JPH0620300B2 - Progressive scan interpolator - Google Patents

Description

Description: TECHNICAL FIELD The present invention relates to a progressive scan interpolation device for improving the image quality of a device such as a television receiver that handles images.

(Prior Art) Currently, the standard method used in television broadcasting and the like is to perform 2: 1 interlace scanning,
Line flicker and a decrease in vertical resolution, which cause it, have become problems. Therefore, a method for solving the above-mentioned problem by interpolating thinned lines in an interlaced signal to form a progressive scanning (non-interlaced) signal is being studied.

Among them, the method of adaptively changing the interpolation method according to the level of the difference between frames of the image is effective because a high resolution can be obtained when the image is stationary and no big problem occurs even when the image moves. Is considered a method. The operation will be described with reference to FIG.

In the figure, the y direction is the vertical direction of the image, and the points A and D of the signal are the intervals y for one line in the interlace.
There is h. The t direction is the time change direction, and the point B and the point C have a time difference tf for one frame. Usually, this value is 1 /
30 seconds. For these points A, B, C and D, the point to be interpolated is located in the center of them and the point is designated as Z.

This interpolation point Z is interpolated by the respective values of A, B, C, D. Here, when the image is in a stationary state, since the signal does not change in the time direction, that is, in the time direction, that is, in the frame. Interpolate between. In this case, the values of B and C are basically the same value, but in consideration of noise reduction and the like, Z
= (B + C) / 2.

On the other hand, when the image is moving, the image is different in each field, so interpolation between frames cannot be performed. Therefore, interpolation is performed from the upper and lower lines of the same field, and Z = (A + D) / 2 may be set.

The above is the basic idea of motion adaptive interpolation, but if it is a simple switching operation it is not clear whether it is motion or still, or discontinuity at the switching point of operation becomes a problem, so
It is desirable to have an intermediate region and operate continuously. The interpolation method in this case is shown by the following equation.

Z = Zt + k (Zy−Zt) (1) where Zy = (A + D) / 2 Zt = (B + C) / 2 where k is motion information, and in the case of perfect stillness, k = 0.
And Z = Zt, and in the case of complete movement, k = 1 and Z = Z
It becomes y. If k is intermediate, Zy and Z
t are mixed.

The problem here is the method of setting k, that is, the detection of motion, but it is basically determined by the absolute value of the inter-frame difference and is expressed by the following equation.

k = F (| B−C |) (2) where F is “0” when | B−C | is “0”, and | B
It increases as -C | increases, and is "1" when it exceeds a certain value.
Becomes

(Problems to be Solved by the Invention) However, in the above-described conventional method, motion detection is performed for each pixel, which causes the following problems.
A prominent example occurs in an image in which a horizontal line having a width of several lines moves up and down. This is shown in FIG. Incidentally, A, B, C, D and Z in FIG. 7 correspond to FIG.

In FIG. 8, the horizontal direction is the vertical direction in the image, and F
The three kinds of signals ₁ , ₁ , F ₂ and F ₃ are three fields showing the change in the vertical direction of each field.

Here, a line with a width of 2 lines is 1 when interlaced.
Suppose you move 3 lines in a frame. Here, considering the interpolation in the central field, the inter-frame difference signal corresponding to each line is as shown in FIG. 9 (a). Here, since the line passes through the center point, there is no difference between frames, and that portion is interpolated by the signal Zt between frames. As a result, the interpolated signal becomes as shown at F ₂ ′ in FIG. 8, and incorrect interpolation is performed at the center of the line. This development occurs even in a portion having a high contrast ratio, resulting in extremely large image quality deterioration.

Such development is not limited to the case described above, but when an image having a high spatial frequency moves, even if the image actually moves, it is considered to be still when the value of the difference between frames happens to be similar, and This frequently occurs due to the interpolation between them.

In addition to noise, vibrations of the video camera and minute vibrations whose motion amount due to conversion from the image film is less than one sample interval are other causes of erroneous detection that is actually considered to be a moving area where still processing should be performed. is there. In this case, the sample value at the edge portion of the image fluctuates considerably, so that the adaptive processing becomes unnecessarily in a moving state and the improvement in image quality is suppressed.

On the other hand, as a method for more accurate motion detection, a method of dividing an image into blocks and performing pattern matching within the blocks is also being considered in high-efficiency coding of moving images, but in this case, block units are used. However, there is a problem that the processing becomes discontinuous at the block boundary and that an erroneous processing is easily performed on an image in which pulse noise is mixed.

SUMMARY OF THE INVENTION Therefore, an object of the present invention is to provide a progressive scanning interpolation device which solves the above-mentioned problems of the conventional technique.

(Means for Solving the Problems) In order to achieve the above object, the present invention provides a progressive scan interpolating device for forming a line to be interpolated in order to convert a television signal from interlaced scan to progressive scan, An absolute value circuit that takes the difference between pixel values between fields or frames around the line to be interpolated and an absolute value circuit, and the output of this absolute value circuit is input, and this input level is a value corresponding to the noise level from "0" Until the output becomes “0”, the gain increases from “0” as the input level increases from the value corresponding to the noise level, and when it further increases, the gain decreases, and the input level And a low-pass filter for the vertical component of the output through the first non-linear circuit, which has a characteristic that it has a saturation region where the output is a constant value above a certain level. Filter and a low-pass filter for the horizontal direction component, and a second non-linear circuit to which the output passed through these low-pass filters is input and which has the same characteristics as the first non-linear circuit, including a dead zone, an intermediate zone, and a saturation zone. A mixing that mixes the interpolated signal formed by the upper and lower lines in the same field as the line to be interpolated and the interpolated signal formed by the preceding and following fields by changing the respective ratios by the signal passed through the second non-linear circuit. The present invention provides a progressive scanning interpolating device comprising a circuit and obtaining an interpolation signal from the output of the mixing circuit.

(Operation) In the progressive scan interpolator with the above configuration, the motion information determined for each pixel has a fairly strong correlation in the spatial domain so that the motion information does not change abruptly to reduce erroneous operation. To do. This is because the moving and still areas of the image do not change rapidly pixel by pixel,
It is based on changes in a certain amount,
It is the same idea that processing can be performed in block units. The motion detection in this case is shown in FIG. 9 (b) as opposed to FIG.
As described above, appropriate interpolation is performed.

(Embodiment) In order to realize the above idea, it is necessary to apply a low-pass filter (LPF) in the spatial domain to the motion information obtained for each pixel, and the configuration diagram of that embodiment is shown in FIG.

In the figure, the input is a 262H delay circuit 1 (H is 1
It is supplied to one input terminal of the 1H delay circuit 2 and the adder 3 via the horizontal scanning period). The output of the 1H delay circuit 2 is supplied to the other input terminal of the adder 3 and the 262H delay circuit 4.

On the other hand, the inputs are supplied to the one input ends of the adder 5 and the subtracter 6, respectively, and the outputs of the 262H delay circuit 4 are supplied to the other input ends of the adder 5 and the subtractor 6, respectively.

The adders 3 and 5 add the respective inputs, multiply by 1/2 and output the results, which are supplied to the delay circuits 7 and 8, respectively.

The subtractor 6 subtracts the signal at the other input terminal (−) from the signal at the one input terminal (+), outputs the subtracted signal, and supplies this to the absolute value (ABS) circuit 9.

The output of the delay circuit 7 is supplied to the input terminal of the subtractor 10.
The output of the delay circuit 8 is supplied to the other input terminal of the subtractor 10 and one input terminal of the mixing circuit 11.

The subtractor 10 subtracts the signal at the input end (−) from the signal at the one input end (+), outputs the subtracted signal, and supplies this to the one input end of the multiplier 12.

On the other hand, the output of the ABS circuit 9 is supplied to the second non-linear circuit 16 via the first non-linear circuit 13, the vertical low-pass filter (LPF) 14 and the horizontal low-pass filter (LPF) 15.

The output of the second non-linear circuit 16 is supplied to the multiplier 12 as its multiplication coefficient (motion information) k [0 ≦ k ≦ 1].

The multiplier 12 multiplies the respective inputs and outputs the result, which is supplied to the other input end of the mixing circuit 11. Then, an interpolation line is obtained from the output of the mixing circuit 11.

2 is a diagram showing the characteristics of the first non-linear circuit 13,
FIG. 3 is a diagram showing the characteristic of the second nonlinear circuit 16.

Here, the base motion information is the absolute value of the interframe difference | B-C | from the output of the ABS circuit 9, but a method of using the values of A and D is also conceivable. In the processing, the calculated absolute value of the inter-frame difference is passed through the first non-linear circuit 13, the output thereof is subjected to the vertical LPF 14 and the horizontal LPF 15, and then passed through the second non-linear circuit 16 to interpolate the information. Used to change the method. Here, the interpolated value is passed through delay circuits 7 and 8 in order to compensate for the delays of the LPFs 14 and 15, and the timing is matched with the motion information.

Detailed description of each part will be given. First, the absolute value of the inter-frame difference | B-C | in FIG. 7 is obtained, and this is calculated as the first and second values.
The non-linear circuits 13 and 16 of
The second and non-linear circuits 13 and 16 are set based on the following idea.

The inter-frame difference signal basically has the same value for a still image, but actually contains noise, so that the output is "0" for an extremely small value corresponding to the noise level. To do. This level is related to the amount of noise and sensitivity to movement. If too much input level area with no output is taken, the output disappears and remains stationary when an image with little change in the spatial area moves. When the level is "0" to "255", the level is "0" to "6". For higher levels, the operation is close to linear, and the gain increases from “0” as the input level increases from the value corresponding to the noise level, and when it further increases, the gain decreases. To
Furthermore, when the input level is above a certain level, the output is limited to a saturation range where it becomes a constant value.

Thus, the first and second nonlinear circuits 13 and 16 are
It has a characteristic consisting of a dead zone, an intermediate zone, and a saturated zone.

The characteristic having such a saturation region is significantly different from the conventional pattern matching circuit having the square characteristic in this nonlinear circuit in order to obtain the mean square error of the difference signal. .

Here, the reason for having a characteristic having a saturated region is as follows. This is because, if pulsed noise is mixed in, if this circuit is made to have a linear or square characteristic, it is possible to prevent even the peripheral portion of the pixel from being regarded as a moving region in the subsequent LPF output.

On the other hand, this means that an isolated point in the frame difference signal is ignored, but when the image moves, the frame difference rarely appears as an isolated image, and it usually occurs to some extent continuously, so there is no problem.

In addition, even if the image moves in the same way, the frame difference varies greatly depending on the degree of spatial change in that part, and a large difference signal is obtained when the amplitude is large, but a small difference signal is obtained when the amplitude is small. If the characteristic having the above-mentioned saturation region is set to solve the above problem, there will be no significant difference between the two, and moving parts will be treated as the same signal. Further, in the case of a characteristic having a saturation region, there is an advantage that the number of operation bits when the later LPF is attempted to be realized by a digital filter is small and the circuit is simple.

Next, regarding the LPF, basically, this filter characteristic should be determined depending on how much spatial frequency component the change in the spatial region of the moving region and the stationary region has. It is desired that the band matches the band. However, in reality, the above-mentioned change greatly differs depending on the image, while the two-dimensional digital filter requires the use of a line delay circuit, so that there is not much freedom in design from the viewpoint of the device scale.

Therefore, as an example that can be constructed without using the line delay circuit much, there is one as shown in FIG. 4 (a).

The filter of FIG. 4 (a) has the same coefficient for each tap in order to narrow the band with as few line delay circuits as possible. {Note that FIG. 4 (b) shows the response (FIR). In the figure, 1H is a delay circuit (line delay circuit) for one line. }for that reason,
Although the characteristic has a large ripple, the characteristic of this filter does not directly affect each pixel value of the image and is used only for changing the interpolation method, so that it does not cause a big problem.

On the other hand, the delay circuit in this filter is 2H + 5T
Therefore, the delay compensation in FIG. 1 and the delay compensation for the original line are necessary accordingly. Further, even if the processing of FIG. 1 is performed only on the luminance signal in the processing of the luminance signal and the color difference signal, it is necessary to perform delay compensation on the color difference signal.

As described above, when the FIR type filter as shown in FIG. 4 (a) is used, although the characteristic is relatively ideal, many line delay circuits are required and the circuit scale is large. turn into.

Therefore, a method using an IIR filter only in the vertical direction is conceivable as a method of simplifying the circuit although it is slightly inferior in characteristics. This example is shown in FIG. 5 (a). In this case, only one line delay circuit is required for the filter, and it is not necessary for delay compensation. {Note that FIG. 5 (b) shows the response (IIR). } The characteristic in this case is asymmetric with respect to the vertical direction, and the filter works effectively when there is a moving region above the current line, but it does not work when there is only the moving region below the current line. Malfunctions are likely to occur at the top of the area. But,
In other parts, the same processing as that using the FIR type filter is performed, and since the horizontal filter is the FIR type, it is considered that the effect can be obtained without much difference.

On the other hand, in the horizontal LPF, the dot delay circuit is not so large, so that the filter having the same coefficient can be realized by the configuration as shown in FIG. 6 (a).
{Note that FIG. 6 (b) shows the response. In the figure,
T is a delay circuit for one dot (dot delay circuit), 1
1T is a delay circuit for 11 dots. Although this filter seems to be a recursive type, it is a non-recursive type filter. By adding a certain sample value and subtracting the same value after a fixed delay, a pair of adder and subtractor are used. A filter with an arbitrary number of taps can be constructed. Here, it is set to about 11 taps so as to have a spatial band equivalent to that of the LPF in the vertical direction.

Note that a filter with such a configuration is
Since many delay circuits are required, it is not a very suitable method in the vertical direction. On the other hand, it is possible to use a recursive horizontal filter, but this is not very desirable from the viewpoint of characteristics.

From such LPF output, k (motion information) used for actual adaptive processing is obtained. In this case, L
If the output of the PF is above a certain level, it is considered that the image is moving without fail, and at a value above that, k = 1 is set. The level setting is based on the assumption that the first nonlinear circuit 13
If the value after passing through each pixel is either “0” or “255”, about 1/15 to 1/10 of the pixels included in the filter tap are changed pixels, that is, the first pixel. When the output of the non-linear circuit 13 is "255", it is considered to be completely moving. If it is less than that, it is processed as an intermediate area.

When the output of LPF is very small, the output is set to "0".
By providing a non-linear circuit having a dead zone at the latter stage of the LPF, it is difficult to be affected by the difference between frames, which is not the motion of the image caused by noise or minute vibration of the image, and the motion is actually performed in a place where still processing should be performed. It is possible to prevent erroneous detection that is regarded as a region. In order to have such a characteristic, the shape of the characteristic (FIG. 3) of the second non-linear circuit 16 in the latter stage of the LPF becomes the same as that of the characteristic of the first non-linear circuit 13 (FIG. 2). Only the output level setting is different.

K obtained by the above processing (depending on the motion information
An interpolation line is formed according to (1).

(Effects of the Invention) As described above, according to the progressive scanning interpolating apparatus of the present invention, as compared with the conventional motion adaptive scanning line system using the inter-frame difference, an image or a linear image having a high spatial frequency component can be obtained. Interpolation errors that occur when the image moves can be greatly reduced, which improves the image quality for the moving image and makes the movement of the adaptive processing more nearly perfect, especially when the image is insensitive before and after the low-pass filter (LPF). Since a non-linear circuit having a saturation region and a saturation region is provided, it is possible to prevent erroneous detection due to noise and minute image vibration, and when a recursive filter is used, the circuit configuration is only slightly increased. It has the feature of high feasibility.

[Brief description of drawings]

FIG. 1 is a diagram showing a configuration of an embodiment of a progressive scan interpolation device according to the present invention, FIGS. 2 and 3 are characteristic diagrams of a non-linear circuit constituting an embodiment of the device of the present invention, and FIGS. Figure 6 shows LP
FIG. 7 is a diagram showing the structure and response of F, FIG. 7 is a diagram for explaining a conventional interpolation method, FIG. 8 is a diagram showing a conventional error interpolation example, and FIG. 9 is a diagram showing a state of motion detection. . 1,4 ... 262H delay circuit 2 ... 1H delay circuit, 3,5 ... adder, 6,10 ... subtractor, 7,8 ... delay circuit, 9 ... absolute value (ABS) circuit, 11 ... mixed circuit, 12 ... multiplier, 13 ... first nonlinear circuit, 14 ... vertical low-pass filter (LPF), 15 ... horizontal low-pass filter (LPF), 16 ... second nonlinear circuit.

Claims (2)

[Claims] 1. A progressive scanning interpolating device for forming a line to be interpolated in order to convert a television signal from interlaced scanning to progressive scanning, by taking a difference in pixel value between fields or frames around the line to be interpolated. Absolute value circuit to make absolute value and the output of this absolute value circuit are input, and the output is "0" from the input level "0" to the value corresponding to the noise level.
The dead zone which becomes, the gain increases from “0” as the input noise increases from the value corresponding to the noise level, and the gain decreases when it further increases, and the input level outputs at a certain level or higher. A first non-linear circuit having a characteristic consisting of a saturation region in which is a constant value, a low-pass filter for the vertical component and a low-pass filter for the horizontal component of the output through this first non-linear circuit, and these low-pass filters A second non-linear circuit, to which the output that has passed through is input, and which has the same characteristics as the first non-linear circuit, including the dead zone, the intermediate zone, and the saturation zone, and the signal that has passed through the second non-linear circuit, The interpolated signal formed by the upper and lower lines in the same field as the line to be processed and the interpolated signal formed by the preceding and following fields are mixed at different ratios. Mixing circuit and becomes more to sequentially scan interpolation apparatus and obtaining an interpolation signal from the output of the mixing circuit. 2. A low-pass filter for a vertical component,
The progressive scan interpolation apparatus according to claim 1, wherein the low-pass filters for the horizontal direction component are each constituted by a recursive digital filter.