JPH0251302B2 - - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION The present invention relates to a predictive encoding/decoding device that predictively encodes an image signal and predictively decodes the predictively encoded signal.

Predictive coding is one method for efficiently coding image signals such as color television signals. When transmitting a signal that has been predictively encoded,
If an error occurs on the transmission path, the received and decoded signal will not be reproduced correctly, and the error will propagate.
In the case of a one-dimensional prediction function, in other words, a prediction function using only sampled pixels on the same line, error propagation occurs one-dimensionally, or in other words, only in the horizontal direction due to predictive decoding. On the other hand, in the case of a two-dimensional prediction function that has better prediction efficiency than one-dimensional prediction function, or in other words, a prediction function that uses pixels from the same line and previous lines, error propagation is reduced to two-dimensional prediction by predictive decoding. In other words, it occurs both horizontally and vertically.

For one-dimensional prediction, if the first few pixels of each line are reset by transmitting a PCM (Pulse Code Modulation) signal, error propagation can be suppressed within one line. By interpolating the line in which the error has occurred using the decoded signal of the previous line, it is possible to prevent deterioration in the quality of the reproduced image. However, for two-dimensional prediction, it is generally not possible to interpolate an error line with the previous line because errors occur in all subsequent lines from the line where the error occurred. For this reason, in conventional predictive coding and decoding of two-dimensional prediction, even if interpolation is performed on an error line, the signal used for interpolation is a decoded signal of a distant line, and the number of error lines to be interpolated is It was not possible to obtain a satisfactory decoded signal by interpolation because the signals were continuous.

The purpose of the present invention is to prevent deterioration of coding efficiency for decoded signals of error lines that are erroneously reproduced due to errors occurring in the transmission path in a predictive coding/decoding device using a two-dimensional prediction function. An object of the present invention is to provide a predictive encoding/decoding device that can obtain a decoded signal without image quality deterioration by performing interpolation.

The predictive coding/decoding device for image signals of the present invention includes:
A two-dimensional predictor configured using only sampling pixels on the current line and the even-numbered previous line and a one-dimensional predictor configured using sampling pixels only on the current line are calculated for each predetermined line. means to perform predictive encoding to obtain a prediction error signal from the image signal, means to convert the prediction error signal into a signal to be transmitted, and a signal for error detection so that transmission errors can be detected on a line-by-line basis. a predictive coding device comprising: means for adding a signal to a transmission signal;
and means for receiving the signal sent from the predictive encoding device and detecting the presence or absence of a transmission error on a line-by-line basis, and performing predictive decoding using a predictor having the same configuration as the transmitting side and decoding an image from the received signal. a means for obtaining a signal; a means for determining which line of the decoded signal of an image contains an error from the signal indicating the presence or absence of a detected transmission error; The present invention is characterized in that it has a predictive decoding device equipped with means for interpolating using a decoded signal of an error-free line.

According to the present invention, since the prediction function of the predictor is composed only of sampled pixels on the current line and the even-numbered previous line, error propagation by predictive decoding is performed on subsequent lines where a transmission path error has occurred. It now propagates only to even-numbered lines,
By interpolating the decoded signals of the even-numbered lines that were incorrectly decoded using the decoded signals of the correct lines for each line, it is possible to reproduce the decoded signals, thereby eliminating transmission path errors. Even if there is a problem, the image signal can be reproduced without reducing the encoding efficiency or deteriorating the image quality.

The present invention will be described in detail below using examples.

FIG. 1 is a block diagram showing the configuration of a first embodiment of the present invention.

In this embodiment, a case will be described in which a picture signal of an NTSC color television signal is sampled at a frequency three times that of the subcarrier f _SC and predictively encoded, and a signal is predictively decoded. The image signal input to the input terminal 1 is sent to the A/D converter 2, where it is sampled at a sampling frequency _S three times the subcarrier f _SC , converted into a digital signal, and supplied to the subtracter 3 . The subtracter 3 subtracts the prediction signal supplied from the predictor 6 from the digitized image signal supplied from the A/D converter 2, outputs a prediction error signal, and supplies the signal to the quantizer 4. . The quantizer 4 quantizes the prediction error signal supplied from the subtracter 3 according to predetermined quantization characteristics,
A quantized prediction error signal is output from the quantizer and is supplied to the code converter 7 and the adder 5. Adder 5 adds the quantized prediction error signal supplied from quantizer 4 and the prediction error signal supplied from predictor 6, outputs a locally decoded signal to the output of adder 5, and supplies it to predictor 6. do. The predictor 6 has a predictor (i.e., a two-dimensional predictor) configured using only the sampled pixels in the even-numbered previous line including the current line, and uses horizontal error propagation to predict the next line. In order to prevent error propagation, at the beginning of each line, the predicted signal for the next sampling time is obtained from the locally decoded signal while resetting the internal value of the predictor 6, and outputs it to the output of the predictor 6, and then performs prediction. The signal is supplied to a subtracter 3 and an adder 5. In addition, the predictor 6 performs the prediction every appropriate line (for example, every 5 lines for both odd-numbered lines and even-numbered lines) in order to stop error propagation in the vertical direction.
It has an error reset function that performs prediction using a dimensional prediction function (in other words, a one-dimensional predictor) and resets the values in the predictor 6 to stop error propagation in the vertical direction due to predictive decoding.

The code converter 7 has code conversion characteristics corresponding to the quantization characteristics of the quantizer 4, and converts the quantized prediction error signal sent from the quantizer 4 into a corresponding equal-length code according to the conversion characteristics. The equal length code is output from the code converter 7 and supplied to the parity addition circuit 8. In the parity addition circuit 8, a parity signal is added to the equal-length code for each horizontal scanning line on the receiving side so that it can be detected whether an error has occurred in the transmission path, and a one-dimensional predictor and a two-dimensional predictor are added to the parity signal. A control signal necessary for decoding, such as a predictor selection signal indicating which of the predictors is selected, is outputted to the output of the parity addition circuit 8 and output terminal 9.
is sent out to the transmission path. The above is an explanation of the operation of the predictive encoding device.

The received signal input to the terminal 11 is supplied to the parity detection circuit 12. The parity detection circuit separates the predicted selection signal, the equal-length code signal, and the parity signal, and detects whether there is a transmission error in the transmission path for the code-length code signal by one line from the equal-length code signal and the parity signal. A parity check signal indicating the presence or absence of an error is obtained by detecting each time, and is sent to the interpolation judgment circuit together with the separated prediction selection signal.
The equal length code signal is supplied to the code inverse converter 13. The code inverse converter 13 has a code inverse conversion characteristic that is opposite to the code conversion characteristic of the code converter 7, and performs corresponding quantization on the equal-length code signal supplied from the parity detection circuit 12 according to the inverse conversion characteristic. The prediction error signal is converted into a predicted error signal, outputted, and supplied to the adder 14.

The adder 14 adds the quantized prediction error signal supplied from the code inverse converter 13 and the prediction signal supplied from the predictor 15, obtains a decoded signal as the output of the adder 14, and performs interpolation with the predictor 15. Supplied to circuit 16. The predictor 15 has the same function as the predictor 6 and performs similar operations to obtain a predicted signal for the next sampling time from the decoded signal, output it to the output of the predictor 15, and supply it to the adder 14. For the line for which error reset is to be performed, the predictor 15 performs prediction using a one-dimensional prediction function (one-dimensional predictor) and resets the value of the predictor 15. The interpolation determination circuit 19 uses the parity check signal and the predictor selection signal to determine whether
The interpolation circuit 1 determines that the line until the predictor selection signal selects the one-dimensional predictor is a line that has been erroneously decoded.
6 outputs an interpolation selection signal for controlling interpolation and supplies it to the interpolation circuit 16. Note that since the parity check signal is determined at the end of each line, the interpolation selection signal is output with a delay of one line.
To compensate for this delay, interpolation circuit 16 includes a one-line delay circuit. In accordance with the selection signal from the interpolation determination circuit 19, the interpolation circuit 16 outputs the decoded signal delayed by one line without interpolating the line without an error, and outputs the decoded signal after being delayed by one line. Interpolation is performed using the decoded signal of the line that does not exist, and the output is output and supplied to the D/A converter 17. The D/A converter 17 converts the interpolated decoded signal into an analog signal and outputs it to the terminal 18 .

The above is an explanation of the operation of the predictive encoding/decoding device.

FIG. 2 shows a specific circuit example of the predictors 6 and 15 in the first embodiment of the present invention. In this circuit example, Z ^-1 indicates the delay of one sampling clock period, and two prediction functions P ^(z) ₁ and P ^(z) ₂ P ^(z) ₁ = 0.5Z ^- are shown by the following equations. ¹ +Z ^-3 -0.5Z ^-4 (1) P ^(z) ₂ =Z ^-2H (2) (However, since f _S = 3f _SC , the number of samples H in one horizontal scanning period is H = 682.5.) A case is shown in which predictive encoding and decoding are performed using a two-dimensional predictor that adaptively switches the data. The adaptive switching method is a method in which the predicted values obtained by two prediction functions are compared with the locally decoded signal, and the prediction function that is closer to the decoded signal is used as the prediction function at the next sampling time. Since this method performs switching using encoded signals, there is no need to transmit switching signals. The locally decoded signal input to the predictor 6 is supplied to a first prediction circuit 21 and a second prediction circuit 22. First prediction circuit 21
are registers 25, 27, 28, and 31 that delay the input signal by one sampling clock cycle and output it, multipliers 24 and 29 with coefficients of 0.5, and subtracter 26.
This is a non-recursive type digital filter consisting of an adder 30 and a characteristic expressed by equation (1). A first predicted signal is obtained using equation (1) and is supplied to terminal a of the switching circuit 34 and the comparison circuit 32. The second prediction circuit 22 is composed of a shift register that delays the input signal by 1365 sampling clock cycles and outputs the delayed signal, and the output of the second prediction signal is supplied to the terminal b of the switching circuit 34 and the comparison circuit 32. .
In addition, in order to initialize the first prediction circuit 21 of the transmitting/receiving predictor 6 and 15 at the beginning of each line in order to stop error propagation in the horizontal direction, the first prediction circuit 21 of each line is
At clock time the contents of registers 25, 27, 28 and 31 of predictors 6 and 15 are set to predetermined values, for example the value of zero.

The comparison circuit 32 compares which predicted signal is closer to the local decoded signal, and switches the signal to 0 if the first predicted signal is closer, and switches to 1 if the second predicted signal is closer to the local signal. A signal is output and supplied to the register 33. The switching signal delayed by one sampling clock period in the register 33 is sent to the switching circuit 34.
If the switching signal is 0, the switch x is connected to the terminal a and the first prediction signal is sent to the switching circuit 3.
The output of 4 is obtained. When the switching signal is 1, switch x is connected to terminal b and a second prediction signal is obtained at the output of switching circuit 34. In addition, in order to reset the error propagation in the horizontal and vertical directions, one-dimensional prediction (one-dimensional predictor ) is selected. The first or second predicted signal selected by the switching circuit 34 is output as a predicted signal of the predictor 6 or 12. In this circuit example, one of the two-dimensional predictors configured by adaptively switching the functions of equations (1) and (2).
A one-dimensional predictor is constructed using the one-dimensional prediction function of equation (1). Another example of a one-dimensional predictor is
When P ^(z) = 0, that is, PCM (PCM: Pulse
It is also possible to send a Code Modulation (Code Modulation) signal.

FIG. 3 is a block diagram showing a specific circuit example of the interpolation circuit 16 shown in FIG. 1. Input terminal 3
The decoded signal supplied to 5 is delayed by one line by shift register 44 in order to match the phase with the interpolation selection signal, and is then sent to terminal a of switching circuit 42 and shift register 38 of interpolation filter circuit 43. The interpolation filter circuit 43 receives the input signal.
A shift register 38 that delays the 681 sampling clock period and outputs the signal; and a shift register 39 that outputs the input signal after delaying the 681 sampling clock period.
This is a non-recursive type digital filter, which is composed of an adder 40, and a multiplier 41 with a coefficient of 0.5, and has filter characteristics shown in equation (3).

H(z)=0.5Z ^-681 +0.5Z ^-684 (3) In this interpolation filter, the average value of the decoded signals at two points where the subcarrier phase is the same one line before is the output of the interpolation filter. Note that correct interpolation cannot be performed for one or two samples at both ends of one line, so separate processing must be performed. The interpolated decoded signal is supplied to terminal b of the switching circuit 42. In the switching circuit 42, according to the interpolation selection signal inputted to the input terminal 36 for controlling the interpolation process, the switch x is connected to terminal a when interpolation is not to be performed, and the switch x is connected to terminal b when interpolation is to be performed.
The decoded signal as it is or the decoded signal that has been interpolated, in other words, the decoded signal that has been subjected to interpolation processing, is transferred from the output of the switching circuit 42 to the output terminal 3.
7 and sent to the D/A converter 17. Note that the shift register 38 of the interpolation circuit 16 and the shift register 22 of the predictor 15 can also be configured to be shared.

FIG. 4 is a block diagram showing a specific circuit example of the interpolation determination circuit 19 shown in FIG. 1. The parity check signal inputted to the terminal 51 has a signal of "0" when there is no error in the transmission line, and a signal of "1" when there is an error, and is supplied to the OR circuit 55. The predictor selection signal input to the terminal 52 has a signal of "0" when a two-dimensional predictor is selected, and a signal of "1" when a one-dimensional predictor is selected, and is supplied to the inverting circuit 53 and becomes "0". ” is inverted to “1” and “1” is inverted to “0” and supplied to the AND circuit 54. An interpolation selection signal is supplied from the output of the register 57 to the other input of the AND circuit 54, and the two input signals are ANDed and the output is sent to the logic circuit 55.
supplied to The OR circuit 55 performs the OR of the two input signals to obtain an interpolation selection signal, which is supplied to the register 56. Registers 56 and 5
7 operates with a clock of one horizontal scanning period, sets each input signal at the beginning of each line, and outputs each set signal. An interpolation selection signal delayed by one line is outputted from the register 56 and supplied to the output terminal 58 and the register 57. The interpolation selection signal outputted from the output terminal 58 of the interpolation determination circuit 19 is supplied to the interpolation circuit 16 as a signal of "1" when interpolation is to be performed, and as a signal of "0" when no interpolation is to be performed. Note that since the propagation of errors in predictive decoding depends on the characteristics of the predictor, if the characteristics of the predictor change, the configuration of the interpolation determination circuit 19 also changes.

As is clear from the above description, according to the present invention, by using a predictor having prediction coefficients configured using only signals of even lines including the current line, the prediction code can be reset while resetting the predictor when necessary. When a transmission error occurs, the receiving side performs error propagation on the decoded signal of the line that was incorrectly decoded by predictive decoding. Interpolation can be performed using the decoded signal of a correctly decoded line without any influence, and a decoded signal in which image quality deterioration due to transmission errors is hardly noticeable can be obtained.

Note that in the first embodiment, the predictors 6 and 12
is not limited to what is shown in FIG. For example, the two-dimensional predictor may be a predictor having the following prediction function using prediction between fields.

P ^(z) =0.5 ^z-1 +Z ^-262H -0.5Z ^-262H-1 (4) However, Z ^-H represents the delay of one line.

Furthermore, the interpolation circuit 16 is not limited to that shown in FIG. The interpolation determination circuit 19 is not limited to that shown in FIG. Furthermore, the sampling frequency is not limited to three times the frequency of the subcarrier. Although the first embodiment shows a case where the prediction error signal is equal-length encoded and decoded, the present invention is not limited to this. For example, the configuration may be such that unequal-length encoding and unequal-length decoding are performed. . Although the predictive encoder is shown as a recursive type, it can also be constructed as a non-recursive type without a feedback loop.

[Brief explanation of the drawing]

FIG. 1 is a block diagram showing the configuration of the first embodiment of the present invention, FIG. 2 is a diagram showing a specific circuit example of the predictors 6 and 15, and FIG. 3 is a specific circuit of the interpolation circuit 16. A diagram showing an example, FIG. 4 is an interpolation judgment circuit 19
FIG. 2 is a diagram showing a specific example of a circuit. 1, 11, 35 and 36 are input terminals, 2 is an A/D converter, 3 and 26 are subtracters, 4 is a quantizer, 5, 14, 30 and 40 are adders, 6 and 15 are predictors, 7 is a code converter, 8 is a parity addition circuit, 9, 18 and 37 are output terminals, 1
2 is a parity detection circuit, 13 is a sign inverter, 1
6 is an interpolation circuit, 17 is a D/A converter, 19 is an interpolation determination circuit, 21 is a first prediction circuit, 22 is a second prediction circuit (shift register), 24, 29 and 41 are multipliers, 25, 27, 28, 31 and 33 are registers, 32 is a comparison circuit, 34 and 42 are switching circuits, 38 and 39 are shift registers, 51, 52 are input terminals, 53 is an inversion circuit,
54 is an AND circuit, 55 is an OR circuit, 56 and 57 are registers, and 58 is an output terminal.

Claims (1)

[Claims] 1 A device that predictively encodes and decodes an image signal is configured using a two-dimensional predictor configured using only sampling pixels on the current line and the even-numbered previous line, and a two-dimensional predictor configured using sampling pixels only on the current line. a means for obtaining a prediction error signal from an image signal by performing predictive coding by switching a one-dimensional predictor for each predetermined line; a means for converting the prediction error signal into a signal to be transmitted; a predictive coding device comprising a means for adding a signal for error detection to a transmission signal so that transmission errors can be detected; means for detecting on a line-by-line basis, means for performing predictive decoding using a predictor having the same configuration as the transmitting side to obtain a decoded signal of an image from the received signal, and determining which line from the signal indicating the presence or absence of a detected transmission error. means for determining whether the decoded signal of the image contains an error, and means for interpolating the decoded signal of the erroneously decoded line using the decoded signal of the error-free line according to the determined result. A predictive encoding/decoding device comprising a predictive decoding device.