KR950002073B1 - Three dimensional filter - Google Patents

Abstract

The method and circuit improves the image data process system, reduces the bit error with data transfer and regeneration process, and simplifies the hardware. The circuit includes the average value computing means which computes the average value between image data, the count generation means which generates the variation coefficients, the 1st multiplication means which miltiplies the variation coefficient to the fixed coefficient, the 2nd multiplication means, the 1st addtion means, the 2nd addition means, and the division means which divides the output signal.

Description

3D filter method and circuit

Figure 1 shows a block diagram of a three-dimensional filter circuit of one embodiment of the present invention.

2 shows original image data of an embodiment of the present invention.

3 shows data including noise of an embodiment of the present invention.

4 shows data including noise of another embodiment of the present invention.

FIG. 5 shows data obtained by average of pixel data shown in FIG. 3 and FIG.

FIG. 6 shows data after performing interpolation of the data shown in FIG.

The present invention relates to a method and circuit for filtering image data, and more particularly, to a three-dimensional filter method and circuit for image data.

Representative examples of three-dimensional data include video data that changes in time, that is, video data. In the image data, a low pass filter or a median filter can be used to remove Gaussian random noise or salt-and-pepper noise. Although one filter is known to be excellent, it has a problem that the hardware implementation for arranging the data in the filter area in order of size is complicated.

The low pass filter requires a large number of filter taps in the case of the FIR filter for a sufficient noise canceling effect, which leads to hardware complexity. On the other hand, in the case of the IIR filter, the number of filter taps can be reduced, but the group delay is not linear, which is not suitable for filtering image data.

The median filter shows an excellent filtering effect using a relatively small number of adjacent pixels, but also has a computational difficulty in arranging all the pixels in size order.

Accordingly, an object of the present invention is to provide a three-dimensional filter method for filtering using a value according to the degree of isolation between neighboring pixels on one field and a value according to the stability of one field with another field. There is.

Another object of the present invention is to provide a three-dimensional filter circuit that can reduce the hardware.

In the three-dimensional filtering method of the present invention for achieving the above object, in the filtering method for filtering the image data, the average value calculation step of calculating the average value with the image data of the upper, lower, left and right by inputting the image data, the average value and the To generate an isolation coefficient for generating the isolation coefficient according to the image data of the top, bottom, left and right and the difference coefficient according to the difference value between the image data of the top, bottom, left and right and the image data of the same position of the previous field. A coefficient generation step, multiplying the coefficients to generate a first value by adding the multiplied coefficients, multiplying the self-mechanical coefficients, multiplying the multiplied coefficients by the image data of the top, bottom, left and right, and adding the multiplied values to a second value Compute and divide the second value by the first value to obtain a final filtered output. Characterized in consisting of steps.

The three-dimensional filter circuit of the present invention for achieving the another object is a filtering circuit for filtering the image data, the average value calculating means for obtaining the average value by inputting the image data, the average value and input the average value Generating an isolation coefficient according to a difference from adjacent image data up, down, left, and right, and generating a gradient coefficient according to a difference between the image data of the previous field displayed at the same position as the image data and the image data of the top, bottom, left, and right. Coefficient generating means for multiplying the coefficients and adding the multiplied coefficients to generate a first value, multiplying the coefficients, multiplying the multiplied coefficients by the image data of the top, bottom, left and right, and adding the multiplied values to a second value And dividing the second value by the first value to obtain a final filtered output. For is constituted by comprising a calculating means.

Referring to the accompanying drawings, the three-dimensional filter method and circuit of the present invention will be described.

First, the three-dimensional filter method will be described.

First, the average of adjacent pixels on top, bottom, left and right is obtained.

Second, the degree of isolation according to the difference between the average value and the pixels of up, down, left, and right in one field is output, and the stability according to the difference between pixels of one field and the other field is output.

Third, the isolation degree and the stability are inputted and the following equation is performed.

Referring to the three-dimensional filter circuit of the present invention using the above method as follows.

1 shows a block diagram of the three-dimensional filter circuit of the present invention.

1, a storage means 10 for storing pixel data, a storage means 20 for storing an output signal of the storage means 10, and a 1H delay means 30 for delaying the pixel data 1H. ), A register 40 for storing the output signal of the 1H delay means 30, a register 50 for storing the output signal of the register 40, a register for storing the output signal of the register 50; 60, a 1H delay means 70 for delaying the output signal of the 1H delay means 30 again by 1H, a register 80 for storing the output signal of the 1H delay means 70, and the register 80 Register 90 for storing the output signal of the signal, an adder 100 for adding the output signal of the registers 60, 90, and an adder for adding the output signal of the registers 20, 40 ( 110, an adder 120 and the register for adding output signals of the adders 100 and 110. An adder 130 for adding the output signal of the emitter 50 and the output signal of the adder 120, a field memory 140 for storing the pixel data in field units, and an output signal of the field memory 140. 1H delay means 150 for delaying 1H, a register 160 for storing the output signal of the 1H delay means 150, an output signal of the storage means 20 and an output signal of the adder 130. Degree of variation for generating stability coefficient according to the difference between the isolation coefficient generating means 170 and the output signal of the storage means 20 and the output signal of the storage means 160 for generating the isolation coefficient according to the difference value. Coefficient generator 180, the isolation coefficient generator 190 for generating an isolation coefficient according to the difference between the output signal of the storage means 40 and the output signal of the adder 130, the storage means 40 Side according to the difference between the output signal of the control signal and the output signal of the storage means 160 An isolation coefficient generating means for generating an isolation coefficient according to a difference between the gradient coefficient generating means 200 for generating the degree coefficient and an output signal of the storage means 60 and an output signal of the adder 130 ( 210, the gradient coefficient generating means 220 for generating the gradient coefficient according to the difference between the output signal of the storage means 60 and the output signal of the storage means 160, the output of the storage means 90 The isolation coefficient generating means 230 for generating the isolation coefficient according to the difference between the signal and the output signal of the adder 130, the output signal of the storage means 90 and the output signal of the storage means 160. A gradient coefficient generator 240 for generating a gradient coefficient according to a difference, a multiplier 250 for multiplying an output signal of the isolation coefficient generator 170 and the gradient coefficient generator 180, and Isolation coefficient generating means 190 and the gradient coefficient generating means 200 A multiplier 260 for multiplying an output signal, a multiplier 270 for multiplying an output signal of the isolation coefficient generating means 210 and the gradient coefficient generating means 220, and the isolation coefficient generating means 230; A multiplier 280 for multiplying the output signal of the gradient coefficient generator 240, a multiplier 290 for multiplying the output signal of the storage means 20, and an output signal of the multiplier 250, the storage means ( A multiplier 300 for multiplying the output signal of the multiplier 260 and an output signal of the multiplier 260, a multiplier 310 for multiplying the output signal of the storage means 60 and the output signal of the multiplier 270, the storage A multiplier 320 for multiplying the output signal of the means 90 with the output signal of the multiplier 280, an adder 330 for adding the output signals of the multipliers 250, 260, the multipliers 270, An adder 340 for adding the output signals of 280, the adders 330, 3 An adder 350 for adding the output signals of 40, an adder 360 for adding the output signals of the multipliers 290, 300, and an adder for adding the output signals of the multipliers 310, 320; 370, an adder 380 for adding the output signals of the adders 360 and 370, and a distribution means 390 for dividing the output signal of the adder 370 into the output signal of the adder 380. Consists of. In the above configuration, the 1H delay means stores the pixel data of one horizontal line using a line memory, and the 1V delay means stores the pixel data of one field using a field memory.

The operation according to the above configuration is as follows.

The storage means 20, 40, 50, 60, 90 store up, down, left, and right data of the pixel data to be processed. The adders 100, 110, and 120 input and add output signals of the storage means 20, 40, 60, and 90 to obtain an average by discarding the least significant two bits, and the adder 130 adds the adder 120. The final average is obtained by inputting and adding the output signal of the < RTI ID = 0.0 >)< / RTI > The 1V delay means 140 stores pixel data corresponding to the previous field. The 1H delay means 140 stores the pixel data corresponding to the previous field. The 1H delay means 150 stores data corresponding to 1H of the 1V delay means 140. The storage means 160 stores one pixel data from the 1H delay means 150. The isolation coefficient generating means 170, 190, 210, 230 compares the data from the storage means 20, 40, 50, 60, 90 with the value from the adder 130 to isolate the difference according to the difference value. Generates degree coefficients. Gradient coefficient generating means 180, 200, 220, 240 correspond to the same position of the data from the storage means 20, 40, 50, 60, 90 and the previous field from the storage means 160. The gradient coefficient according to the difference value with the pixel data is generated. The multiplier 250 outputs a result (a ₁ , b ₁ ) multiplied by the output coefficients a ₁ , b ₁ from the isolation coefficient generating means 170 and the gradient coefficient generating means 180. The multiplier 260 outputs a result (a ₂ , b ₂ ) of the isolation coefficient generating means 190 and the output signals a ₂ , b ₂ from the gradient coefficient generating means 200. The multiplier 270 outputs a result (a ₃ , b ₃ ) of the isolation coefficient generating means 210 and the output signals a ₃ , b ₃ from the gradient coefficient generating means 220. The multiplier 280 outputs a result (a ₄ , b ₄ ) of the isolation coefficient generating means 230 and the output signals a ₄ , b ₄ from the gradient coefficient generating means 240. The multiplier 290 outputs a result (a ₁ b ₁ x ₁ ) of the pixel data x ₁ multiplied by the output signal a ₁ b ₁ of the multiplier 250. The multiplier 300 outputs a result (a ₂ b ₂ x ₂ ) of the pixel data x ₂ multiplied by the output signal a ₂ b ₂ of the multiplier 260. The multiplier 310 outputs a result (a ₃ b ₃ x ₃ ) of the pixel data x ₃ multiplied by the output signal a ₃ b ₃ of the multiplier 270. The multiplier 320 outputs a result (a ₄ b ₄ x ₄ ) of the pixel data x ₄ multiplied by the output signal a ₄ b ₄ of the multiplier 280. The adder 330 outputs the result (a ₁ b ₁ + a ₂ b ₂ ) of the output signals a ₁ b ₁ and a ₂ b ₂ of the multipliers 25 and 260. The adder 340 outputs the result of adding the output signals of the multipliers 270 and 280 (a ₃ b ₃ , a ₄ b ₄ ) (a ₃ b ₃ + a ₄ b ₄ ). Adder 350 is the output signals of the adders _{(330, 340) (a 1} b 1 + a 2 b 2, a 3 b 3 + a 4 b 4) the result of adding the (a ₁ b ₁ + a ₂ b ₂ + a ₃ b ₃ + a ₄ b ₄ ) The adder 360 adds the output signals a ₁ b ₁ x ₁ and a ₂ b ₂ x ₂ of the multipliers 290 and 300 (a ₁ b ₁ x ₁ + a ₂ b ₂ x ₂ ). Outputs The adder 370 adds the result of adding the output signals of the multipliers 310 and 320 (a ₃ b ₃ x ₃ , a ₄ b ₄ x ₄ ) (a ₃ b ₃ x ₃ + a ₄ b ₄ x ₄ ) The adder 380 outputs the output signals of the adders 360 and 370.

Output the result (a ₁ b ₁ x ₁ + a ₂ b ₂ x ₂ + a ₃ b ₃ x ₃ + a ₄ b ₄ x ₄ ). The dividing means 360 divides the output signal a ₁ b ₁ x ₁ + a ₂ b ₂ x ₂ + a ₃ b ₃ x ₃ + a ₄ b ₄ x ₄ of the adder 380 from the adder 350. The output signal (a ₁ b ₁ + a ₂ b ₂ + a ₃ b ₃ + a ₄ b ₄ ) is divided and output. Through the above process, the filtered value x of the pixel data x ₀ to be processed is obtained.

The noise canceling function of an embodiment of the present invention will be described with reference to FIGS. 2 to 6 as follows.

2 shows normal pixel data without noise. 2 shows an image composed of 7 * 7 pixels, with 9 pixels in the center having brightness of 100 and the rest having values of 20. FIG.

3 and 4 illustrate two temporally successive images in which different Gaussian random noises are applied to the image of FIG. 2.

FIG. 5 shows an image of the average value output from the adder 130 shown in FIG. 1 when the image shown in FIG. 3 is input. At this time, the average value of the top, bottom, left and right edges will be ignored. If ₀ is the (3, 3) th pixel, the mean value x _m = 82 corresponds to this value, and each a ₁ , a ₂ , a ₃ , a ₄ is Obtained together.

a ₁ = 0, a ₂ = 0, a ₃ = 2, a ₄ = 2

TABLE 1

Further, x _old = 100 is obtained from FIG. 4, and thus the change coefficients b ₁ , b ₂ , b ₃ , and b ₄ are obtained as follows according to Table 2 below.

b ₁ = 1, b ₂ = 1, b ₃ = 4, b ₄ = 8

TABLE 2

6 shows the output image of the final filter. It can be seen that the noise is greatly reduced, unlike the use of a simple average filter as shown in FIGS. In addition, it can be seen that even in the sharp edge portion that changes from 20 to 100, it is preserved without deterioration.

The second embodiment of the present invention is described as follows.

In the image stabilization system, a method of obtaining a field motion vector from a plurality of local motion vectors that change in time will be described.

First, the 1V storage means 140 and the 1H storage means 150 located at the front of FIG. 1 need not be used. Instead, a register for storing respective local motion vectors and a register for storing the motion vector of the previous field are needed.

For example, four local motion vectors are (-3, -1), (-3, -1), (5, 2), (-3, -1), and the field motion vector of the previous field is (0). , 0), the averaged circuit calculates the mean values x _m and y _m for the x and y directions, respectively, and the isolation coefficient generating means and the gradient coefficient generating means in the x direction for each local motion vector. A weighting factor is generated for the sum of the difference absolute value and the difference absolute value in the y direction. In this example, y _m = -1 and x _m = 0. Thus, in the case of the first local motion vector, the arithmetic mean and the absolute difference value in the y direction are 2, the absolute difference value in the x direction is 0, the sum thereof is 2, and as a result, the isolation coefficient is 8. Since the value of the field motion vector of the previous field is (0, 0), the absolute value of the y direction difference between the value and the first local motion vector (-3, -1) is 3, and the absolute value of the difference value of the x direction is 1. The gradient coefficient is 4. By this method, we obtain a ₂ = 8, b ₂ = 4, a ₃ = 1, b ₃ = 2, a ₄ = 8, and b ₄ = 4 for the remaining local motion vectors. You get an average value of (-3, -1).

The isolation coefficient generation is generated according to Table 3 below.

TABLE 3

Gradient coefficient generation is generated according to Table 4 below.

TABLE 4

Gradient coefficient generation is generated according to Table 4 below.

The above-described isolation coefficient generating means and gradient coefficient generating means obtain a absolute value of the difference between two input data and generate a coefficient according to the magnitude of the value, and the isolation coefficient generally increases as the difference absolute value increases. It has a declining characteristic.

Therefore, the three-dimensional filter method and circuit according to the present invention can firstly increase the reliability of the image data processing system.

Second, it is possible to reduce bit errors that can occur during data transfer or restoration.

Third, it does not use a circuit for classifying data in order of magnitude like a median filter and can simplify hardware.

Claims (18)

A filtering method for filtering image data, comprising: an average value calculating step of inputting image data to obtain an average value with up, down, left and right image data; An isolation coefficient for generating an isolation coefficient according to a difference value between the average value and the image data of up, down, left, and right, and a degree of change according to a difference value between the image data of the top, bottom, left, and right and the image data of the same position of the previous field A coefficient generation step for generating a coefficient; Multiply the coefficients and add the multiplied coefficients to generate a first value, multiply the coefficients and multiply the multiplied coefficients by the image data of the top, bottom, left and right to add the multiplied values to generate a second value And a calculating step of dividing two values by the first value to obtain a final filtered output. The three-dimensional filter method of claim 1, wherein the isolation coefficient is a coefficient indicating how much a difference between neighboring pixels of image data displayed at the same time is. The three-dimensional filter method of claim 1, wherein the gradient coefficient is a coefficient indicating how much difference exists between data displayed at the same position of consecutively displayed image data. A filtering circuit for filtering video data, the filtering circuit comprising: an average value calculating means for inputting the video data to obtain an average value with up, down, left, and right video data; An isolation coefficient for generating an isolation coefficient according to a difference value between the average value and the image data of up, down, left, and right, and a degree of change according to a difference value between the image data of the top, bottom, left, and right and the image data of the same position of the previous field Coefficient generating means for generating a coefficient; First multiplication means for multiplying each of the isolation coefficients and the gradient coefficients from the coefficient generating means; Second multiplication means for multiplying the signal from the first multiplication means and the image data of the top, bottom, left and right, respectively; First adding means for adding signals from the first multiplication means; Second adding means for adding signals from said second multiplication means; And a dividing step for dividing the output signal from the second adding means into the output signal from the first adding means. A filtering circuit for filtering image data, comprising: an average value calculating means for calculating an average value by inputting the image data and adjacent data up, down, left, and right with the image data; Isolation coefficient generation means for generating an isolation coefficient according to a difference value between the average value and the image data of up, down, left, and right sides; Gradient coefficient generating means for generating gradient coefficients according to the difference value between the image data in the top, bottom, left and right and the image data indicated at the same position of the previous field; First multiplication means for multiplying the isolation coefficient generating means and the output signals from the gradient coefficient generating means, respectively; Second multiplication means for multiplying the top, bottom, left and right image data and output signals of the first multiplication means; First adding means for adding output signals of the first multiplication means; Second adding means for adding output signals of the second multiplication means; Third adding means for adding output signals of the first adding means; Fourth adding means for adding output signals of the second adding means; And distribution means for dividing an output signal of said fourth adding means into an output signal of said third adding means. 6. The apparatus of claim 5, wherein the average value calculating means comprises: first storage means for storing pixel data; Second storage means for storing the output signal of the first storage means; First H delay means for delaying the pixel data by 1 H; A first register for storing an output signal of the first 1H delay means; A second register for storing an output signal of the first register; A third register for storing an output signal of the second register; Second 1H delay means for delaying the output signal of the first 1H delay means by 1H; A fourth register for storing the output signal of the second 1H delay means; A fifth register for storing the output signal of the fourth register; A first adder for adding output signals of the fifth and seventh registers; A second addition time for adding output signals of the second and third registers; A third adder for adding output signals of the first and second adders; A fourth adder for adding an output signal of the fourth register and an output signal of the third adder; A field memory for storing the pixel data in field units; Third 1H delay means for delaying the output signal of the field memory by 1H; And an eighth register for storing the output signal of the third 1H delay means. 6. The apparatus of claim 5, wherein the isolation coefficient generating means comprises: first isolation coefficient generating means for generating an isolation coefficient according to a difference value between an output signal of the second storage means and an output signal of the fourth adder; Second isolation coefficient generating means for generating an isolation coefficient according to a difference between an output signal of the third storage means and an output signal of the fourth adder; Third isolation coefficient generating means for generating an isolation coefficient according to a difference between an output signal of the fifth storage means and an output signal of the fourth adder; And a fourth isolation factor coefficient generating means for generating an isolation coefficient according to the difference between the output signal of the seventh storage means and the output signal of the fourth adder. 6. The apparatus of claim 5, wherein the gradient coefficient generating means comprises: first gradient coefficient generating means for generating a stability coefficient according to a difference between an output signal of the second storage means and an output signal of the eighth storage means; Second gradient coefficient generating means for generating gradient coefficients according to the difference between the output signal of the third storage means and the output signal of the eighth storage means; Third gradient coefficient generating means for generating gradient coefficients according to the difference between the output signal of the fifth storage means and the output signal of the eighth storage means; And a fourth gradient coefficient generator for generating gradient coefficients according to the difference between the output signal of the seventh storage means and the output circuit of the eighth storage means. 6. The apparatus of claim 5, wherein the first multiplication means comprises: a first multiplier for multiplying the first isolation coefficient generating means and the output signal of the first gradient coefficient generating means; A second multiplier for multiplying the output signal of the second isolation coefficient generating means and the second gradient coefficient generating means; A third multiplier for multiplying by the output signal of the third gradient coefficient generating means and the third isolated coefficient coefficient generating means; And a fourth multiplier for multiplying the output signal of the fourth isolation coefficient coefficient generating means and the fourth gradient coefficient generating means. 6. The apparatus of claim 5, wherein the second multiplication means comprises: a fifth multiplier for multiplying the output signal of the second storage means with the output signal of the first multiplier; A sixth multiplier for multiplying the output signal of the third storage means by the output signal of the second multiplier; A seventh multiplier for multiplying the output signal of the fifth storage means by the output signal of the third multiplier; And an eighth multiplier for multiplying the output signal of the seventh storage means by the output signal of the fourth multiplier. 6. The apparatus of claim 5, wherein the first adding means comprises: a fifth adder for adding output signals of the first and second multipliers; And a sixth adder for adding output signals of the third and fourth multipliers. 6. The three-dimensional filter circuit of claim 5, wherein the second adding means comprises a seventh adder for adding output signals of the fifth and sixth adders. 6. The apparatus of claim 5, wherein the third adding means comprises: an eighth adder for adding output signals of the fifth and sixth multipliers; And a ninth adder for adding output signals of the seventh and eighth multipliers. 3. The three-dimensional filter circuit according to claim 2, wherein the fourth adding means comprises tenth adding means for adding output signals of the eighth and ninth adders. A filtering circuit for filtering image data, comprising: an average value calculating means for inputting the image data to obtain an average value; The average value is input to generate an isolation coefficient according to the difference between the image data and the image data adjacent up, down, left, and right, and with the image data of the upper, lower, left, and right of the image data of the previous field displayed at the same position as the image data. Coefficient generation means for generating a gradient coefficient according to the difference; Multiply the coefficients and add the multiplied coefficients to generate a first value, multiply the coefficients and multiply the multiplied coefficients by the image data of the top, bottom, left, and right to add the multiplied values to produce a second value and the first And a calculating means for dividing two values by the first value to obtain a final filtered output. 16. The apparatus of claim 15, wherein the average value calculating means comprises: first storage means for storing pixel data; Second storage means for storing the output signal of the first storage means; First 1H delay means for delaying the pixel data by 1H; A first register for storing an output signal of the first 1H delay means; A second register for storing an output signal of the first register; A third register for storing the output signal of the second register; Second 1H delay means for delaying the output signal of the first 1H delay means by 1H; A fourth register for storing the output signal of the second 1H delay means; A fifth register for storing the output signal of the fourth register; A first adder for adding output signals of the fifth and seventh registers; A third adder for adding output signals of the first and second adders to add output signals of the second and third registers; A fourth adder for adding an output signal of the fourth register and an output signal of the third adder; A field memory for storing the pixel data in field units; Third 1H delay means for delaying the output signal of the field memory by 1H; And an eighth register for storing the output signal of the third 1H delay means. The method of claim 15, wherein the coefficient generating means comprises: first isolation coefficient generating means for generating an isolation coefficient according to a difference between an output signal of the second storage means and an output signal of the fourth adder; Third isolation factor issuing means for generating an isolation coefficient according to a difference between an output signal of said third storage means, an output signal of said fourth adder and an output signal of said fourth adder; Second isolation coefficient generating means for generating an isolation coefficient according to a difference between an output signal of the seventh storage means and an output signal of the fourth adder; And a fourth isolatedness coefficient generating means for generating an isolation coefficient according to the difference of the output signal of said fifth storage means. 16. The apparatus of claim 15, further comprising: a first multiplier for multiplying said calculating means by said first isolation coefficient generating means and an output signal of said first gradient coefficient generating means; A second multiplier for multiplying the output signal of the second isolation coefficient generating means and the second gradient coefficient generating means; A third multiplier for multiplying the output signal of the third isolation coefficient generating means and the third gradient coefficient generating means; A fourth multiplier for multiplying the output signal of the fourth isolation coefficient generating means and the fourth gradient coefficient generating means; A fifth multiplier for multiplying the output signal of the second storage means by the output signal of the first multiplier; A sixth multiplier for multiplying the output signal of the third storage means by the output signal of the second multiplier; A seventh multiplier for multiplying the output signal of the fifth storage means by the output signal of the third multiplier; An eighth multiplier for multiplying the output signal of the seventh storage means by the output signal of the fourth multiplier; A fifth adder for adding output signals of the first and second multipliers; A sixth adder for adding output signals of the third and fourth multipliers; A seventh adder for adding output signals of the fifth and sixth adders; An eighth adder for adding output signals of the fifth and sixth multipliers; A ninth adder for adding output signals of the seventh and eighth multipliers; And a tenth adder for adding output signals of the eighth and ninth adders.