JPH0785259B2 - Space filter - Google Patents

Description

TECHNICAL FIELD The present invention relates to a digital filter, and more particularly to a spatial filter having a two-dimensional spread.

[Prior Art] Spatial filters for digital images are widely used for image feature extraction, noise removal, image sharpening, and the like.
This type of space filter is used, for example, as shown in FIG.
When it consists of × 5 elements, the circuits that realize them are 25 latch circuits 20 and 25 coefficient multipliers K _{11 to} K.
₅₅ and the adder circuit 21 is composed of four adders for adding the signals S ₁₁ and S ₂₁ , so that it is composed of a total of 24 adders, and the number of parts is very large. However, wiring between circuits and parts was complicated. Further, in FIG. 2, even if the coefficients are fixed and the coefficient multiplier is simplified in order to simplify the hardware, different hardware must be configured for each required filter characteristic.

[Problems to be Solved by the Invention] The conventional space filter has a very large number of parts, and the circuit and wiring between parts are complicated.

The present invention has been made in view of the above-mentioned conventional example, and an object of the present invention is to provide a space filter capable of forming several standard space filters with a small number of parts.

[Means for Solving the Problems] As one means for solving this problem, for example, the spatial filter of the embodiment shown in FIG. 1 has adders 5, 6, 7 and a selector 10 as a first filter circuit. , 11 and a one-dimensional filter 14 as a second filter circuit and selectors 11, 13 for selecting the output of the one-dimensional filter 14.

[Operation] In the configuration shown in FIG. 1, among the image data of different scanning lines input from the input ports 1 to 4, the input port 1
The image data of 2 is added by the adder 5. Adder 5
The selector 10 selects whether to delay the output of the above, and inputs it to the adder 7. On the other hand, the image data input from the input ports 3 and 4 are added by the adder 6 and multiplied by the coefficient by the coefficient unit 9. When the A side of the selectors 11 to 13 is selected, the output of the coefficient unit 9 is added to the image data of the input ports 1 and 2 by the adder 7 and output to the output port 15 through the one-dimensional filter 14. On the other hand, selectors 11 to 13
When the B side of is selected, the output of the coefficient unit 9 is calculated by the primary filter 14 and added to the output of the selector 10 to be output.

Embodiments Embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[Explanation of Spatial Filter Circuit (FIG. 1)] FIG. 1 shows a spatial filter circuit according to an embodiment of the present invention.
Reference numeral 4 is an input port for inputting image data, 5-7 are adders for performing addition between image data, 8 is a delay circuit for delaying image data by two pixels, and 9 is a coefficient unit for multiplying coefficient Km.
Reference numerals 10 to 13 are selectors for selecting and outputting one of the two input data A and B, 14 is a one-dimensional filter, and 15 is an output port for filtering image data.

According to the space filter of this embodiment, various space filters can be constructed as described later, but the core of the filter is described for the sake of explanation. Consider the smoothing filter. In this case, the selectors 10 to 13 are all set to select A.

The input image data is raster-scanned data, and the input port 1 has n−1, n, n + 1 ... Lines, the input port 2 has n, n + 1, n + 2 ... Lines, and the input port 3 has n + 1, n + 2, n + 3 ... Lines are input in synchronization with each other. The adder 5 successively adds the data of the input port 1 and the data of the input port 2, the output of which is input to the A side of the selector 10 and is connected to one input of the adder 7 as a signal line 16.

On the other hand, by connecting all the input lines of the input port 4 to the ground, the data of the input port 3 passes through the adder 6 as it is, multiplied by Km by the coefficient unit 9, and the other input of the adder 7 through the selector 11. Connected to. At this time Km
Is 1, the output signal 17 of the adder 7 becomes I ₀ + I ₁ + I ₂ . This signal 17 is input to the A side of the selector 12 and transmitted to the one-dimensional filter 14. The output of the one-dimensional filter 14 is transmitted to the output port 15 by the selector 13.

[Description of One-Dimensional Filter (FIG. 3)] FIG. 3 is a diagram showing details of the one-dimensional filter 14, wherein 30 to 34 are coefficient units, 35 is a latch, and 36 is an adder. In the case of the smoothing filter described above, the coefficients K _{0 to} K ₄ of the coefficient units 30 to 34 are sequentially set,
It may be set to 0,1,1,1,0. If the input of the one-dimensional filter 14 at time t is represented by Dt, the output 37 has K ₀ · Dt− ₄ + K ₁ · Dt− ₃ + K ₂・ Dt− ₂ + K ₃ · Dt− ₁ + K ₄ · Dt is obtained, so the output becomes Dt− ₃ + Dt− ₂ + Dt− ₁ according to the coefficient. Since the data of the (n-1) th line is input to the input port 1, the data of the nth line is input to the input port 2, and the (n + 1) th line is input to the input port 3, Dt is the data of the input ports In- ₁ and In. , In + ₁ , it is represented by In- ₁ + In + In + 1.

Therefore, the signal output to the output port 15 is O = In-1t-3 + In-1t-2 + In-1t-1 + Int-3 + Int-2 + Int-1 ++ In + 1t-3 + In + 1t-2 + In + 1t. −
It becomes 1.

[Explanation of another filter example (Fig. 1) (Fig. 3) (Fig. 4)
FIG. 4) FIGS. 4 (a) and 4 (b) show examples of cores of other filters that can be configured by the space filter of this embodiment.

FIG. 4 (a) shows the smoothing filter described above.

FIG. 4 (b) shows km = 2, K ₀ = 0.K ₁ = 1, K ₂ = 2, K ₃ = 1, K ₄ = 0.
It is a filter which is obtained when the above is added to emphasize noise removal and image sharpness by weighting. At this time, the input of the one-dimensional filter 14 is Dt = In− ₁ + ₁ + 2In + In + ₁ , the output of the one-dimensional filter 14 is Dt− ₃ + 2 · Dt− ₂ + Dt− ₁ , and the signal output to the output port 15 is O = In-1 t-3 + 2In-1 t-2 + In-1 t-1 + 2In t-3 + 4In t-2 + In t-1 + In + 1 t-3 + 2In + 1 t-2 + In-1 t
It is represented by -1, which can also be written as follows, and it can be seen that the filtering of FIG. 4 (b) is performed.

FIG. 4 (c) is also a smoothing filter, and in order to realize this, the selectors 10 to 13 of the embodiment in FIG. 1 are set and used so as to select the B side. Km = 1, K ₀ = 0, K ₁
= 1, K ₂ = 1, K ₃ = 1, K ₄ = 0, and I _{0 is connected} to the n− ₁ line, I _{1 is connected} to the n + ₁ line, I ₂ is connected to the n line, and I ₃ is connected to the ground. The signal of I ₂ passes through the adder 6, the coefficient unit 9, and the selector 12 to obtain Int-3 + Int-2 + Int-1 by the one-dimensional filter 14.

On the other hand, I ₀ and I ₁ are added by the adder 5, and the delay circuit 8
After being delayed by 2 pixels, the output of the one-dimensional filter 14 is added by the adder 7. The addition result is guided to the output port 15 through the data selector 13, and the following formula is obtained.

O = In-1 t-2 + In t-3 + In t-2 + In t-1 In + 1 t-2 FIG. 4 (d) is a commonly used second-order differential filter for image sharpening, and FIG. 4 (c). Similarly, the selectors 10 to 13 are set so that the B side is selected, and Km = -1, K ₀ = 0, K ₁ =
By setting -1, K ₂ = 4, K ₃ = -1, K ₄ = 0, O = In-1 t-2 + In t-3-4 In t-2 + In t-1 + In + 1 t-2 is obtained. Therefore, it is sufficient to apply a coefficient of -1 externally. This coefficient may be included in the output of the selector 13,
Alternatively, the adder 7 may be used as an adder / subtractor by providing the output of the selector 10 with a coefficient unit of -1.

Both FIG. 4 (e) and FIG. 4 (f) are well-known first-order differential filters for boundary detection, and are elements that constitute a prewrite operator. In FIG. 4 (e), the I ₀ (or I ₁ ) port and the I ₂ (or I ₃ ) 0 port are used, I ₀ is connected to the (n- _{1) th} line, I ₂ is connected to the (n + _{1) th} line, and the like. Leave the port on the ground. The selectors 10 to 13 are set so as to select the A side, and Km = −1, K ₀ = 0, K ₁ = 1, K ₂ = 1 and K ₃
= 1 and K ₄ = 0, the input Dt of one filter 14 is Dt = In−
_{Since 1−In} + ₁ , the output port 15 has O = In−1t−3 + In−1t−2 + In−1t−1 + 0 + 0 + 0 + In + 1t−3 + In + 1t−2 + In + 1t−.
It becomes 1.

In the case of FIG. 4 (f), the n- _1st , nth, and n + _1th lines are connected to I ₀ , I ₂ , and I ₁ , respectively, and the selectors 10 to 13 are set to select the A side, Km = 1, K ₀ , K ₀ = 0, K ₁ = 1, K ₂ = 0, K
_{When 3} = -1, K ₄ = 0, the input Dt of the one-dimensional filter is Dt
= In- ₁ + In + In + ₁ , and its output is Dt- _3- Dt- ₁ , so O = In-1t-3 + 0-In-1t-1 + Int-3 + 0-nt-1 at the output port 15. + In-1 t-3 +0 -In + 1 t-1 is output.

FIG. 4 (g) and FIG. 4 (h) are the elements constituting the sobel operator in the first-order differential filter. Figure 4 input connections and a selector (e), Figure 4 the same as FIG. 4 (f) (g) in n K ₂ = _2, that for the FIG. 4 (h) and Km = 2 Can be achieved with.

Although the well-known filter constructed by the 3 × 3 matrix has been described in the present embodiment, it is easy to construct a filter having other kernels, and the configuration of the present embodiment provides a 5 × 4 matrix. It is possible to carry out a filter having a core.

In the present embodiment, I ₂ and I _{3 are used} as the inputs of the first group, and I ₀ and I _{1 are used} as the inputs of the second group, but two multiplications are described. However, these may be independently one. However, 3 or more may be used. Although the one-dimensional filter has been described for five pixels, it does not impair the present invention as long as it has one pixel or more.

[Advantages of the Invention] As described above, according to the present invention, it is possible to provide a spatial filter having a small number of parts, easy wiring, and capable of forming various two-dimensional filters.

BRIEF DESCRIPTION OF THE DRAWINGS FIG. 1 is a block diagram of a space filter of this embodiment, FIG. 2 is a block diagram of a conventional 5 × 5 two-dimensional filter, and FIG. 3 is a one-dimensional filter of this embodiment. 4A to 4H are diagrams showing an example of a two-dimensional filter that can be realized in the present embodiment. In the figure, 1 to 4 ... input port, 5 to 7,36 ... adder, 8
...... Delay circuit, 9,30 to 34 …… Coefficient multiplier, 10 to 13 …… Selector, 14 …… One-dimensional filter, 15 …… Output port, 35 ……
It is a latch.

Claims (1)

[Claims] 1. Input means for inputting a plurality of digital image data of different scanning lines, and first filter means for adding and synthesizing predetermined ones of a plurality of image data input at the same time from said input means. A delay unit to which an output signal of the first filter unit is input; and a remaining image data not input to the first filter unit out of a plurality of image data input simultaneously from the input unit, A spatial filter comprising: a second filter means for combining; an adding means; a third filter means for filtering image data in the scanning line direction; and a data path selecting means, wherein the data path selecting means comprises: In the state of 1, the output of the first filter means and the output of the second filter means are input to the adding means, and the output of the adding means A signal is input to the third filter means, an output of the third filter means is set to be connected to an output port, and in a second state, an output of the first filter means is input to the delay means. The output of the second filter means is the input of the third filter means, the output of the delay means and the output of the third filter means are the inputs of the adding means, and the output of the adding means is A spatial filter characterized by being set to connect to an output port.