JP2572067B2 - Defocus device - Google Patents

Description

Description: BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a two-dimensional digital filter for image processing, and more particularly to a defocus device using a two-dimensional recursive digital filter having a linear phase characteristic.

[Prior Art] Conventionally, as a technique for obtaining a defocus effect by a digital filter, a method using a two-dimensional transversal filter for performing a moving average calculation of neighboring pixels is known. In this two-dimensional transversal filter, a delay line group having a plurality of taps is arranged two-dimensionally in a horizontal and vertical manner. These tap outputs are multiplied by an arbitrary weighting factor, and by adding them, It is intended to obtain a desired impulse response. That is, by appropriately selecting the arrangement of the taps and the weighting coefficient, a characteristic close to the optical defocus characteristic can be obtained.

Further, as a method realized by a simpler device, there is known a method in which a two-dimensional low-pass filter is configured by a recursive digital filter and is used. this is,
The value delayed by one delay line is multiplied by an appropriate coefficient and fed back to the input to provide a first-order low-pass filter characteristic, which is applied in the horizontal and vertical directions to remove spatial harmonic components. To form a two-dimensional digital filter.

[Problems to be Solved by the Invention] However, the above-described two-dimensional transversal filter can provide a filter close to the optical defocus characteristic, but requires a huge amount of circuitry to increase the defocus amount. There are drawbacks.

Further, the two-dimensional recursive filter described above can obtain a large defocus effect with a small amount of circuit, but since the phase characteristic does not become linear, when processing is performed on an image, the processing result becomes horizontal and vertical. There is a disadvantage of becoming asymmetric. Furthermore, since the impulse response of the first-order low-pass filter is significantly different from the impulse response of optical defocus, there is a disadvantage that the defocused image becomes unnatural.

Therefore, an object of the present invention is to construct a two-dimensional digital filter having a characteristic close to optical defocus and having a linear phase characteristic, similarly to the two-dimensional transversal filter, using a recursive filter.
An object of the present invention is to solve the disadvantages of the conventional method, that is, the asymmetry in image processing and the unnaturalness of the defocus effect, with a small number of circuits.

[Means for Solving the Problems] In order to achieve the above object, the present invention provides a recursive digital filter having a linear phase characteristic when performing real-time defocus processing on a digital image signal. What is claimed is: 1. A defocusing device for operating at least one of a vertical component, a horizontal component, and an oblique component of an image signal, wherein said recursive digital filter is delayed by one clock period and / or one horizontal scanning period. And feedback delay means for providing
A recursive portion that provides an infinite loop with a coefficient of 1 and a transversal portion that cancels the response of the infinite loop to zero after a lapse of a clock period and / or a horizontal scanning period according to a predetermined defocus amount. This is a cascade connection in which a recursive part is connected to a stage subsequent to the versal part.

[Function] The present invention relates to a recursive digital filter having a linear phase characteristic for performing an average calculation of peripheral pixels in a real-time defocus processing apparatus used to obtain the same effect as optical defocus on a digital image signal. A defocus processing device that achieves almost the same effect as optical defocus by applying digital, horizontal, vertical and oblique directions to digital image signals is realized with a small number of elements. It is.

That is, the recursive digital filter in the horizontal direction has a recursive portion having an infinite loop with a coefficient of 1 (a portion formed by 7 and 8 in FIG. 1) and the above-mentioned after a lapse of the number of clocks corresponding to the defocus amount. By connecting the transversal part (the part constituted by 5 and 6 in FIG. 1) for canceling the response due to the infinite loop to zero in cascade such that the recursive part is connected after the transversal part, Linear phase characteristic ".

Here, the “linear phase characteristic” means that the delay amount is the same for all frequency signal components included in the input signal. Therefore, when an image is processed by a filter having a linear phase characteristic, image distortion does not occur. Also,
The “linear phase characteristic” is equivalent to the impulse response being symmetrical. More specifically, as can be seen from the impulse response of the “present invention” shown in the middle part of FIG. 3, the response is bilaterally symmetric, and a recursive filter having a “linear phase characteristic” It is called a filter.

Similarly, each of the recursive digital filters in the vertical direction and the oblique direction includes a respective recursive part and transversal part.

Next, regarding the above-mentioned "cancellation of the response by the infinite loop to zero after the elapse of the number of clocks corresponding to the defocus amount" with reference to FIG.
Details will be described below.

As long as the filter output has a delayed negative component as shown in FIG. 2C, the impulse response is not symmetrical.
Since the delayed negative component is not output, a symmetrical impulse response can be obtained as a result.

That is, since the signal component of the input signal C of the infinite loop (see FIG. 1) is zero from time τ to 4τ in FIG. 2, the output signal of the infinite loop is the input signal (plus component) at time 0 which is a loop. The input signal at time 0 is repeatedly output as shown in FIG. 2 (e). At time 5τ,
Since the input signal C (see FIG. 1) of the infinite loop is a minus signal obtained by inverting and delaying the input signal at time 0, the plus signal delayed in the infinite loop is canceled and the output signal of the infinite loop is zero. Is output. Further, after time 6τ, the delayed zero output signal and the zero input signal (C signal) of the infinite loop are added, so that the delayed negative component is not output.

Thus, after the time has elapsed by the number of clocks corresponding to the defocus amount, the response by the infinite loop can be canceled to zero.

[Embodiment] Next, an embodiment of a defocus device to which the present invention is applied will be described with reference to the drawings.

As shown in FIG. 1, the present defocus device is configured by cascade connection of a horizontal direction filter unit 1, a vertical direction filter unit 2, and an oblique direction filter unit 3. Here, the diagonal direction filter unit 3 is switched on / off by switches 4A and 4B.

The horizontal filter unit 1 has three delay elements 5, 8 and 1
0, a subtractor 6, an adder 7, and a multiplier 9. Here, the delay amount of each delay element is as follows.

Now, the sampling interval τ of the input digital image signal
If m is a positive odd number that determines the degree of defocus and M is a predetermined upper limit of (m−1) / 2, the delay element 5 is m · τ, the delay element 8 is τ, The delay element 10 has a delay amount of {M− (m−1) / 2} · τ.

The vertical filter unit 2 includes three delay elements 11, 14 and 1
6, a subtractor 12, an adder 13, and a multiplier 15.
Here, the delay amount of each delay element is as follows.

Now, based on the time Τ for one scanning line of the input image,
Assuming that a positive odd number that determines the degree of defocus is n, and a predetermined upper limit of (n−1) / 2 is N, the delay element 11 is n · T, the delay element 14 is Τ, and the delay element 16 is ｛N-
(N-1) / 2｝ · Τ. Note that the ratio between the values of m and n is determined by the ratio between the vertical and horizontal lengths of the pixel.

The oblique direction filter unit 3 includes four delay elements 17, 20 and 21.
, 24, subtractors 18 and 22, adders 19 and 23, and multiplier 25. Here, the delay element 17 is 5 (Τ + τ), the delay element 20 is Τ + τ, the delay element 21 is 5 (Τ−τ), the delay element
24 has a delay amount of Τ−τ.

Next, details of the operation of the present defocus device will be described step by step. For reference, FIG. 2 shows the waveform of each part when an impulse is given as an input signal when m = 5 and M = 4.

The digital image data (FIG. 2A) applied to the input line A is delayed by m · τ by the delay element 5 and appears on the line B (FIG. 2B). Further, the signal on line B is subtracted from the signal on line A that has not been delayed by the subtractor 6 and output to line C (FIG. 2C).

Next, the signal of the line D (FIG. 2D) obtained by delaying the output of the adder 7 by .tau.
To obtain the signal of the line E (FIG. 2-e).

The signal of the line E is multiplied by the coefficient 1 / m by the multiplier 9 to obtain the signal of the line F (FIG. 2-f).

In this way, the moving average calculation of the pixel is performed in the horizontal direction. Since this is a signal whose phase is delayed by (m-1) τ / 2 in the horizontal direction, the same applies to any value of m. In order to compensate for the delay by the amount M · τ, the signal is delayed by {M− (m−1) / 2} · τ by the delay element 10 to obtain the signal of the line G as the output of the horizontal filter unit 1 (second
Figure-g).

Next, the output of the horizontal filter unit 1 is added to the vertical filter unit 2. In the vertical filter unit 2, the delay amount τ in the horizontal filter unit 1 is replaced by the delay amount Τ, and almost the same. Since the above operation is performed, details of the operation are omitted.

The signals filtered in the horizontal and vertical directions appear on the output line H in this manner. This signal is a signal obtained by performing an averaging operation of pixels in a square centered on a certain pixel.

The image obtained by the processing up to this point is close to optical defocus. However, while the two-dimensional impulse response of the optical defocus is circular, while the two-dimensional impulse response in the processing system up to this point is square, if processing is performed on an image having many horizontal and vertical components, a kind of Moire occurs, resulting in an unsightly image. Therefore, filtering in the oblique direction is further performed.

The signal on the output line H of the vertical filter unit 2 enters the oblique filter unit 3 when the switch 4 is tilted to the a side, and enters the delay element 26 when the switch 4 is tilted to the b side.

The oblique direction filter unit 3 includes delay elements 17 and 20, a subtractor 18,
The processing unit includes a processing unit in the diagonal right direction including the adder 19, a processing unit in the diagonal left direction including the delay elements 21 and 24, the subtractor 22, and the adder 23, and a multiplier 25. The defocus amount is set to an appropriate value, for example, 5 based on the balance between the naturalness of defocus and the degree of moiré.

Further, the delay elements 17 and 20 have delay amounts of 5 (Τ + τ) and Τ + τ, respectively, and the processing unit in the diagonally right direction applies the elements 5, 6, and
Performs the same operation as 7,8.

Delay elements 21 and 24 are 5 (Τ−τ) and Τ, respectively.
The processing unit in the diagonally left direction has a delay amount of −τ, and performs the same operation as that of the elements 5, 6, 7, and 8 of the horizontal direction filtering unit 1 on the pixel in the diagonally lower left direction.

Finally, the output of the diagonal direction filter unit 3 is obtained by multiplying by 1/25 by the multiplier 25.

Since passing through the oblique direction filter section 3 causes a delay of 4 °, when the switch 4 is tilted to the b side to make the oblique direction filter 3 through, the same amount of delay is performed by the delay element 26 to correct the phase. .

The defocus amount control unit 27 controls the overall defocus amount for the above filters. The defocus amount control unit 27 controls the defocus amounts m and n of the horizontal direction filter unit 1 and the vertical direction filter unit 2, and switches 4A and 4B in a range where the defocus amount is small.
To the b side, and when the defocus amount is large, switch
4A and 4B are tilted to the a side to perform control for performing oblique filtering.

[Effects of the Invention] The present invention provides a two-dimensional digital filter having a characteristic close to the characteristic of optical defocus and having a linear phase characteristic by using a recursive filter.
The disadvantages of the conventional method, that is, the unnaturalness of the asymmetry and defocus effect in image processing, have been eliminated, and natural defocus processing close to the effect of the optical system, symmetrical in the horizontal and vertical directions, has been made possible.

FIG. 3 shows a one-dimensional impulse response of an optical system, the present invention, and a conventional low-pass filter having a first order coefficient.
It can also be seen from the figure that the present invention has characteristics very close to those of the optical system as compared with the conventional technology.

Further, since a recursive filter is used in the present invention, the device can be configured with a very small number of elements. For example, in a horizontal filter, if a filter with m = 51 is to be realized by a transversal filter,
50 individual delay elements, 50 adders, and one multiplier are required for the delay amount τ. In contrast, in the present invention, the delay amount
One delay element of 51τ, one delay element of delay amount τ, one subtractor, one adder and one multiplier can be composed, and the number of elements can be dramatically reduced. Can be.

[Brief description of the drawings]

FIG. 1 is a block diagram showing an embodiment of a defocus processing apparatus according to the present invention, FIGS. 2 (a) to 2 (g) are timing diagrams showing signals of respective parts shown in FIG. 1, and FIG. FIG. 4 is a comparative explanatory diagram of an impulse response of the defocus device of FIG. 1 horizontal filter section, 2 vertical filter section, 3 diagonal filter section, 4A, 4B changeover switch, 26 delay element, 27 defocus amount control section,
m: Defocus amount (positive odd number), M: (m-1)
/ 2 upper limit, n ... defocus amount (positive odd number), N ...
... Upper limit of (n−1) / 2, τ... Sampling interval of input image data, Δ... Time for one scanning line of input image.

Claims (1)

(57) [Claims] When performing real-time defocus processing on a digital image signal, a recursive digital filter having a linear phase characteristic is provided with at least one of a vertical component, a horizontal component, and an oblique component of the digital image signal. A defocusing device for acting on a component in one direction, wherein the recursive digital filter includes feedback delay means for delaying by one clock period and / or one horizontal scanning period, and a recursive part for providing an infinite loop of coefficient 1. After a lapse of a clock period and / or a horizontal scanning period according to a predetermined defocus amount, a transversal portion for canceling the response by the infinite loop to zero is connected to a recursive portion at a stage subsequent to the transversal portion. Characterized by continuous connection Defocus device.