JP2010055268A - Two-dimensional space filter device - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To prevent a circuit scale in a two-dimensional space filter device from increasing by reducing the number of multipliers without retaining a large amount of pixel data during a computing process. <P>SOLUTION: First to fourth multiplier-accumulators 5, 7, 9, 11 multiply individual coefficients corresponding to individual pixels for one column in an auxiliary scanning direction on a symmetrical axis of a filter matrix and on the left and right sides thereof, and accumulate the integration results to multiply and accumulate all the columns. Shifters Z1, Z2, Z3, Z4 shift the multiplication and accumulation result data of the respective columns. A first total adder 17 adds up the whole multiplication and accumulation result data shifted by the shifters Z1-Z4, such that numerical values whose absolute values in symmetrical positions of the columns are equal are multiplied. <P>COPYRIGHT: (C)2010,JPO&INPIT

Description

  The present invention relates to a two-dimensional spatial filter device, and has a function of two-dimensionally scanning an image using a predetermined spatial filter and filtering the image, such as a multi-function peripheral (MFP: Multi Function Peripheral). The present invention relates to an improvement of a two-dimensional spatial filter device suitable for being mounted on an image forming apparatus.

  Conventionally, in an image forming apparatus, various filtering processes such as an image smoothing process and an edge extraction process are performed using a two-dimensional spatial filter.

  As an example of the filter process, for example, a process of simply smoothing an image, as shown in FIG. 7A, an image composed of a large number of pixels S0, S1,. A filter matrix (spatial filter or operator) in which numerical values (filter coefficients) are assigned to each element arranged in three elements vertically and horizontally is superimposed around the pixel S0, and each element and each pixel superimposed on each other In addition, the numerical values in FIG. 5B and the pixel data (S0 to S8) are integrated, and the added data are added together.

That is, the following equation (1) is calculated,
S0 ′ = 1 × S1 + 1 × S2 + 1 × S3 + 1 × S4 + 1 × S0 + 1 × S5 + 1 × S6 + 1 × S8 + 1 × S8 (1)
The pixel data S0 ′ after this calculation is used as new pixel data corresponding to the pixel S0.

  In addition to such filter processing, there are those that perform weighted smoothing processing and those that perform Sobel arithmetic processing useful for extracting edge components.

  However, in order to perform such filter processing, for example, in the filter matrix in which numerical values are assigned to each of the three elements in the vertical and horizontal directions, as described above, nine 3 × 3 multipliers are required. There is a problem that the circuit scale tends to increase.

  Thus, for example, a two-dimensional spatial filter circuit disclosed in Japanese Patent Laid-Open No. 5-135169 (Patent Document 1) has been proposed.

That is, in Patent Document 1, among the plurality of elements constituting the operator, the numerical values assigned to the plurality of elements that are symmetrical with respect to the vertical axis, the horizontal axis, or the diagonal diagonal axis are multiplied by the same absolute value. A plurality of pixel data is added to each other by inverting the sign of one of the plurality of pixel data as necessary, and the number of elements of the conventional filter is required. Therefore, it is possible to reduce the number of multipliers having a large circuit scale, and to reduce the circuit scale.
Japanese Patent Laid-Open No. 5-135169

  However, when the above-mentioned Patent Document 1 has a 3 × 3 filter matrix configuration as described above, for example, 9 × 3 pixel data is input and multiplied by the same coefficient using symmetry. Since the pixel data is added first, and the added pixel data is multiplied by the coefficient, it is necessary to input nine pixel data, and the input pixel is delayed by one pixel from now. Furthermore, it is necessary to prepare image data delayed by two pixels, and the configuration is still complicated, and there is room for simplification of the circuit configuration.

  The present invention has been made to solve such a problem, and it is easy to reduce the number of multipliers and multipliers without holding a large amount of pixel data to be filtered, and it is difficult to increase the circuit scale. An object is to provide a dimensional spatial filter device.

  In order to solve such a problem, the first configuration according to the two-dimensional spatial filter device of the present invention is such that a plurality of elements arranged vertically and horizontally are arranged symmetrically with respect to each other with a predetermined axis as a symmetry axis. A filter matrix to which numerical values having the same absolute value are assigned is superimposed on the individual pixels forming the image, and the individual numerical values of the filter matrices that are superimposed on each other are added to those pixels. Is a two-dimensional spatial filter device that sequentially adds and adds the pixels, and adds and adds the pixels to each pixel corresponding to one column in the sub-scanning direction on the left and right sides of the symmetrical target axis. A plurality of product-sum calculators that multiply the corresponding individual coefficients and sum the results to perform product-sum operation for each column, and a product-shift result data for each column are respectively shifted. And the sum of product-summation result data obtained by summing up the sum-of-products result data obtained by shifting each of these shifters and the numerical values having the same absolute value at the symmetrical positions of the respective columns. 1 total adder.

  In the first configuration according to the present invention, a configuration in which the shifter is formed by a pipeline is possible.

  A second configuration according to the two-dimensional spatial filter device of the present invention is a filter in which a plurality of elements arranged vertically and horizontally are assigned numerical values having the same absolute value at vertically symmetrical positions with a predetermined axis as a symmetry axis. The matrix is superimposed around the individual pixels forming the image, the individual values of the filter matrix superimposed on each other are added to those pixels, the pixel data after this addition are added together, and the pixels are sequentially added. A two-dimensional spatial filter device that performs integration and addition by changing, and for each pixel for one column in the sub-scanning direction, a plurality of pixels for which the results are equal to each other are summed up and down the target axis A summing unit, a plurality of summing units for summing up the corresponding individual coefficients with respect to the summation result data from each of the summation units, and a summation unit for summing the summation result data. Are provided and the total adder, a.

  The 3rd structure concerning the two-dimensional spatial filter apparatus of this invention has the 1st and 2nd structure mentioned above, and is comprised.

  In the first configuration according to the two-dimensional spatial filter device of the present invention as described above, each of one column in the sub-scanning direction is individually detected by a plurality of product adders on the left and right sides of the symmetrical target axis. The corresponding individual coefficients are added to the pixels and the results are summed, and the product-sum result data of each column is shifted by a plurality of shifters, and each of these shifters is shifted to the product-sum result data to be shifted. In addition, since the sum of product-sum operation result data in which numerical values having the same absolute value at the symmetrical positions of the respective columns are integrated is added by the first total adder, in the filter processing for the symmetrical pixels, the filter Even without holding a large amount of pixel data to be processed, it is easy to reduce the number of multipliers and multipliers, and it is difficult to increase the circuit scale.

  In the first configuration according to the present invention, in the configuration in which the shifter is formed by a pipeline, the configuration can be further simplified.

  In the second configuration according to the two-dimensional spatial filter device of the present invention, for each pixel in one column in the sub-scanning direction, a plurality of adders above and below the target axis are arranged with pixels having the same result. , The corresponding individual coefficients are added to the summation result data from each of the adders by a plurality of integrators, and the summation results data are added together by the second summation calculator. Therefore, in the filtering process for vertically symmetric pixels, it is easy to reduce the number of multipliers and multipliers without holding a large amount of pixel data to be filtered, and it is difficult to increase the circuit scale.

  Since the third configuration according to the two-dimensional spatial filter device of the present invention has the first and second configurations described above, in the filtering process for any of the vertically symmetric and left-right symmetric pixels, the filter Even without holding a large amount of pixel data to be processed, it is easy to reduce the number of multipliers and multipliers, and the circuit scale can be greatly reduced.

  Hereinafter, an embodiment of a two-dimensional spatial filter device according to the present invention will be described with reference to the drawings.

  FIG. 1 is a schematic block diagram showing a first configuration according to the two-dimensional spatial filter device of the present invention, which uses a symmetrical filter matrix on the left and right.

  First, the filter matrix will be described before describing the first configuration of the two-dimensional spatial filter device of the present invention.

  FIG. 2 shows an image made up of 49 pixels (A1, A2,..., G6, G7) of 7 × 7 pixels in the form of a matrix, and FIG. 3A shows a Gaussian filter as an example. FIG. 3B shows the filter coefficients corresponding to the filter matrix (G0A1 to G0G7) in the form of a coefficient matrix. It shows.

  As can be seen from FIG. 3B, the coefficient matrix in the Gaussian filter has four symmetry axes, left, right, up, down, and diagonal, and 49 filter coefficients are 10 filter coefficients as shown in FIG. Can be filtered and optimized.

  In FIG. 2, the pixel (D4) in the image is the central pixel, and the filter matrix (G0D4) is overlapped with the pixel (D4) so that the pixel data of the central pixel (D4) is filtered. If it is obtained as the pixel-of-interest data (GAU), it is originally obtained by the following equation (2), but in the configuration of the present invention, it is obtained by a simplified method as described later.

  GAU = A1 * G0A1 + A2 * G0A2 + A3 * G0A3 + A4 * G0A4 + B1 * G0B1 + B2 * G0B2 + ... G7 * G0G7 (2)

  In FIG. 1, the buffer 1 is for each pixel of one column in the sub-scanning direction (vertical direction in the figure: A1, B1, C1, D1, E1, F1, G1 directions) in the pixels (A1 to G7) in FIG. Are input in the order of A1 to B1 to G1, and output to the buffer 3 in the FIFO (first-in first-out) system in the order of input of A1 to G1.

  The buffer 1 outputs the pixel data (A1 to G1) for one column to the buffer 3, and when the pixel data (A1 to G1) is output from the buffer 3, the pixel data (A2 to G2) for the next column. ) Are output to the buffer 3, and this is sequentially repeated.

  The buffer 3 temporarily stores pixel data (A1 to G1, A2 to G2, A3 to G3, A4 to G4, A5 to G5, A6 to G6, A7 to G7) for each pixel for one column in the sub-scanning direction. Are formed from flip-flops FFA, FFB, FFC, FFD, FFE, FFF, FFG, and the like.

  When the buffer 3 completes temporarily holding the pixel data (A1 to G1) for one column, the buffer 3 outputs the pixel data (A1 to G1) for one column to the first product-sum calculator 5 at a time, The next column of pixel data (A2 to G2) is input from the buffer 1 and output to the second product-sum calculator 7, and the next column of pixel data (A3 to G3) is input to the buffer 1. Are output to the third product-sum calculator 9, and the pixel data (A 4 to G 4) for the next one column are input from the buffer 1, and are input to the fourth product-sum calculator 11. It has the function to output to and repeat this sequentially.

  The first product-sum calculator 5 integrates the pixel data (A1 to G1) for one column and the filter coefficients (see FIG. 3B) of the corresponding filter matrix (G0A1 to G0G7), and adds the integrated value. Thus, the product-sum operation result data is output to the first shifter Z1.

  That is, the filter coefficient (0) of the filter matrix (G0A1) is integrated with the pixel data (A1), the filter coefficient (1) is added to the pixel data (B1), and the filter coefficient (2) is added to the pixel data (C1). However, the filter coefficient (4) for the pixel data (D1), the filter coefficient (2) for the pixel data (E1), the filter coefficient (1) for the pixel data (F1), and the pixel data (G1). The filter coefficients (0) are sequentially integrated, and these integrated values are added.

  Similarly, the second product-sum calculator 7 integrates the pixel data (A2 to G2) for one column and the filter coefficients of the corresponding filter matrix (G0A2 to G0G2), and adds the integrated value. It has a function of outputting to the second shifter Z2.

  Similarly, the third product-sum calculator 9 integrates the pixel data (A3 to G3) for one column and the filter coefficients of the corresponding filter matrix (G0A3 to G0G3), and adds the integrated value. The product-sum operation result data is output to the third shifter Z3.

  The fourth product-sum calculator 11 integrates the pixel data (A4 to G4) for one column and the filter coefficients of the corresponding filter matrix (G0A4 to G0G4), and adds the integration values to perform product-sum calculation. The result data is output to the fourth shifter Z4.

  For example, as shown in FIG. 5, the first to fourth product-sum calculators 5 to 11 apply filter coefficients of the filter matrix (G0A1 to G0G1) to the pixel data from the flip-flops FFA to FFG forming the buffer 3. Can be integrated by the integrators 13a, 13b, 13c, 13d, 13e, 13f, and 13g, and the sum of the integrated values can be calculated by the adder 15.

  Returning to FIG. 1, the first shifter Z1 is formed by connecting eight unit shifters Z1a, Z1b, Z1c, Z1d, Z1e, Z1f, Z1g, Z1h, which are conventionally known shift registers, in series. 1 has the function of sequentially shifting the product-sum operation result data input from the product-sum-adder 5 at a predetermined timing, and from the second and eighth unit shifters Z1b, Z1h. The product-sum operation result data is output to the first total adder 17.

  The product-sum calculation result data from the second unit shifter Z1b corresponds to pixel data (A1 to G1), and the product-summation result data from the eighth unit shifter Z1h is pixel data (A7 to G7). It corresponds to.

  The second shifter Z2 is formed by serially connecting seven unit shifters Z2a, Z2b, Z2c, Z2d, Z2e, Z2f, and Z2g, which are similar shift registers, and is input from the second product-sum calculator 7. The unit-sum shifters Z2a to Z2g are sequentially shifted at a predetermined timing, and the sum-of-products result data is sent from the third and seventh unit shifters Z2c and Z2g to the first A function of outputting to the total adder 17 is provided.

  The product-sum calculation result data from the third unit shifter Z2c corresponds to the pixel data (A2 to G2), and the product-summation result data from the seventh unit shifter Z1g is the pixel data (A6 to G6). It corresponds to.

  The third shifter Z3 is formed by serially connecting six unit shifters Z3a, Z3b, Z3c, Z3d, Z3e, and Z3f made of the same shift register. The unit shifters Z3a to Z3f are sequentially shifted at a predetermined timing with the summation result data, and the product-summation result data is first summed from the fourth and sixth unit shifters Z3d and Z3f. The function of outputting to the device 17 is provided.

  The product-sum operation result data from the fourth unit shifter Z3d corresponds to pixel data (A3 to G3), and the product-summation result data from the sixth unit shifter Z1f is pixel data (A5 to G5). It corresponds to.

  The fourth shifter Z4 is formed by serially connecting five unit shifters Z4a, Z4b, Z4c, Z4d, and Z4e made of similar shift registers, and the product-sum calculation input from the fourth product-sum calculator 11 The function has a function of sequentially shifting the result data at predetermined timings to the unit shifters Z4a to Z4e, and a function of outputting the product-sum calculation result data from the fifth unit shifter Z4e to the first total adder 17. is doing.

  The product-sum operation result data from the fifth unit shifter Z4e corresponds to pixel data (A4 to G4).

  The first total adder 17 has a function of inputting the product-sum calculation result data from the shifters Z1, Z2, Z3, and Z4 and calculating and outputting the sum, and outputs the total sum of the filter matrix. To do.

  Next, the operation of the first configuration described above, which is an example in which the filter matrix is symmetrical, will be briefly described.

  Pixel data (A1 to G1, A2 to G2, A3 to G3, A4 to G4, A5 to G5, A6 to G6, and A7 to G7) of each pixel for one column in the sub-scanning direction are sequentially input to the buffer 1. Then, pixel data for each pixel for each column is sequentially output to the buffer 3.

  In the buffer 3, each column of pixel data (A1 to G1, A2 to G2, A3 to G3, A4 to G4, A5 to G5, A6 to G6, A7 to G7) is temporarily stored in the flip-flops FFA to FFG. When the holding is completed, the pixel data (A1 to G1) for one column is output to the first product-sum calculator 5 at a time, and the pixel data (A2 to G2) for the next column is the second The pixel data (A3 to G3) for the next column is output to the third product-sum calculator 9 and the pixel data (A4 to G4) for the next column is output. Is output to the fourth product-sum calculator 11.

  The first to fourth product-sum calculators 5 to 11 integrate pixel data (A1 to G4) for one column and filter coefficients (G0A1 to G0G4) corresponding to these, and add the integrated values. Then, the product-sum operation result data is output to the first to fourth shifters Z1 to Z4.

  The first shifter Z1 sequentially shifts the product-sum operation result data input from the first product-sum calculator 5 to the unit shifters Z1a to Z1h, and then the second and eighth unit shifters Z1b. , Z1h, the product-sum operation result data corresponding to the pixel data (A1 to G1) and the product-sum operation result data corresponding to the pixel data (A7 to G7) are output to the first total adder 17.

  The second shifter Z2 sequentially shifts the product-sum calculation result data input from the second product-sum calculator 7 to the unit shifters Z2a to Z2g, and then the third and seventh unit shifters Z2c. , Z2g, the product-sum operation result data corresponding to the pixel data (A2 to G2) and the product-sum operation result data corresponding to the pixel data (A6 to G6) are output to the first total adder 17.

  The third shifter Z3 sequentially shifts the product-sum calculation result data input from the third product-sum calculator 9 to the unit shifters Z3a to Z3f, and the fourth and sixth unit shifters Z3d. , Z3f, the product-sum operation result data corresponding to the pixel data (A3 to G3) and the product-sum operation result data corresponding to the pixel data (A5 to G5) are output to the first total adder 17.

  That is, from the first to third shifters Z1 to Z3, product-sum calculation result data relating to the pixels on the left or right side of the symmetrical target axis is obtained.

  The fourth shifter Z4 sequentially shifts the product-sum operation result data input from the fourth product-sum calculator 11 from the fifth unit shifter Z4e in the process of shifting the unit shifters Z4a to Z4e. The product-sum calculation result data corresponding to A4 to G4) is output to the first total adder 17.

  That is, from the fourth shifter Z4, the product-sum operation result data relating to the pixels on the symmetrical target axis is obtained.

  The first total adder 17 inputs the product-sum calculation result data from the first to fourth shifters Z1 to Z4, calculates the sum, and outputs it. This total value is used as, for example, the calculated value of the pixel D4 in FIG. 2, and is sequentially calculated in the same manner for the other pixels.

  Therefore, a filter can be realized by shifting the result data obtained by multiplying the image data and the matrix table by one column without inputting and holding 49 pixel data of 7 × 7, and the integrator. The number is greatly reduced from 49 to 28.

  As described above, the first configuration according to the two-dimensional spatial filter device of the present invention corresponds to the symmetrical target axis of the filter matrix and individual pixels for one column in the sub-scanning direction on the left or right side. The first to fourth product-sum calculators 5-11 for summing the individual coefficients and summing the results and summing the products for each column, and a plurality for shifting the product-sum result data for each column. 1 to 4 shifters (shifters) Z1 to Z4, and product-sum result data shifted from these shifters Z1 to Z4, and the absolute values of the data in the symmetrical positions of the columns are equal to each other. And a first total adder 17 for adding up all the product-sum calculation result data to which the numerical values are integrated.

  For this reason, it is not necessary to store a large amount of pixel data to be filtered in the calculation process, and it is easy to reduce the number of multipliers and multipliers and increase the circuit scale. There are difficult advantages.

  Further, regarding the number of unit shifters Z1 to Z4, when the total value of the filter matrix coefficients is smaller than the depth of the input image, the first configuration according to the present invention rather than creating 7 × 7 reference matrices. As described above, the structure for shifting the operation result is advantageous because the circuit scale is reduced.

  In addition, each shifter Z1 to Z4 can be formed by a conventionally known pipeline, which is easy to implement and contributes to a simpler configuration.

  Next, a second configuration according to the two-dimensional spatial filter device of the present invention will be described.

  FIG. 6 is a schematic block diagram showing a second configuration according to the two-dimensional spatial filter device of the present invention, which uses a vertically symmetric filter matrix. In addition, the same code | symbol is attached | subjected to the part which is common in FIG. 5 mentioned above.

  In FIG. 6, the buffer 3 is composed of the flip-flops FFA to FFG as described above, and inputs and temporarily holds the pixel data for each column from the buffer 1.

  In the buffer 3, the flip-flops FFA and FFE are connected to the adder 19a, the flip-flops FFB and FFF are connected to the adder 19b, and the flip-flops FFC and FFG are connected to the adder 19c. The adders 19a to 19c have a function of adding the respective pixel data and outputting them to the integrators 13a, 13b, 13c and 13d.

  The integrators 13a to 13d integrate the addition results from the adders 19a to 19c with the filter coefficients of the filter matrix (G0A1 to G0D1), and output the result to the total adder 15 as the second total adder. have.

  The total adder 15 has a function of calculating the total sum of the integration result data of the integrators 13a to 13d.

  Next, the operation of the above-described second configuration in which the filter matrix is vertically symmetric is briefly described.

  Pixel data (A1 to G1, A2 to G2, A3 to G3, A4 to G4, A5 to G5, A6 to G6, A7 to G7) for each pixel for one column are sequentially input to the buffer 3 through the buffer 1. Then, in the flip-flops FFA to FFG of the buffer 3, each column of pixel data (A1 to G1, A2 to G2, A3 to G3, A4 to G4, A5 to G5, A6 to G6, A7 to G7) Are temporarily stored, and pixel data (A1 to G1) for one column is output to the adders 19a to 19c.

  The adder 19a adds the respective pixel data (A1 to A7 and E1 to E7), and the adder 19b adds the respective pixel data (B1 to B7 and F1 to F7), and the adder 19c. Then, the respective pixel data (C1 to C7 and G1 to G7) are added and output to the integrators 13a to 13d.

  That is, from the adders 19a to 17c, summation result data obtained by integrating numerical values having the same absolute value at the symmetrical positions of the respective columns on the upper side or the lower side of the vertically symmetrical target axis is obtained.

  The accumulators 13a to 13d accumulate the filter coefficients of the filter matrix (G0A1 to G0D1) with respect to the addition results from the adders 19a to 17c, and in the total adder 15, integration result data from the accumulators 13a to 13d. Calculate the sum of.

  The total value from the second total adder 15 is used as, for example, the operation value of the pixel D4 in FIG. 2, and the other pixels are sequentially calculated in the same manner.

  Therefore, when performing product-sum of one column of image data and matrix table by using vertical symmetry, it is possible to reduce the number of integrators from 7 to 4, and as a whole, vertical symmetry Since there are four calculations using, the number of integrators can be reduced from 28 to 16.

  As described above, in the second configuration according to the two-dimensional spatial filter device of the present invention, for each pixel in one column in the sub-scanning direction, pixels having the same result are summed up and down the target axis. A plurality of summers 19a to 19c to be calculated, a plurality of multipliers 13a to 13d that accumulate respective individual coefficients corresponding to the summation result data from each of the summers 19a to 19c, and their integration results And a second total adder 15 for adding the whole data.

  Therefore, by utilizing the vertical symmetry of the Gaussian filter coefficients, it is easy to reduce the number of multipliers and multipliers without increasing the amount of pixel data to be filtered for vertically symmetrical pixels, and the circuit scale is increased. It is difficult to let it.

  By the way, in the two-dimensional spatial filter device of the present invention, it is possible to constitute the third configuration with the above-described first and second configurations, and for any pixel that is vertically symmetric or laterally symmetric. In the filtering process, it is easy to greatly reduce the number of multipliers and multipliers without holding a large amount of pixel data to be filtered, and there is an advantage that the circuit scale can be easily reduced.

  Note that the two-dimensional spatial filter device of the present invention is not limited to the Gaussian filter, and can be implemented for various filter processes.

It is a schematic block diagram which shows embodiment of the 1st structure which concerns on the two-dimensional spatial filter apparatus of this invention. It is a figure explaining the matrix used for the two-dimensional spatial filter apparatus of this invention. It is a figure explaining the matrix used for the two-dimensional spatial filter apparatus of this invention. It is a figure explaining the matrix used for the two-dimensional spatial filter apparatus of this invention. It is a block diagram which shows the structural example of the product-sum calculator used for the two-dimensional spatial filter apparatus of FIG. It is a schematic block diagram which shows embodiment of the 2nd structure which concerns on the two-dimensional spatial filter apparatus of this invention. It is a figure explaining operation | movement of a general two-dimensional spatial filter apparatus.

Explanation of symbols

1, 3 Buffer 5 First product-sum calculator 7 Second product-sum calculator 9 Third product-sum calculator 11 Fourth product-sum calculator 13a, 13b, 13c, 13d, 13e, 13f, 13g 15 Adder (second total adder)
17 1st total adder 19a, 19b, 19c Adder Z1 1st shifter Z1a, Z1b, Z1c, Z1d, Z1e, Z1f, Z1g, Z1h, Z2a, Z2b, Z2c, Z2d, Z2e, Z2f, Z2g , Z3a, Z3b, Z3c, Z3d, Z3e, Z3f, Z4a, Z4b, Z4c, Z4d, Z4e Unit shifter Z2 Second shifter Z3 Third shifter Z4 Fourth shifter

Claims (4)

A filter matrix in which numerical values having the same absolute value are assigned to symmetrical positions with respect to a predetermined axis as the symmetry axis is superimposed on the elements arranged vertically and horizontally around the individual pixels forming the image. A two-dimensional spatial filter device that integrates the numerical values of the filter matrices superimposed on each other, adds the pixel data after the integration to each other, sequentially changes the pixels, and performs the integration and the addition And
On the left and right sides of the symmetrical target axis, the corresponding individual coefficients are added to the individual pixels for one column in the sub-scanning direction, and the results are summed to sum the products for each column. A plurality of multiply-accumulate arithmetic units,
A plurality of shifters that respectively shift the product-sum result data of each column;
The product sum result data shifted by each of these shifters, wherein the sums of product sum results obtained by adding the numerical values having the same absolute value at the symmetrical positions of the respective columns are added together. An adder;
A two-dimensional spatial filter device comprising: The two-dimensional spatial filter device according to claim 1, wherein the shifter is formed of a pipeline. A filter matrix in which a numerical value having the same absolute value is assigned to a vertically symmetrical position with respect to a predetermined axis as a symmetry axis is superimposed on a plurality of elements arranged vertically and horizontally, centering on individual pixels forming an image. A two-dimensional spatial filter device that integrates the numerical values of the filter matrices superimposed on each other, adds the pixel data after the integration to each other, sequentially changes the pixels, and performs the integration and the addition And
A plurality of adders for summing up and down the target axis pixels for which the results are equal for each column of pixels in the sub-scanning direction;
A plurality of integrators for integrating the corresponding individual coefficients with respect to the summation result data from each of these summers;
A second total adder for adding up all the integration result data;
A two-dimensional spatial filter device comprising: A two-dimensional spatial filter device comprising the configuration according to claim 3 in the configuration according to claim 1 or 2.

Images