JPS5937631B2 - Composite predictive coding/decoding device for color television signals - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION The present invention relates to a composite predictive encoding/decoding device for color television signals in which subcarriers are modulated with color signals and frequency multiplexed.

As a method for encoding television signals, there is a frame predictive encoding method that utilizes frame correlation of television signals.

The frame predictive coding method uses the sample value of the previous frame as a predicted value, and quantizes and codes the prediction error, and is an effective coding method for television signals in which there is little movement of the subject. Encoding methods that further improve coding efficiency include pre-prediction, which uses the sample value one sampling time before as the predicted value, and composite prediction, which combines line prediction and frame prediction, which uses the sample value one scanning line before as the predicted value. There are encoding methods. However, although conventional composite predictive coding methods are effective for black and white television signals, prediction errors increase for composite color television signals that include carrier color signals whose phases are inverted for each line and frame. The drawback was that it could not be applied. For this reason, the conventional method has been to separate and demodulate a frequency multiplexed color television signal into baseband signals (for example, luminance signal Y and chrominance signals I and Q in the NTSC system), and then individually perform composite predictive coding. . In this case, it is necessary to separate and demodulate and re-modulate and multiplex the color signals, which not only complicates the apparatus, but also causes amplitude and phase errors that occur in the demodulation process to deteriorate image quality. An object of the present invention is to combine a prediction filter adapted to a frequency multiplexed color television signal including a carrier color signal, and to prevent an increase in frame prediction error due to the carrier color signal component described above, thereby decomposing and resynthesizing the color television signal. An object of the present invention is to provide an encoding/decoding device that can perform direct predictive encoding without performing any additional steps.

According to the present invention, direct predictive coding can be performed without disassembling and resynthesizing a composite color television signal, so that coding and transmission of color television signals can be realized with a simple device and with little deterioration in image quality. Compatibility with black and white television signals can also be achieved. Hereinafter, an explanation will be given using an NTSC color television signal as an example. The subcarrier frequency fsc is selected when fH is the line frequency.

Further, the following relationship exists between frame frequencies FF and FH. Therefore, the phase of the subcarriers will be reversed from line to line and from frame to frame.

If the time function of the luminance signal Y is y(t), and those of the color signals I and Q are i(t) and q(t), then the NTSC color television signal u(t) is expressed as follows.

If one frame period is TF, then TF=1/FF, so the {NTSC color television signal U(t
+TF) is obtained from equations (1), (odor, and (3)).

If the subject is stationary, then the frame prediction error e(t) at this time will be.

As is clear from equation (6), even when the subject is stationary, the frame prediction error does not become zero, and the carrier color signal component appears doubled, so it is normal for NTSC signals. Frame prediction alone has little effect.
However, as is well known, the energy of the carrier color signal component of the NTSC signal is concentrated near frequencies that are odd multiples of half the line frequency JH.

Therefore, if an efficient prediction filter for this frequency component can be realized, the above-mentioned prediction error can be made sufficiently small, and composite predictive coding becomes possible as in the case of black-and-white television signals. FIG. 1 shows a principle diagram of a composite predictive encoding device of the present invention.

For convenience of explanation, all the encoding and decoding device systems are sample value systems, and the transfer function of the first prediction filter 26 is referred to as P1(z), and the transfer function of the second prediction filter 27' is referred to as P2(z). , the transfer function of the third prediction filter 28 is P3
(z), the input television signal is U(z), and the quantization error generated in the quantization circuit 25 is N(z), then the subtracter 21
, 22 and 23, the output signals E1(z), E2(z) and E3(z) are as follows. As is clear from these equations, if the quantization error N(z) is sufficiently small, E1
(z) is a prediction error component when U(z) is predicted by the prediction filter 26, and E2(z) contradicts E1(t).
This is a prediction error component when predicted by the filter 28. Therefore, if P1(z) is a frame selection element, E,
When the subject is stationary, (z) becomes a transport color signal component approximately expressed by equation (6). A feature of the present invention is that the carrier color signal component is suppressed by the second prediction means by utilizing the fact that the spectrum of the carrier color signal component is concentrated near frequencies that are odd multiples of FH/2. be.

In other words, if the function {1-P2(z)} is a function that is sufficiently smaller than 1 at frequencies that are odd multiples of FH/2, E2(z) is determined by suppressing the carrier color signal component from equation (8). It becomes a signal. If the subject is moving, E1
In (z), a frame prediction error of the luminance signal component appears in addition to the carrier color signal component.

Also in this case, the carrier color signal component is passed through the second predictive filter 2.
7, only the luminance signal component appears in -E2(z). Therefore, predictions similar to those for black-and-white television signals can be made for E2(z). For example, as the third prediction means, the previous value prediction is used, and the function {1
-P3(z)} is set to be sufficiently smaller than 1 near the zero frequency, as in the case of black-and-white television signals,
The prediction error E3(z) becomes a signal concentrated at zero, allowing efficient predictive coding.

Also, the function {1-P. (z)} is set so that not only the zero frequency but also the number of subcarriers Fsc is sufficiently smaller than 1, E. Since the residual carrier color signal component included in (z) is also suppressed, more efficient predictive coding can be performed. In addition,(
As is clear from equations 7), (8), and (9), there is no change in E3(z) even if the transfer characteristics of the prediction filters 26, 27, and 28 are interchanged. Therefore, any of the prediction filters 26, 2T, and ε28 may have the transfer characteristics P1(z), P2(z), and P3(z). The output V(z) of the quantization circuit 25 is E3
(z) plus the quantization error N(z), so
It is immediately derived from the Afj formula.

Therefore, the transfer characteristic of the encoding device for the signal U(z) is { 1-P1(z)}{ 1-P2(z)}{ 1-
P3(z)}, so the transfer characteristic W(z) of the decoding device
, the transfer characteristic from the input of the encoding device to the output of the decoding device becomes 1, and a decoded signal U(z) is obtained.

In other words, the transfer characteristic of the decoding device may be set to be the opposite of the transfer characteristic of the encoding device. Examples will be described below.

FIG. 2 is a block diagram showing the configuration of an embodiment of the present invention. An input color television signal u(t) to the encoding device is converted by an analog/digital converter 31 into a PCM (pulse code modulation) signal U(z) with a sampling frequency Fs.

Fs is selected to be 2 to 3 times the signal bandwidth and an integral multiple of FH. P
The CM signal U(z) is supplied to a subtracter 21, and the difference between it and the output of the first prediction filter 26 is calculated. The output of the subtracter 21 is sent to the subtracter 22, and the difference between it and the output of the second prediction filter 2T is calculated. The output of the subtracter 22 is further sent to a subtracter 23, and the difference between it and the output of the third prediction filter 28 is calculated. The output of the subtracter 23 is quantized by a quantization circuit 25, and the output of the quantization circuit is sent to a code converter 39, where the code is converted, smoothed through a buffer memory 38, and sent out to a transmission line. There is an overwhelming probability that the quantization circuit output will be at zero level, and the probability of occurrence generally decreases as the output level moves away from zero, so the code converter 39
By performing code conversion that matches this probability distribution,
The number of transmission bits can be greatly reduced. The output of the quantization circuit 25 is also supplied to the adder 13 and added to the output of the third prediction filter 28, and the result is supplied to the prediction filter 28 and the adder 12 to form a feedback loop. Similarly, adder 12 and adder 11 also constitute a feedback loop. The decoding device on the receiving side includes a buffer memory 4.
The received code stored in 2 is converted by the code decoder 40 into a code corresponding to the quantized output on the transmitting side, and then sent to the adder 13'.
supplied to The adder 13' adds the output of the eighth prediction filter 28', which has the same function as the transmission side prediction filter, and the output of the code decoder 40, and sends the result to the adder 12'.
and fed back to the prediction filter 28'. Similar loops are constructed for the second and first prediction filters 27' and 26'. If the decoding device is configured like this, the transfer characteristic W(z) from the input of adder 13' to the output of adder 1V will be the inverse characteristic of that of the encoding device as shown in equation (self). , a decoded signal n(Z) is obtained at the output of the adder 1V. n(Z) is converted by the digital/analog converter 41 into a color television signal called ti(t). n(t) is the original color television signal U(t
) plus quantization error. Since the dynamic range of the input to the quantization circuit is sufficiently suppressed by the prediction filter, the quantization error can be sufficiently reduced. Next, a specific method of configuring a prediction filter having transfer characteristics P1(z), P2(z), and P3(z) will be described.

First, P1(z) may be delayed by one frame using a normal frame delay element, that is, a shift register or a memory element. That is, the frame frequency Fl
l, for the sampling frequency Fs, realize the delay of the sample value F (= Fs / FF). In this case, P1 (z) = Z - F, so the function {1 - P1 (z)} is , it can be seen that it becomes zero at frequencies that are integral multiples of the frame frequency.

Next, P1(z) can be set such that the function {1-P2(z)} is an odd multiple of 1/2 of the line frequency FH and is sufficiently smaller than 1, so it is sufficient to include the term.

For example, a predictive filter that satisfies this requirement is shown in Fig. 3.
What is shown in FIGS. 4 and 5 can be easily realized. FIG. 3 shows a case where the prediction filter 27 is composed of an H sample value delay element 3 and a polarity inverter 4.

In this case, P2(z)=-Z-H, so the function {1-
P2(z)} is also a filter that suppresses the carrier color signal component.

FIG. 5 shows a case in which two H sample value delay elements 3 are connected in series, which also serves as a filter for suppressing the carrier color signal component.

The carrier color signal component can be suppressed by any of the three prediction filters mentioned above, but since the line correlation of the signal generally decreases as the prediction distance increases, the suppression effect is almost the same as that shown in FIGS. The number decreases in the order of Fig. 5.

On the other hand, the passage characteristic for the luminance signal component increases approximately twice in the configuration shown in FIG. 3, is almost flat in the configuration shown in FIG. 4, and is suppressed in the configuration shown in FIG. Since the prediction error caused by the second prediction filter is further suppressed by the third prediction filter, the most efficient combination of the third prediction filter ξ should be selected from among the three prediction filters mentioned above. All you have to do is choose. Next, a method for realizing the transfer characteristic P3(z) of the third predictive filter 28 will be described.

Since the third prediction means is mainly for suppressing the luminance signal component, the function {1-P3(z)} is (1-Z-1)
It is sufficient to include the section. Therefore, the simplest configuration is P
This method uses a one-sample value delay element for 3(z), and this is the same as normal previous value prediction. As another configuration example of P3(z), the input of the third prediction circuit (E2(z) in FIG.
), it is effective to have a configuration in which {1-P3(z)} is sufficiently smaller than 1 even at subcarrier frequency Fsc.

FIG. 6 shows an example of this, which is composed of four one-sample value delay elements D, a multiplier 5 with a magnification of 0.5, a subtracter 6, and an adder 7. In this case the function {1-P3(z)} is
It can be seen that if the sampling frequency Fs is selected to be about three times Fsc, the formula (self) becomes sufficiently smaller than 1 at the zero frequency and the subcarrier frequency Fsc.

Therefore, the third prediction means can suppress not only the residual luminance signal component but also the residual carrier color signal component.
A predictive filter for such a purpose can be modified in various ways by changing the relationship between Fs and Fsc. As described in the embodiments above, according to the present invention, direct predictive coding can be performed without adding any operation to a frequency multiplexed color television signal, and efficient coding transmission with little deterioration in image quality can be realized. Furthermore, since the prediction error signal entering the quantization circuit contains almost no carrier color signal component, it is possible to control the encoding operation in the same way as in the case of a monochrome television signal. In other words, depending on the movement of the subject, if the movement is very large, it is possible to apply controls such as subsampling and field repetition to prevent an increase in the amount of generated information, and to greatly reduce the number of transmission codes. Can be compressed.

[Brief explanation of the drawing]

FIG. 1 is a block diagram showing the configuration of a composite predictive encoder of the present invention, and FIG. 2 is a block diagram showing the configuration of an encoding/decoding apparatus. FIGS. 3, 4, 5, and 6 are block diagrams showing specific examples of the construction of the prediction filter. 21, 22, 2
3, 6... Subtractor, 11, 12, 13, 1V,
12! , 13', 7... Adder, 26, 27,
28, 267, 27', 287...Prediction filter, 25...Quantization circuit, 31...Analog/digital converter, 41...Digital/
Analog converter, 39... Code converter, 40.
... code decoder, 38, 42 ... buffer memory, 3 ... 1 line delay element, 4, 5...
...multiplier.

Claims (1)

[Claims] 1. In a predictive coding transmission device for a composite color television signal including a carrier color signal, the prediction function (1-P_1) based on sample values in the previous frame is at least near a frequency that is an integer multiple of the frame frequency, and is based on one or several lines before. The prediction function (1-P_2) based on the sample value is at least 2 of the line frequency.
Near frequencies that are odd multiples of 1/2, the prediction function (1-P_3) based on sample values within the same scanning line is at least near zero frequency, and the first and second prediction functions are sufficiently smaller than 1, respectively.
and a third prediction means, and means for quantizing, encoding, and transmitting a composite prediction error generated by the first, second, and third prediction means according to a predetermined rule, the encoding device The overall transfer characteristic is (1-P_1)(
1-P_2) (1-P_3), and by setting the decoding device so that the transfer characteristic is the opposite of that of the encoding device, predictive coding can be performed without decomposing the decoded color television signal. A composite predictive encoding/decoding device.