JPH0447785A - Television receiver - Google Patents

Abstract

PURPOSE:To prevent quality of picture of a telop from being deteriorated even when correction by a moving vector is applied to the entire pattern by replacing a video signal of a contour of a still picture with a video signal of a relevant part of a preceding frame written in a picture memory. CONSTITUTION:The television receiver is provided with a still picture detection circuit 5 detecting a contour of a still picture and with a picture memory 27 in which a video signal is sequentially written when the correction by a moving vector is not implemented. When the correction by a moving vector is implemented, the video signal of the contour of the still picture is replaced with a video signal of the relevant part of a preceding frame written in the picture memory 27. Thus, the quality of picture is improved more than the case with the superimposition of the video signal whose phase is shifted by the correction by a moving vector. Thus, the deterioration in a reproduced picture of a still picture on an original pattern such as a telop at the correction by a moving vector at panning is not caused.

Description

DETAILED DESCRIPTION OF THE INVENTION [Field of Industrial Application] The present invention relates to a television receiver having a correction mode using a motion vector, such as a MUSE type decoder.

[Summary of the Invention] The present invention synthesizes video signals of multiple fields while shifting the phase according to the motion vector transmitted from the encoder side, thereby producing an image with higher resolution than the image reproduced from one field of video signal. In a television receiver having a correction mode using a motion vector to be reproduced, the frame difference of the video signal is below a first level and the high frequency component is above a second level without performing motion vector correction. It has a still image detection circuit that detects the outline of a still image, and an image memory into which the video signal is sequentially written when correction using the motion vector is not performed. By replacing the video signal of the outline with the video signal of the corresponding part of the previous frame written in the image memory, the entire screen can be corrected using motion vectors when panning an image that includes a subtitle. This is to prevent the image quality of the telop portion from deteriorating even when the telop is displayed.

[Prior Art] In order to increase the image resolution and improve the image quality in television reception @ (decoder), it is necessary to widen the transmission band from the encoder to the decoder in the spatial frequency domain of the video signal. There is a limit to the expansion of the transmission band in transmission channels such as satellite broadcasting.

For this reason, for example, the MUSE system and VHF are used to transmit high-definition signals over one channel of satellite broadcasting.
EDTV to improve image quality using transmission channels such as
In ACTV, etc., each pixel is divided into a still pixel whose corresponding video signal has a low temporal frequency (temporal frequency) and a video pixel whose temporal frequency is high. ) by transmitting a still image signal with a wide spatial frequency band, and by transmitting a video signal with a relatively narrow band in one field for video pixels, the transmission band of the transmission line is substantially expanded. There is.

Correspondingly, the decoder side uses motion area detection to determine whether each pixel is a still pixel or a video pixel, and for still pixels, reproduces a high-resolution image by superimposing still image signals from multiple fields, and For pixels, the original image is reproduced by interpolating the video signal of one field within the field, and for pixels in the intermediate region between still pixels and video pixels, the two reproduced signals are weighted and mixed. .

Also, if the camera is panning, the screen will be processed as a video and the resolution will be poor, but since the human eye moves to follow the image of a specific object on the screen, it will be processed as a still image. The same level of resolution is required. Therefore, for example, in the MUSE method, the encoder side detects a motion vector indicating the amount of movement of the entire screen and transmits it to the decoder side, and the decoder side uses the motion vector to obtain a resolution comparable to that of a still image.

That is, the decoder side generally calculates the difference between, for example, two frames of a video signal, and determines that a pixel for which the amount of difference is smaller than a predetermined level is a still pixel, and a pixel for which the amount of difference is larger than a predetermined level is a moving pixel. For still pixels, four fields of signals are superimposed. When performing correction using motion vectors on the decoder side, after correcting the phases of four fields of video signals whose phases are shifted due to panning etc. using motion vectors, the difference between two frames is calculated. becomes smaller. Therefore, the panned image can be reproduced as a high-resolution original image as a collection of still pixels.

[Problems to be Solved by the Invention] However, when telops such as weather forecasts and breaking news are displayed on the screen during such panning, MU
When motion area detection is performed after gradually shifting the phases of four fields of video signals based on motion vector correction on the SE decoder side, the pixels of the telop will be determined to be moving pixels. Therefore, such a telop requires relatively low resolution to be processed as a video pixel.

Therefore, fix the camera on the encoder side → panning →
If you change it from fixed to..., the resolution of the caption on the screen will change from high to bottom to high, etc., so the caption will become blurred only when panning, and the image quality will deteriorate. There is a problem with deterioration. In particular, since telops are generally images with very clear outlines, deterioration in image quality is easily noticed by viewers.

In view of the above, the present invention provides a television receiver capable of performing correction using motion vectors, in which images such as subtitles that remain stationary on the original screen while performing correction using motion vectors during panning, etc. The purpose is to prevent image quality deterioration from occurring in reproduced images.

[Means for Solving the Problems] A television receiver according to the present invention synthesizes video signals of multiple fields while shifting the phase according to a motion vector transmitted from an encoder side, as shown in FIG. 1, for example. Therefore, in a television receiver having a correction mode using a motion vector that reproduces an image with a higher resolution than the image reproduced from one field of video signal, the frame difference of the video signal is the first without performing motion vector correction. A still image detection circuit (5) detects the contour of a still image that is below level REI and whose high frequency component is above a second level RE2, and when the video signal is not corrected by the motion vector. It has an image memo U (27) that is written sequentially, and when correction is performed using the motion vector, the video signal of the outline of the still image is converted to the corresponding previous frame written in the image memo IJ (27). This is done by replacing the video signal with a partial video signal.

[Operation] According to the present invention, when the motion vector is not corrected, the image memo! A high-resolution video signal of a still image such as the telop is written to J(27>).Then, when performing correction using the motion vector during panning, etc., the still image with the image being panned etc. as the background is written. Since the high-resolution video signal of the corresponding part before the previous frame written in the image memo jJ (27) is used as the video signal of the outline of the still image, the video signal whose phase is shifted due to motion vector correction is used. The image quality can be improved compared to the case where the images are superimposed.

In this case, the inside of the still image is determined to be a still pixel even after correction using a motion vector, so a high-resolution video signal is reproduced like the outline.

[Embodiment] Hereinafter, an embodiment of the present invention will be described with reference to the drawings. In this example, the present invention is applied to a MUSE type decoder for high-definition signals.

FIG. 1 shows the decoder of this example, and in this FIG.
(1) is an input terminal, and a baseband signal of the MUSE system is supplied to this input terminal (1) from a satellite broadcasting tuner (not shown). This baseband signal is supplied to a reverse transmission P correction circuit (3) and a control signal separation circuit (4) via an A/D converter (2) with a sampling frequency of 16.2 MHz. This control signal separation circuit (4) indicates the motion vector MV supplied from the encoder side and whether the current field is a completely still image (completely parallel moving plane), a quasi-still image (semi-parallel moving plane), or a normal image. A motion control signal M that separates motion information and becomes a high level ``1'' when the image is a completely still image or a semi-still image, and a low level ``O'' during other periods.
Generate C. Also, the motion control signal is at low level “0”.
When , the motion vector is set to 0.

(5) as a whole represents a still image detection unit for detecting a still image area such as a telop, which will be described later; (6) and (7) each represent a frame memory for one frame delay; and its reverse transmission F correction circuit ( 3) is supplied to the still image detection unit (5) and frame memory (6), and the output of this frame memory (6) is sent to the frame memory (
7). The motion vector MV is supplied from the control signal separation circuit (4) to the previous stage frame memory (6), and during panning etc., the time difference between writing and reading of this frame memory (6) is used as the motion vector MV.
By changing the period from one frame according to the motion vector, correction is performed using the motion vector. Incidentally, as for the color difference signal, it is only necessary to perform interpolation as in the conventional example, so the processing circuit thereof will be omitted.

Furthermore, the input and output of the previous frame memory (6) and the output of the subsequent frame memory (7) are supplied to the subtracter (8), and the input and output of the subsequent frame memory (7) are supplied to the adder (9).
), the sum signal output from this adder (9) is supplied to a well-known interpolation circuit (10), and the difference signal between two frames output from the subtracter (8) is supplied to a well-known interpolator (10). The motion area detection circuit (11) is supplied to the motion area detection circuit (11).
) is supplied to the interpolation circuit (10).

From this interpolation circuit (10), a reproduced luminance signal D is produced by weighted mixing of the still image signal and the moving image signal reproduced by interpolation.
Y is output, and this reproduced luminance signal DY is supplied to a signal replacement section (12) to be described later.

In the still image detection section (5), the luminance signal Y is supplied to a frame delay circuit (13) consisting of a frame memory, and the human power and output of this frame delay circuit (13) are commonly used by an adder (14) and a subtractor ( The frame difference signal ΔY1 output from the subtracter (15) is supplied to the non-inverting input section of the comparison circuit (17) via the absolute value circuit (16). ) is supplied with a first constant level reference signal REI. In addition, the frame sum signal ΣY1 output from the adder 14) is filtered through a two-dimensional bypass filter in the spatial frequency domain (
HPF) (20) and the absolute value circuit (21) to the non-inverting input of the comparator circuit (22).
22), a second constant level reference signal RE is applied to the inverting input of
Supply 2.

The binary frame difference signal ΔY2, which is the inverted output of the comparator circuit (17), is at a high level of 1 in a still image area where the absolute value of the frame difference signal ΔY1 is less than RE1 without motion vector correction. In other moving image areas, the low level is ``0''.

On the other hand, the binary frame sum signal ΣY2, which is the output of the comparison circuit (22), has the absolute value of the high frequency component of the frame sum signal ΣY1 without being corrected by the motion vector.
The high level is "1" in the outline of the still image area where the value is 2 or more, and the low level is "0" in other areas. These signals ΔY2 and ΣY2 are respectively connected to a two-dimensional extension circuit (18
) and (23) to the AND circuit (19),
The signal TESI output from this AND circuit (19) is supplied to the non-inverting input section of a comparator circuit (25) via a temporal filter (24) which is a low-pass filter on the time axis. A third constant level reference signal RE3 is supplied to the inverting input of the circuit.

The signal TBS2 output from this comparison circuit (25) is approximately the binary frame difference signal ΔY2 and the binary frame sum signal ΣY.
This signal is an AND output with ``2'' and has a high level of ``1'' only in the outline of a still image area such as a telop, and a low level of ``0'' in other areas.

Therefore, the signal TES2 is called a "static telop edge signal". In this case, two-dimensional stretching circuits (18) and (23) are used to widen the contour,
The temporal filter (24) prevents signals with a very fast rise or large amplitude from being mistakenly detected as contours when they move, and allows only the contours of signals that are completely stationary to be detected. used for

In the signal replacement section (12) in FIG.
Y is supplied to the input part of the frame memo IJ (27) and one input part of the data selector (28), and the replacement signal RY read from this frame memo 'J (27) is supplied to the other input part of the data selector (28). The output of the data selector (28) is supplied to a processing circuit including a D/A converter (not shown) via a connection terminal (62).

Also, (26) is an AND circuit, and this AND circuit (
A control signal separation circuit (4) is connected to the two input sections of (26).
) is supplied with the motion control signal MC from the still image detection section (5) and the still telop edge signal TES2 from the still image detection section (5), and the output of the AND circuit (26) is connected to the negative logic write enable terminal RB of the frame memory (27) and the data. It is supplied to the switching control terminal of the selector (28).

The output of this AND circuit (26) becomes high level ``1'' only at the outline of the still image area where the still telop edge signal TBS2 is high level ``1'' in the field or frame in which motion vector correction is performed. When the signal TES2 is at a high level P1'', writing to the frame memo IJ (27) is prohibited during the Al period, and the data selector (28) writes data to the frame memo IJ (27).
) is selected. Therefore, the AND circuit (2
The output of 6> is called 1 write inhibit signal WIHJ.

That is, the frame memory (27) is mainly a memory for storing the image of the telop, but the telop does not generally appear on the entire screen, and on the contrary, when the telop appears on the entire screen. There is no need to perform correction using motion vectors. Therefore, the capacity of the frame memo IJ (27) is not necessary for the planning surface, and it is also possible to use one with a capacity that can divide the entire screen into 8 parts and store images from about 2 of them. Therefore, the frame memory (27) can also serve as a part of the image memory used in the interpolation circuit (10), so the scale of the additional circuit can be made extremely small. .

Next, the two-dimensional expansion circuit (18) of the still image detection section (5)
), specific configuration examples of the two-dimensional/S pass filter (20) and the temporal filter (24) will be explained.

FIG. 2 shows a configuration example of a two-dimensional bypass filter (HPF) (20), and in this FIG. 2, (30).

(31) are delay circuits connected in cascade with each delay time being one horizontal period (IH), (34), (37), (38)
each has a delay time of one sample period (L D, 16.2
MHz), it supplies power to this filter to the previous stage delay circuit (30), and connects this input and the delay circuit (30) with the input power to this filter.
31) in an adder (29), and the addition result and the output of the delay circuit (30) are added to a multiplier (32) with a coefficient of 1/4 and a multiplier (33) with a coefficient of 1/2, respectively. ), and the output of the delay circuit (30) is supplied to the subtracter (42) via a delay element (34).

Further, the output of the adder (35) is supplied to the adder (36) and the delay element (37), and the output of the delay element (37) is supplied to the delay element (36) and the delay element (37).
38) to the adder (36), and this adder (
36) and the output of the delay element (37), each with a coefficient of 1.
/4 multiplier (39) and multiplier with coefficient 1/2 (40
) to the adder (41), and this adder (41
) is supplied to the input section on the subtraction side of the subtractor (42). The difference output of this subtracter (42) becomes the output of this filter.

The left half of the circuit in Figure 2 is a vertical low-pass filter consisting of a 3-tap transversal filter, and the right half is a 3-tap transversal filter.
The function as a noise pass filter is achieved by subtracting the signal that has passed through a two-dimensional low pass filter consisting of a tap transversal filter and the like from the original signal in a subtracter (42).

FIG. 3 shows an example of the configuration of a two-dimensional extension circuit (18),
In FIG. 3, the delay time is IH delay circuit (43
) to (46) are connected in cascade, and the delay time is the delay element of ID (
48) to (51) are connected in cascade, and the first delay circuit (43) is connected in cascade.
) is supplied with human power to this extension circuit, and this human power and the outputs of the delay circuits (43) to <46) are connected to a 5-man powered OR circuit (
47), and the output of this OR circuit (47) is supplied to the first delay element (48), and this output and the delay element (48)
The outputs of ~(51) are supplied to a five-person OR circuit (52). The output of this OR circuit (52) becomes the final output of this extension circuit.

The stretching circuit shown in FIG. 3 stretches one pixel's worth of dots manually input by 5 lines in the vertical direction and 5 dots in the horizontal direction. In this case, in order to extend the target pixel so as to surround it, for example, an interpolation circuit (10
), the luminance signal Y may be delayed by two lines in the vertical direction and two dots in the horizontal direction.

FIG. 4 shows an example of the configuration of the temporal filter (24),
In FIG. 4, a 1-bit input to this filter is supplied to one of the human parts of an adder (53) with a 5-bit output, and the output of this adder (53) is fed to a frame delay circuit <
54) and a coefficient of 15/160 through a multiplier (55) to the other part of the adder (53), and the output of this adder (53) is fed to a multiplier (56) with a coefficient of 1716. supply This multiplier (56) is a circuit for extracting the maximum bit of the addition result of the adder (53) (5 vias), and one bit of this maximum digit is the final output of this filter. become.

In this case, the multiplier (55) is connected to the frame delay circuit (54).
The output is multiplied by 15/16 and the result is rounded off and supplied to the adder (53), and operates as a kind of limiter. Therefore, in the example in Figure 3, the output becomes "1" only when the human power reaches a high level of "1" 16 times in a row, and then when the human power reaches a low level of "0", the output also follows. It becomes “o”.

In other words, the example in FIG. ) by setting the coefficient to 7/8 and the coefficient of the multiplier (56) to 178, it is easy to input "
It can be changed so that 1'' is output.

Incidentally, when the output is regulated to 1 bit by itself as in the example in FIG. 4, the comparator (25) in FIG. 1 can be omitted. However, if the output stage multiplier (56) is omitted in this example in FIG. 4, a kind of noise level can be set by adjusting the reference signal RE3 of the comparator (25) in FIG. I can do it.

Referring to FIGS. 5 and 6, we will explain the overall operation when the camera on the encoder side is moved from a fixed state to a panning state and the image is played back by the decoder shown in FIG. 1. In this case, Assume that on the encoder side, a subtitle consisting of the characters "News" is displayed in a stationary state in a part of the panned image, as shown in FIG.

First, when the camera is in a fixed state, the entire image is a still image, so the interpolation circuit (10) of the decoder shown in Figure 1 synthesizes four fields of brightness signals and reproduces one frame of brightness signals. High-resolution reproduced luminance signal D
Y is output, and this high resolution signal DY is supplied to the data selector (28) and frame memory (27).

In this case, even if the motion vector correction is performed, the motion vector MV is O, so the reproduced image is the same, and the reproduced luminance signal DY and the replacement signal RT read from the frame memory (27) are the same, so Data selector (
No matter which manual power is selected by 28), the luminance signal output from the connection terminal (62) is the same. Furthermore, the static telop edge signal TBS2 is '1' at the contour of the still image, but the signal TES2 is generally '1' inside the still image.
In the area where the signal TBS2 becomes 1'', image data such as the previous frame has already been written to the corresponding portion of the frame memory (27), so writing of the signal replacement unit (12) is prohibited. The output of the connection terminal (62) is the same regardless of the level of the signal WIH.

Next, assuming that the luminance signal of the n-th frame is Y (n) and the camera is panned from this n-th frame, the luminance signal Y (n-1) of the (n-1)th frame and the luminance signal Y ( For example, as shown in FIG. 6 and B, the telop area (59) is the same and the background is shifted in parallel. Since the telop area is an area that can be ignored compared to the entire image, the motion vector MV corresponding to the amount of translation of the background and motion information indicating the degree of translation are transmitted from the encoder to the decoder. The control signal separation circuit (4) in FIG. supply to.

In this case, the frame difference signal ΔY1 becomes a signal close to O in at least the telop area (59) as shown in FIG. 6C, and the binary frame difference signal ΔY2 becomes a signal close to O in at least the telop area (59) as shown in FIG. It becomes "1".

Furthermore, the level of the telop area (59) is different from the level of the background image, and in the frame sum signal ΣY1, as shown in FIG. has been amplified. Therefore,
The signal obtained by extracting the high-frequency component from the signal ΣY1 using the two-dimensional bypass filter (20) exceeds the level of the reference signal RE2 at least at the edge portion of the telop area (59), as shown in FIG. 6F. As a result, the binary frame sum signal DY2 becomes '1' at least at its edge portion, as shown in FIG. 6G.

The signal TESI obtained by ANDing the signal obtained by two-dimensionally extending the signal ΔY2 and the signal obtained by two-dimensionally extending the signal ZY2 is, as shown in FIG. 6H, the signal TESI of the telop area (59). The contour vicinity part (60) near the rising part and the contour vicinity part (60) near the falling part (
61) is 1'', and in other areas it is 0''.

In this example, the signal TBSI is at least in both neighborhoods (60
), (61) continuously 1' over a series of frames.
”, the static telop edge signal TES2 supplied to the signal replacement unit (12) is the same as the signal TESI.

To represent the signal flow in FIG. 6 two-dimensionally with reference to FIG. 5, the contour of the telop on the encoder side in FIG. 5A is shown by the broken line in FIG. 5B-D. At this time, the signal obtained by expanding the binary frame difference signal ΔY2 in Figure 6 using the two-dimensional expansion circuit (18) becomes "1" in the shaded area in Figure 5B.
The signal obtained by expanding the binary frame sum signal ΣY2 in FIG. 5G by the two-dimensional expansion circuit (23) becomes '1'' in the shaded area in FIG. 5C. Then, the static telop edge signal TE
S2 is 1 near the outline of the telop indicated by diagonal lines in the fifth figure.
”, which corresponds to the shaded area in Figure 5B and the 5th signal.
This corresponds to the common area with the shaded area in Figure C. In this example, the outputs of the two-dimensional expansion circuits (18) and (23) are ANDed, so for example, in FIG. Also, the area (58) in which the noise level is 1"' in FIG. 5C is advantageously excluded from the vicinity of the contour in FIG. 5.

In the portions (60) and (61) near the contour of the telop area (59) during which the signal TBS2 is 1'', the first
In the signal replacement unit (12) in the figure, the write inhibit signal WI
H becomes ``bi'' and frame memo! Writing to J (27) is prohibited and the data selector 28) is switched to the frame memo IJ (27> side. Therefore, in the vicinity of the outline of the telop area (59), the frame memo IJ (27) has already been written. The high-resolution luminance signal written before the previous frame is read out, and this read signal is supplied to the connection terminal (62) side as the replacement signal RY.However, the area near the outline of the subtitle is the edge of the telop. The background image is almost not included.

In this case, if the same image data of consecutive fields corresponding to the vicinity of the contour are translated in parallel according to the motion vector in the frame memory (6), the image data of those consecutive fields will become different from each other. Therefore, the interpolation circuit (10) performs interpolation only within one field as a moving pixel, and the resolution of the reproduced luminance signal DY deteriorates. However, this reproduced luminance signal DY is frame memo 'I (27) 1. :: It is neither written nor supplied to the connection terminal (62). In general, during panning, the frame difference signal △Y1 is close to 0 in areas other than telops, such as all-black, all-white margins, blue sky areas, etc., and the frame difference signal △
A signal replacement unit (12) converts the luminance signal in the area where Y1 is near O.
If the signal is replaced with a signal read from the frame memory (27) of the frame memory (27), the load on the signal replacement section (12) becomes heavy. However, in this example, since the signal is replaced only in the vicinity of the outline of the still image, the signal is not replaced in areas such as blue skies, and the signal replacement unit (12)
There is an advantage that the burden on the user is light and the capacity of the frame memo 'J (27) is small.

Further, inside the area near the outline (for example, inside the area surrounded by diagonal lines in the fifth diagram), the static telop efficiency signal TES2 is corrected by a low level ``0'''' motion vector. However, even if this internal image is shifted, the signal level remains the same, and the interpolation circuit (10) performs reproduction as a still pixel, so a high-resolution luminance signal is reproduced.

As described above, according to the decoder of this example, in the vicinity of the outline of a telop that is stationary on the screen during panning, the motion vector is corrected without outputting the low-resolution luminance signal reproduced as a video pixel. Since the high-resolution luminance signal written in the frame memory (27) is read and output before correction, the high-resolution luminance signal is always the same as the background image before and during panning. This has the advantage that the image quality of the reproduced image can be improved.

It should be noted that the present invention is not limited to the above-mentioned embodiments, but can be applied to, for example, an EDTV system decoder, etc., and can be applied to zooming, tilting, etc., and various configurations can be adopted without departing from the gist of the present invention. It is.

G [Effects of the Invention] According to the present invention, when performing correction using a motion vector, the video signal of the contour part of a still image such as a telop is used as a high-resolution video signal of the corresponding part of the previous frame written in the image memory. Since it is replaced with a video signal, there is an advantage that the reproduced image of a still image on the original screen, such as a subtitle, does not suffer from deterioration in image quality when correction is performed using a motion vector during panning or the like.

[Brief explanation of the drawing]

Fig. 1 is a block diagram showing a MUSE type decoder according to an embodiment of the present invention, Fig. 2 is a block diagram showing an example of a two-dimensional bypass filter, and Fig. 3 is a block diagram showing an example of a two-dimensional bow extension circuit. 4 is a configuration diagram showing an example of a temporal filter, FIG. 5 is a line diagram for explaining the detection operation near the outline of the telop in the embodiment, and FIG. 6 is a diagram showing the area near the outline of the telop in the embodiment. FIG. 3 is a diagram showing signal waveforms of various parts when detecting. (4) is a control signal separation circuit, (5) is a still image detection unit, (12) is a signal replacement unit, (15) is a subtracter, (
20) is a two-dimensional bypass filter, (24) is a temporal filter, (27) is a frame memory, and (28) is a data selector. Agent Hidemori Matsukuma. Figure 6 Figure 6 October 26, 1990

Claims (1)

[Claims] A motion vector that reproduces an image with a higher resolution than an image reproduced from a single field video signal by synthesizing video signals of multiple fields while shifting the phase according to a motion vector generated on the encoder side. In a television receiver having a correction mode using
a still image detection circuit that detects an outline of a still image whose high frequency component is below a second level and above a second level; and an image memory into which the video signal is sequentially written when the motion vector is not corrected. When performing the correction using the motion vector, the video signal of the outline of the still image is replaced with the video signal of the corresponding part of the previous frame written in the image memory. television receiver.