit in the above configuration will be described in detail below. The three-row image reading unit 21 reads the pixel values of three adjacent rows of the input image and stores them in the three-row image storage unit 22. In FIG. 20, line j−1,
It stores three rows, the j-th row and the j + 1-th row.

In each group sort section 23, the three-row image storage section 2
The images stored in No. 2 are divided in the column direction, and the divided sets are sorted and stored again in the three-row image storage unit 22. This is shown in FIG. The three pixel values in each column are
Sorted in ascending order. Each group sort section 23 sorts all the columns stored in the three-row image storage section 22.

In the group median value order determination unit 24, the i−1th column, the ith column, and the i + th column stored in the three-row image storage unit 22.
The order of the median value of each set is examined using the sorting result of each set of one column. This is the same as the set median order determining unit 2 in the first embodiment.

The median determination unit 25 is the same as the median determination unit 3 in the first embodiment. The minimum value determination unit 26 is the minimum value determination unit 4 in the first embodiment.
Is the same as. The maximum value determination unit 27 is the same as the maximum value determination unit 5 in the first embodiment. Once the processing of the three-row image reading unit 21 and each group sort unit 23 is performed for the j-th row, the group median order determination unit 24, the median determination unit 25, the minimum value determination unit 26, and the maximum value determination unit 27.
Is performed for all i.

In comparison with the second embodiment, the third embodiment is speeded up by omitting the processing performed by the sort result processing storage unit 22 in the second embodiment. However, each time the median filter is applied to each pixel,
It is necessary to read five to six pixel values from the three-row image storage unit 22, and when the read overhead is large, the second embodiment is faster. Therefore, the third
This embodiment is effective when the three-line image storage unit 22 is assigned to a storage medium that can be accessed at high speed, such as a cache of the CPU.

[Fourth Embodiment] As a fourth embodiment, an example will be described in which it is desired to apply a median filter of a 2 × 2 size window to an input image at high speed. Figure 2
1 shows a state in which the median filter processing of a 2 × 2 size window is performed. FIG. 21 shows the position (i, j)
It seems that the median filter is applied to. Figure 21
The frame indicated by the solid line is the window. Since there is no pixel at the center of the 2 × 2 size window, the position (i, j) is arranged in the upper left corner of the window here. Of course, this is only an example, and the present invention is similarly applied to the arrangement in which the upper right corner, the lower left corner, and the lower right corner are arranged.

When the four pixel values in the window in FIG. 21 are arranged in descending order, they are 255, 31, 26 and 25.
Here, since the median of four data does not have an intermediate rank, generally, the second data, the third data, or the average of the second data and the third data is used as the median. To be Therefore, if the second data is a median, for example, 31 becomes a median in FIG. 21, and the position (i, j) of the output image is displayed.
31 will be written in.

Note that the second data or the third data may be used. The three pixels are appropriately selected from the four pixels, and the median value T of the three pixel values is selected.
, T is either the second or third of the four pixel values. That is, it can be said that T is a median of four pixel values.

FIG. 22 is a block diagram of a processing device according to the fourth embodiment of the present invention. This processing device is configured by a median value selection unit 30 that selects three pixel values from four pixel values and uses the median value as the pixel value of the output image in FIG. Now, let's check the number of comparisons. The number of comparisons to obtain the median value among the three pixel values is 2 with probability 1/3.
Since the number of comparisons is 3 with a probability of 2/3, the average number of comparisons in the fourth embodiment is 2 + 2/3, and a maximum of 3 comparisons.

On the other hand, in general, as a method for obtaining the median of four data, a method of sorting in ascending order and ending when the data of the second rank is determined is often used. According to this method, three comparisons are required to determine the data of the first rank and two comparisons are required to determine the data of the second rank, so that a total of five comparisons are required. From the above, it can be seen that the present invention is extremely fast as a method for obtaining the median of four data.

[Fifth Embodiment] As a fifth embodiment, an embodiment will be described in which it is desired to apply a median filter of a 2 × 2 size window to an input image at high speed. 2x
The description regarding the 2-size median filter is as described in the fourth embodiment. In the fifth embodiment, the case where the average of the second data and the third data is used as the median of four data will be described.

FIG. 23 is a block diagram of a processing device according to the fifth embodiment of the present invention. This processing device divides four pixel values into two groups of two in the figure, and examines the magnitude relationship between the two pixel values of each group.
A small value determination unit 32 that compares the larger pixel value of each group and outputs the smaller pixel value, and a smaller value determination unit 32 that compares the smaller pixel value of each group and outputs the larger pixel value. A large value determination unit 33, an average value calculation unit 3 for calculating an average of the output of the small value determination unit 32 and the output of the large value determination unit 33.
It is 4.

The processing of each unit in the above configuration will be described in detail below. Each group size comparison unit 31 divides the four pixel values into two groups of two, and examines the size relationship of the two pixel values of each group. FIG. 24 shows the situation. Set 1 and set 2
Are divided into two groups, and the magnitude relationship between the two pixel values in each group is indicated by an inequality sign.

The pixel value determined to be large in each of the set 1 and the set 2 is input to the small value determination unit 32, and the magnitude relationship is checked. The pixel value determined to be small in each of the set 1 and the set 2 is input to the large value determination unit 33, and the size relationship is examined.

The pixel value U determined to be large by the small value determination unit 32 is the maximum value among the four pixel values. The pixel value V determined to be small by the large value determination unit 33 is the minimum value among the four pixel values. Then 4
Among the pixel values, the two pixel values excluding U and V are either the second largest pixel value or the third largest pixel value. There is no need to find out which is the second and which is the third, and by averaging these two values,
The desired median can be obtained. Therefore, the median can be obtained by averaging the output of the small value determination unit 32 and the output of the large value determination unit 33 by the average value calculation unit 34.

Here, the number of comparisons will be examined. Each group size comparison unit 31 performs the comparison twice, and the small value determination unit 32
Is compared once, and the maximum value determination unit 33 determines 1
Since the comparison is performed twice, the number of comparisons in the fifth embodiment is four.

On the other hand, in general, when the median of four data is obtained as the average of the data of the second rank and the data of the third rank, sorting is performed in ascending order and the second data and the third data are sorted. The method of obtaining the second data is used. In this case, 3 to confirm the data of the first rank
Since a comparison is required twice to determine the data of the second rank and twice to determine the data of the third rank, a total of 6 is required.
Need to compare twice.

From the above, it can be seen that the present invention is extremely fast as an implementation method when the median of four data is obtained as the average of the data of the second rank and the data of the third rank. As an advantage of the fifth embodiment from the fourth embodiment, it is possible to obtain higher performance by taking an average. On the other hand, in the fourth embodiment,
An advantage of the fifth embodiment is that processing can be performed at higher speed.

As described above, the median of 9 pieces of data can be obtained at high speed. Further, it is possible to realize a 3 × 3 median filter which is faster than the conventional method. In addition, a 2 × 2 median filter that is faster than the conventional method can be realized.

[0073]

As described above in detail, according to the present invention, a median value of a plurality of data can be obtained by a simple method and at high speed in a data processing such as image data using a median filter. be able to.

[Brief description of drawings]

FIG. 1 is a block diagram of a processing device according to a first embodiment.

FIG. 2 is a diagram for explaining each group sort unit.

FIG. 3 is a diagram for explaining each group sort unit.

FIG. 4 is a diagram for explaining each group sort unit.

FIG. 5 is a diagram illustrating a process of a median determination unit.

FIG. 6 is a diagram illustrating a process of a median determination unit.

FIG. 7 is a diagram illustrating a process of a median determination unit.

FIG. 8 is a diagram illustrating a method of obtaining a median of 9 pieces of data.

FIG. 9 is a diagram illustrating a method of obtaining a median of nine pieces of data.

FIG. 10 is a diagram illustrating a method of obtaining medians of nine pieces of data.

FIG. 11 is a diagram illustrating a method of obtaining a median of nine pieces of data.

FIG. 12 is a diagram illustrating a method of obtaining a median of 9 pieces of data.

FIG. 13 is a flowchart for explaining a method of obtaining a median of 9 pieces of data.

FIG. 14 is a flowchart for explaining a method of obtaining medians of nine pieces of data.

FIG. 15 is a diagram showing a specific example of a median filter.

FIG. 16 is a block diagram of a processing device according to a second embodiment.

FIG. 17 is a diagram illustrating the processing of each unit of the processing device according to the first embodiment.

FIG. 18 is a diagram showing a state in which median filter processing is collectively performed for each row.

FIG. 19 is a block diagram of a processing device according to a third embodiment.

FIG. 20 is a diagram showing a storage example of a three-row image storage unit.

FIG. 21 is a diagram showing how a median filter process is applied to a 2 × 2 size window.

FIG. 22 is a block diagram of a processing device according to a fourth embodiment.

FIG. 23 is a block diagram of a processing device according to a fifth embodiment.

FIG. 24 is a diagram showing a configuration example of each group size comparison unit.

[Explanation of symbols]

1 ... Each set sorting unit, 2 ... Set median value order determining unit, 3 ... Median determining unit, 4 ... Minimum value determining unit, 5 ... Maximum value determining unit.

─────────────────────────────────────────────────── ───

[Procedure amendment]

[Submission date] January 9, 2003 (2003.1.9)

[Procedure Amendment 1]

[Document name to be amended] Statement

[Name of item to be corrected] Claim 1

[Correction method] Change

[Correction content]

[Procedure Amendment 2]

[Document name to be amended] Statement

[Name of item to be corrected] Claim 2

[Correction method] Change

[Correction content]

[Procedure 3]

[Document name to be amended] Statement

[Name of item to be corrected] 0015

[Correction method] Change

[Correction content]

[0015]

The data processing apparatus of the present invention obtains the value of the central data when 9 pieces of data are arranged in numerical order, and sets 9 pieces of data and 3 pieces of data as a set. The data is sorted into the first data, the second data, and the third data in descending order of numerical value for each group.
Setting means for setting the first set, the second set, and the third set in the descending order of the value of the second data of each sorted set, and the second set of the second data rather than the value of the first set of the third data. Is large and the value of the first data of the third group is larger than the value of the second data of the second group, the second data of the second group is the central data when the above nine data are arranged in numerical order. And a value of the third data set that is smaller than the value of the second data set of the second group and smaller than the value of the second data set of the first set. Is smaller than the value of the third data of the first set, the second determination means for determining the second data of the second set as the central data when the above nine data are arranged in numerical order.
When the value of the second data of the set is small and the value of the first data of the third set is larger than the value of the second data of the second set, the third data of the first set
A third judging means for judging the data having the smallest value among the data , the first data of the second set, and the first data of the third set to be the central data when the nine data are arranged in numerical order; When the value of the second set of second data is larger than the value of the first set of third data and the value of the third set of first data is smaller than the value of the second set of second data, the first set of third data, The data of the largest value in the third data of the second group and the first data of the third group is judged to be the central data when the nine data are arranged in numerical order.
And the judgment means of.

Claims (12)

[Claims] 1. A data processing device for obtaining a value of central data when 9 pieces of data are arranged in numerical order. 9 pieces of data are divided into 3 sets of 3 pieces, and each set is divided. Are sorted into the first data, the second data, and the third data in descending order of numerical value, and the first data set, the second data set, and the third data set in descending order of the value of the second data of each sorted data set. And the setting means for setting, the value of the second data of the second group is larger than the value of the third data of the first group, and the value of the third data of the third group is larger than the value of the second data of the second group, The second set of 2nd data is judged to be the central data when the above 9 data are arranged in numerical order.
And the value of the second data of the second group is smaller than the value of the third data of the first group, and the value of the first data of the third group is smaller than the value of the second data of the second group, The 2nd set of 2nd data is judged to be the central data when the above 9 data are arranged in numerical order.
And the value of the second data of the second group is smaller than the value of the third data of the first group, and the value of the first data of the third group is larger than the value of the second data of the second group, The third judgment that the data with the largest value among the third data of the first set, the third data of the second set, and the first data of the third set is the central data when the nine data are arranged in numerical order. Means, and the value of the second data of the second group is larger than the value of the third data of the first group, and the value of the first data of the third group is smaller than the value of the second data of the second group, the first group A fourth judging means for judging the data having the largest value among the third data, the second data set, the third data set, and the third data set, as the central data when the nine data items are arranged in numerical order. A data processing device comprising: 2. A data processing method for obtaining a value of central data when 9 pieces of data are arranged in numerical order, wherein 9 pieces of data are divided into 3 sets of 3 pieces, and each set is divided into 3 sets. The first data, the second data, and the third data are sorted in descending order of numerical value, and the first data set, the second data set, and the third data set are arranged in descending order of the second data value of each sorted data set. When the value of the second data of the second group is larger than the value of the third data of the first group and the value of the first data of the third group is larger than the value of the second data of the second group, the second data of the second group Is determined as the central data when the above 9 data are arranged in numerical order, and the value of the second data of the second group is smaller than the value of the third data of the first group, and the value of the second data of the second group. When the value of the first data of the third set is smaller than that of the second data of the second set If the value of the second data of the second group is smaller than the value of the third data of the first group and the value of the first data of the third group is larger than the value of the second data of the second group, The data with the largest value among the first set third data, the second set third data, and the third set first data is determined to be the central data when the nine data are arranged in numerical order, and If the value of the second data of the second set is larger than the value of the third data of the set and the value of the first data of the third set is smaller than the value of the second data of the second set, the third data of the first set, A data processing method, characterized in that the data having the largest value among the second data of the third set and the first data of the third set is determined to be the central data when the nine data are arranged in numerical order. 3. A data processing device for obtaining a median value of 9 data, wherein 9 data are divided into 3 sets of 3 data, and each set is in ascending or descending order of data values. A first processing means for rearranging, a first group having the largest median value, a second group having a second largest median value of the first group, and a third group having a smallest median value. And a second processing means for checking the magnitude relationship between the minimum value of the first set and the median value of the second set, and the magnitude relationship between the median value of the second set and the maximum value of the third set. When the minimum value of the first set is greater than or equal to the median value of the second set and the median value of the second set is greater than or equal to the maximum value of the third set by the second processing means, or the minimum value of the first set is the second value. If the median value of the second set is less than or equal to the maximum value of the third set, the median value of the second set is determined to be the median value of the above nine data The second determining means and the second processing means, when the minimum value of the first set is larger than the median value of the second set and the maximum value of the third set is larger than the median value of the second set, The minimum value of the set and the maximum value of the third set and the second
Third judging means for judging the minimum value of the maximum values of the set as the median value of the nine data, and the minimum value of the first set being smaller than the median value of the second set by the second processing means. And the maximum value of the third set is smaller than the median value of the second set, the minimum value of the first set, the maximum value of the third set, and the second value
4. A data processing device, comprising: a fourth judgment means for judging the maximum value of the minimum values of the set to be the median value of the nine pieces of data. 4. An image processing apparatus in which, for each pixel of an input image, the median value of nine pixel values in a 3 × 3 size window centered on each pixel is the pixel value of each pixel in the output image. In, the pixels in the window are arranged in the row or column direction 3
When divided into three sets of pixels, one set which is not included in the previous window centering on the processed pixel adjacent to the direction different from the above direction among the three sets of windows is arranged in ascending or descending order. A first processing means for changing, a storage means for holding a result of rearranging the two sets included in the previous window in ascending or descending order when processing the processed pixels, and a maximum median value. Of the first set, the second set having the next largest median value of the first set, and the third set having the smallest median value; Second processing means for checking the magnitude relationship between the two groups of median values and the magnitude relationship between the median value of the second group and the maximum value of the third group, and the minimum value of the first group by the second processing means When the median of the second set is greater than or equal to the median of the second set and is greater than or equal to the maximum of the third set , Or if the minimum value of the first set is less than or equal to the median value of the second set and the median value of the second set is less than or equal to the maximum value of the third set, then the median value of the second set is set to The second judging means for judging the median value and the second processing means make the minimum value of the first set larger than the median value of the second set and the maximum value of the third set larger than the median value of the second set. If the minimum value of the first set, the maximum value of the third set, and the second
Third determining means for determining the minimum value of the maximum values of the set as the median value of the nine pixels, and the minimum value of the first set being smaller than the median value of the second set by the second processing means. And the maximum value of the third set is smaller than the median value of the second set, the minimum value of the first set, the maximum value of the third set, and the second value
An image processing apparatus comprising: a fourth determination unit that determines the maximum value of the minimum value of the set as the median value of the nine pixels. 5. An image processing apparatus in which, for each pixel of an input image, the median value of nine pixel values in a 3 × 3 size window centered on each pixel is the pixel value of each pixel in the output image. In, a storage means for storing pixels in three rows or three columns of an input image, and the stored pixels are divided into groups of three pixels arranged in the column direction or the row direction, and the pixel values are arranged in ascending or descending order for each group. A first processing means for rearranging to, a first group with the largest median value, a second group with the next largest median value of the first group, and a third group with the smallest median value. 1, a second processing means for checking the magnitude relationship between the minimum value of the first set and the median value of the second set, and the magnitude relationship of the median value of the second set and the maximum value of the third set; By this second processing means, the minimum value of the first set is equal to or greater than the median value of the second set and the median value of the second set is the third set. If the minimum value of the first set is less than the median value of the second set and the median value of the second set is less than or equal to the maximum value of the third set, the median value of the second set is greater than or equal to the maximum value. Is determined as the median value of the nine pixels, and the second processing means determines that the minimum value of the first set is greater than the median value of the second set and the maximum value of the third set is the maximum value. If greater than the median of the two sets, the minimum value of the first set and the maximum value of the third set and the second value
Third determining means for determining the minimum value of the maximum values of the set as the median value of the nine pixels, and the minimum value of the first set being smaller than the median value of the second set by the second processing means. And the maximum value of the third set is smaller than the median value of the second set, the minimum value of the first set, the maximum value of the third set, and the second value
An image processing apparatus comprising: a fourth determination unit that determines the maximum value of the minimum value of the set as the median value of the nine pixels. 6. An image processing apparatus, wherein, for each pixel of an input image, a median value of four pixel values in a 2 × 2 size window including each pixel is set as a pixel value of each pixel in the output image, An image processing apparatus comprising: a processing unit that selects any three pixel values from the number of pixels and sets a median value of the three pixel values as a pixel value of an output image. 7. An image processing apparatus, wherein for each pixel of an input image, a median value of four pixel values in a 2 × 2 size window including each pixel is set as a pixel value of each pixel in the output image, Pixels in the window are divided into two groups, each group consisting of two pixels, and a first processing means for checking the magnitude relationship between two pixel values for each group, and two larger pixel values for each group. The magnitude relationship between the smaller output pixel value of the two pixel values and the first output means for outputting the smaller pixel value of the two pixel values and the smaller size value of the two smaller pixel values of each set are checked, and the two pixel values are checked. Second output means for outputting the larger pixel value of the values, and second processing means for averaging the output of the first output means and the output of the second output means as the pixel value of the output image An image processing apparatus comprising: 8. A data processing method for obtaining a median value of 9 data, wherein 9 data are divided into 3 sets, each set of 3 data, and each set is arranged in ascending or descending order of data values. Rearrangement, the first group with the largest median value, the second group with the next largest median value of the first group, and the third group with the smallest median value are determined, The relationship between the median value of the second set and the relationship between the median value of the second set and the maximum value of the third set is examined, and the minimum value of this first set is equal to or greater than the median value of the second set and the second set. The median of is greater than or equal to the maximum of the third set, or the minimum of the first is less than or equal to the median of the second and the median of the second is less than or equal to the maximum of the third. The median value of the second set is determined as the median value of the nine data, and the minimum value of the first set is larger than the median value of the second set and the maximum value of the third set is the second set. When it is larger than the median value, the minimum value among the minimum value of the first set, the maximum value of the third set and the maximum value of the second set is determined as the median value of the above nine data, and the minimum value of the first set is determined. If the value is less than the median value of the second set and the maximum value of the third set is less than the median value of the second set, then the minimum value of the first set, the maximum value of the third set, and the minimum value of the second set Is determined as the median value of the above nine pieces of data. 9. An image processing method in which, for each pixel of an input image, a median value of nine pixel values in a 3 × 3 size window centered on each pixel is set as a pixel value of each pixel in the output image. In, the pixels in the window are arranged in the row or column direction 3
When divided into three sets of pixels, one set which is not included in the previous window centering on the processed pixel adjacent to the direction different from the above direction among the three sets of windows is arranged in ascending or descending order. In other words, for the two groups included in the previous window, the results of rearranging in ascending or descending order when the processed pixels are processed are retained, and the first group having the maximum median and the next group after the first group are retained. The second group having a large median value and the third group having a minimum median value are determined, and the magnitude relationship between the minimum value of the first group and the median value of the second group and the median value of the second group are determined. If the minimum value of the first set is greater than or equal to the median value of the second set and the median value of the second set is greater than or equal to the maximum value of the third set, or The minimum value of the set is less than or equal to the median value of the second set, and the median value of the second set is less than or equal to the maximum value of the third set. And the median value of the second set is determined as the median value of the nine pixels, and the minimum value of the first set is greater than the median value of the second set and the maximum value of the third set is the median value of the second set. If it is larger than the value, the minimum value among the minimum value of the first set, the maximum value of the third set and the maximum value of the second set is determined as the median value of the nine pixels, and the minimum value of the first set is determined. Is smaller than the median value of the second set and the maximum value of the third set is smaller than the median value of the second set, among the minimum value of the first set, the maximum value of the third set and the minimum value of the second set. An image processing method, characterized in that the maximum value is determined to be the median value of the nine pixels. 10. An image processing method in which, for each pixel of an input image, a median value of nine pixel values in a 3 × 3 size window centered on each pixel is set as a pixel value of each pixel in the output image. In, the pixels of three rows or three columns of the input image are stored, and the stored pixels are arranged in the column direction or the row direction.
Divide into groups of pixels, sort each group in ascending or descending order, the first group with the largest median value, the second group with the next largest median value of the first group, and the second group with the smallest median value. It is determined that there are three groups, and the magnitude relationship between the minimum value of the first group and the median value of the second group and the magnitude relationship between the median value of the second group and the maximum value of the third group are examined. If the minimum value is greater than or equal to the median value of the second set and the median value of the second set is greater than or equal to the maximum value of the third set, or the minimum value of the first set is less than or equal to the median value of the second set and of the second set When the median value is less than or equal to the maximum value of the third set, the median value of the second set is determined to be the median value of the nine pixels, and the minimum value of the first set is lower than the median value of the second set. If it is larger and the maximum value of the third set is larger than the median value of the second set, the minimum value of the minimum value of the first set, the maximum value of the third set and the maximum value of the second set is set to 9 pixels above. Judged as the median of However, when the minimum value of the first set is smaller than the median value of the second set and the maximum value of the third set is smaller than the median value of the second set, the minimum value of the first set and the maximum value of the third set An image processing method, characterized in that the maximum value among the minimum values of the second set is judged to be the median value of the nine pixels. 11. An image processing method in which, for each pixel of an input image, a median value of four pixel values in a 2 × 2 size window including each pixel is used as a pixel value of each pixel in the output image, An image processing method, wherein any three pixel values are selected from the pixels and the median value of the three pixel values is set as the pixel value of the output image. 12. An image processing method, wherein for each pixel of an input image, a median value of four pixel values in a 2 × 2 size window including each pixel is set as a pixel value of each pixel in the output image, Pixels in the window are divided into two groups, each group consisting of two pixels, and the magnitude relationship between the two pixel values in each group is examined, and the magnitude relationship between the two larger pixel values in each group is examined. The smaller pixel value of the two pixel values is output, the magnitude relationship between the two smaller pixel values of each set is checked, and the larger pixel value of the two pixel values is output, An image processing method, wherein an average of the two output pixel values is set as a pixel value of an output image.