This invention relates to an inter-frame coding apparatus for a video signal.

It is known that an inter-frame coding method is used for transmitting a video signal in order to reduce the amount of information. This inter-frame coding method has a good compression effect on a video signal for a picture containing little motion such as a still picture, but is less effective in compression of the information amount for a video signal containing a great deal of motion. To remove this drawback, a motion compensated inter-frame coding method is known in which motions contained in a video signal are detected between frames and the video signal is coded by using the detected motions. One example of this method is disclosed in U.S. Patent No. 4,371,895, "CODED VIDEO SIGNAL TRANSMITTING AND RECEIVING SYSTEM". In the inter-frame coding, when a channel error is involved in transmitted data, the error damages a reproduced picture. To cope with this problem, a refreshing operation which transmits a video signal encoded without exploiting temporal correlation is periodically conducted. One example of encoding used for the refreshing operation is a previous value difference coding.

However, in the inter-frame coding method which uses motion compensation, a error moves spatially in a reproduced picture because of the motion compensation, resulting in the enlargement of the error. In this case, it is frequently difficult to remove the error by the refreshing operation and, at worst, the error cannot be removed fully.

Summary of the Invention

Accordingly, it is an object of the present invention to provide an inter-frame coding apparatus for a video signal which can securely refresh a picture when the motion compensation is used.

According to the present invention, there is provided a video signal coding apparatus comprising: a motion detector for detecting a motion contained-in the video signal to deliver a motion vector, an inter-frame prediction coder for coding the video signal in response to the motion vector, an intra-frame coder for coding the video signal by using a spatial correlation of the video signal for refresh operation, a control circuit for alternatively selecting the inter-frame prediction coder and the intra-frame coder to code the video signal, and a motion compensation inhibiting circuit for inhibiting the operation of the motion detector for a predetermined period with respect to an area where the refresh operation has been completed.

Brief Description of the Drawings

• Fig. 1 shows a refresh operation in conventional motion compensated inter-frame coding;
• Fig. 2 shows a refresh operation according to the present invention;
• Fig. 3 schematically shows an embodiment of the present invention in a case where a motion vector is transmitted to a receiver;
• Fig. 4 is a block diagram of a refresh control circuit shown in Fig. 3;
• Fig. 5 is a block diagram of a motion compensated inter-frame coder shown in Fig. 3;
• Fig. 6 is a block diagram of a motion compensated inter-frame decoder shown in Fig. 3;
• Fig. 7 is a block diagram of a variable length coder used in Fig. 3;
• Fig. 8 is a block diagram of a variable length decoder used in Fig. 3;
• Fig. 9 schemaitcally shows a second embodiment of the present invention in a case where a motion vector is not transmitted to a receiver;
• Fig. 10 is a block diagram of a motion compensated inter-frame coder shown in Fig. 9;
• Fig. 11 is a block diagram of a motion compensated inter-frame decoder shown in Fig. 9;
• Fig. 12 in a block diagram of a variable length coder used in Fig. 9; and
• Fig. 13 is a block diagram of a variable length decoder used in Fig. 9.

Description of the Preferred Embodiment

A refresh operation in the conventional motion compensated inter-frame coding will first be explained. In Fig. 1, it is assumed that coding is carried out at every block with N lines in the vertical direction, and a refreshing operation is conducted by N lines in the raster scanning direction every frame. In this case, it is also assumed that a vertical range of motion compensation does not exceed N lines. The hatched area shows the N lines where refresh is to be carried out. Now, it is supposed that the N lines including a block B are refreshed at the time t_l and an error 1 is generated with respect to an arbitrary block A. At the time t₂, the motion compensated inter-frame coding and decoding for the block B which is in the vicinity of the block A are conducted by a latest vector V₂ which has been detected with respect to the block B and which has a magnitude and direction as shown in Fig. 1. In this case, the coding by using the vector V₂ on a transmitting side is executed correctly. On a receiving side, however, when decoding of the block B by using the vector V₂ is carried out, since the block A includes the error, this error in the block A moves to the block B which has just been refreshed at the time t_l. Therefore, the moved error becomes an error 2. At the time t₃, the error 2 included in the block B is moved to become an error 3 by a latest vector V₃ which has been detected with respect to a block C which is in the vicinity of the block B and which has a magnitude and direction as shown in this Figure. In this way, as time lapses the error shifts or enlarges, and it cannot be removed by the conventional refresh operation.

Next, the principle of the present invention will be described with reference to Fig. 2. The above-described drawback in the prior art is caused by the fact that motion compensation is conducted even in the N lines where the refresh operation has been just undertaken, just as in other places. Therefore, if the motion compensation is inhibited at least in the N lines where refreshing operation has been just finished, the above drawback can be overcome. This principle will be explained in detail, referring to Fig 2. As in Fig. 1, it is supposed that an error is generated in a block A at the time t_l. In the present invention, N lines (the hatched area) including a block are refreshed and the motion compensation is inhibited (hereinunder it is called "motion compensation inhibition line") at least in the K lines where the refresh operation has just been finished (the stippled area). At the time t₂, since the motion compensation is inhibited at least in the N lines where refresh operation has been undertaken, there is no spatial shift of the error in the block A to the block B unlike the case in Fig. 1. Therefore, there is no destruction of any picture element within the motion compensation inhibition line, and the error 1 generated in the block A at the time t₁ is refreshed and removed. At the time t₃, since the error has been removed at the time t_2' there is no possibility of the error movement or the error enlargement. Consequently, in the present invention, since there is no destruction of the picture element due to the fact that an error has moved to the area where the refresh operation has already been completed, the error is quickly removed by only one refresh operation. It is noticed that when the vertical range for motion compensation is N (N_v> N), the N_v line is designated as the motion compensation inhibition line.

Hereinunder, the embodiments of the present invention will be described in detail with reference to the accompanying drawings. Fig. 3 schematically shows an embodiment of an inter-frame encoding and decoding apparatus using the motion compensation. A digitaized television video signal is fed to a motion compensation inter-frame coder 5 from an input signal line 1. A refresh control circuit 2 produces a refresh execution signal and a motion compensation inhibition signal and supply them to the motion compensation inter-frame coder 5 via signal lines 3 and 4 to control the refresh operation and the inhibition of the motion compensation. In lines where the refresh operation is designated, coding such as, for example, previous value difference coding which is not inter-frame prediction coding is conducted. A prediction error e and a motion vector v, delivered from the motion compensation inter-frame coder 5 are supplied to a variable length coder 9 through signal lines 6a and 6b. The refresh . execution signal and the motion compensation inhibition signal which have equalized their phases to the other_' . signals e and v from the motion compensation inter-frame coder 5 are supplied to the variable length coder 9 through a signal line 7 and a signal line 8, respectively. In this case, when the motion compensation inhibition signal indicates the inhibition, the motion vector becomes zero. The motion vector, the prediction error signal which are converted into variable length codes, and the refresh execution signal are transmitted to a transmission line 10 to be fed to a receiving side.

In the receiving side, a variable length decoder 11 decodes the prediction error signal, the moticn vector and the refresh execution signal, and supplies them to a motion compensation inter-frame decoder 15 through a signal line 12a, a signal line 12b and a signal line 13, respectively The motion compensated inter-frame decoder 15 basically decodes the prediction error signal in accordance with the motion vector to reproduce a television video signal. But, when the refresh operation is conducted in a certain area (N lines), the decoder 15 reproduces the television video signal in accordance with a coding method which has been used on the transmitting side in the refresh operation, for example, the previous value difference coding. In this embodiment, since the motion compensation inhibition corresponds to the motion vector of zero, the decoder 15 reproduces the television video signal in accordance with the inter-frame prediction coding without the motion compensation when the motion compensation is inhibited. In this way, when the moving vector is transmitted from the transmitting side to the receiving side, a correct decoding can be performed without feeding the motion compensation inhibition signal. The signal decoded by the motion compensation inter-frame decoder 15 is a digital signal and supplied to an output signal line 16 as the television videc signal.

The refresh circuit 2 will next be explained with reference to Fig. 4. A refresh clock which is generated in a control circuit 17 is supplied to a refresh counter 19 through a signal line 18, which produces the refresh execution signal by, for example, counting the line number. The refresh execution signal generated by the refresh counter 19 moves by the N lines at every frame or several frames. The refresh execution signal is supplied to the motion compensation inter-frame coder 5 shown in Fig. 3 through the signal line 3 and to a register 21. The register 21 delays the refresh execution signal by the time of the N lines in response to a set pulse which is supplied from the control circuit 17 through a signal line 20, and transmits to a signal line 4 as the motion compensation inhibition signal. This inhibits the motion compensation in the N lines where refresh operation has just been completed.

Now the motion compensation inter-frame coder 5 will be explained with reference to Fig. 5. The digitized television signal is supplied to a subtracter 22 and a motion vector detector 30 through the input signal line 1. The motion vector detector 30 detects an optimum signal, namely the motion vector, based on, for example, a preceding frame of the video signal and a present frame of the video signal. The motion compensation inhibition signal is supplied to the motion vector detector 30 through the signal line 4, and when the motion compensation is inhibited, the motion vector designating inter-frame prediction, i.e., the motion vector of zero is delivered. The motion vector delivered from the motion vector detector 30 is supplied to a line 6b and a variable delay circuit 29. The variable delay circuit 29 controls a delay time of a locally decoded signal which is delivered from a frame memory 28 in response to the given motion vector and supplies it to the subtracter 22 and an adder 24. The subtractor 22 delivers a difference between the present frame video signal from the signal line 1 and the variably delayed video signal generated about one frame period before which is supplied from the variable delay circuit 29.

The result obtained by the subtracter 22, namely, a prediction error signal, is quantized by a quantizer 23 and supplied to a selector 31 and the adder 24. The adder 24 reproduces the locally decoded signal by using the quantized prediction error signal and the signal generated about one frame period before which is delivered from the variable delay circuit 29. The locally decoded signal is supplied to a DPCM encoder 25 and a selector 27. The DPCM encoder 25 codes the locally decoded signal by the use of an intra-frame coding method to produce the previous value difference, and supplies it to the selector 31 as a refresh signal and to a DPCM decoder 26. The DPCM decoder 26 is subjected to inter-frame local decoding and supplies the decoded signal to the selector 27. The selector 27 selects the decoded signal delivered from the DPCM decoder 26 and supplies it to the frame memory 28 when the refresh operation is performed in accordance with the refresh execution signal which is supplied through the signal line 3. When refresh is not carried out, the selector 27 selects the locally decoded signal delivered from the adder 24 and supplies it to the frame memory 28. The selector 31 selects the refresh signal delivered from the DPCM encoder 25 instead of the prediction error signal in accordance with the refresh execution signal, and supplies the selected one to the variable length encoder 9 (Fig. 3) through the signal line 6a. The motion compensation inhibition signal and the refresh execution signal are fed to the lines 7 and 8 through delay circuits 14a and 14b.

Referring to Fig. 6, the motion compensation decoder 15 (Fig. 3) will be explained. In the Figure, the -prediction error signal delivered from the variable length decoder 11 (Fig. 3) is supplied to an adder 34 through the line signal 12a. The motion vector is supplied to a variable delay circuit 33 through the line signal 12b so that the delay amount is controlled in response to the speed and direction of the motion vector. As explained above, since the motion vector is zero when motion compensation is inhibited, inter-frame decoding without motion compensation is conducted. The signal decoded about one frame period before is supplied from the frame memory 37 to the variable delay circuit 33, and the delay time is controlled. The adder 34 performs motion compensated decoding by adding the prediction error signal supplied through the signal line 12a and the signal decoded about one frame period before, which is subjected to the delay time control and which is supplied from the variable delay circuit 33. Thereby, the adder 34 feeds the added signal to a selector 36. A DPCM decoder 35 decodes the intra-frame coded signal delivered through the signal line 12a in an intra-frame decoding manner when the refresh operation is conducted, and supplies the decoded signal to the selector 36 as a refresh signal. The selector 36 conducts a change-over action in accordance with the refresh execution signal which is fed through the signal line 13, such that it selects the output of the DPCM decoder 35 when refreshing is executed and it selects the output of the adder 34 in the other case. The output delivered from the selector 36 is supplied to the frame memory 37 and, at the same time, is transmitted to a signal line 16 as an output from the motion compensation inter-frame decoder 15, i.e., a reproduced digital television video signal.

The structure of the variable length encoder 9 (Fig. 3) will be described with reference to Fig. 7. The prediction error signal (or the refresh signal) supplied through the line 6a is coded in a variable length coding circuit 38 and is fed to a multiplex circuit 40. The refresh execution signal supplied through the line 7 is coded a coding circuit 39 and is fed to the multiplex circuit 40. The motion vector from the line 6b is controlled by the motion compensation inhibition signal from the line 8 in a reset circuit 41 and is fed to a variable length coding circuit 42.

Thus, when motion compensation is inhibited,in the reset circuit 41, the motion vector becomes zero. The multiplex circuit 40 multiplies the prediction error signal, the refresh execution signal and the motion vector signal from the circuits 38, 39 and 42, respectively, and feeds them to the transmission line 10.

Fig. 8 shows the structure of the variable length decoder 11 (Fig. 3). The prediction error signal (or the refresh signal), the motion vector signal and the refresh execution signal sent through the transmission line 10 are separated by a distribution circuit 43 and are decoded in variable length decoding circuits 44 and 45, and a decoding circuit 46, respectively. The respective signals are sent to the motion compensation inter-frame decoder 15_.

Next, a second embodiment of the invention, in which a motion vector is not transmitted to a receiving side, will be explained. If, for example, the system which is reported on pages 631-670 of Bell System Technical Journal, March, 1979 under the title of "Motion Compensated Television Coding: Part I" by A. N. Netravali, et al., is used, the motion vector is detected only by the use of the locally decoded signal, and, in this case, it is unnecessary to transmit the motion vector from a transmitting side to the receiving side.

In Fig. 9 showing a second embodiment of the present invention, only a prediction error signal is supplied from a motion compensation inter-frame coder 5' to a variable length coder 9' through a line 6a. Namely, the motion vector is not transmitted. The variable length coder 9' transmits the prediction error signal, a refresh execution signal and a motion compensation inhibition signal in a multiplex manner to the receiving side. In the receiving side, a variable length decoder 11' decodes the prediction error signal, the refresh execution signal and the motion compensation inhibition signal and supplies them to a motion compensation inter-frame decoder 15' through respective lines 12a, 13 and 13a.

In Fig. 10 showing the block diagram of the motion compensation inter-frame coder 5', a motion vector detector 30', which controls a variable delay circuit 29,detects a motion vector by receiving a locally decoded video signal delivered from a selector 27 and a video signal delivered from an intermediate tap of a frame memory 28. In this figure, the motion vector detected by the detector 30' is not transmitted to the receiving side. In Fig. 11 showing the block diagram of the motion compensation inter-frame decoder 15', a motion vector detector 30" detects the motion vector in response to a video signal delivered from a selector 36 and a video signal delivered from an intermediate top of a frame memory 37 as the same of Fig. 10, and controls a variable delay circuit 33.

Fig. 12 shows the block diagram of the variable length coder 9'. In this figure, the prediction error signal is coded by a variable length coding circuit 38, and the refresh execution signal and the motion compensation inhibition signal are coded by respective coding circuits 39 and 47. A multiplex circuit 40' multiplexes those coded signals and transmits them to a transmission line 10.

Fig. 13 shows the block diagram of the variable length decoder 11'. In this figure, the multiplexed signals are separated into the coded prediction error signal, the coded refresh execution signal and the coded motion compensation inhibition signal by a distribution circuit 43'. The coded prediction error signal is decoded by a variable length decoding circuit 44 and transmitted to a line 12a. The coded refresh execution signal and the coded motion compensation inhibition signal are decoded by respective decoding circuits 46 and 48, and supplied to respective lines 13 and 13a.

In the paper described above, coding is carried out at every line, but the inhibition of the motion compensation in the second embodiment is executed for a period corresponding to the number of lines in the vertical direction of the motion compensation, as in the

first embodiment.

In the embodiments, the intra-frame coding such as the DPCM of the locally coded signal is used for producing the refresh signal. Alternatively, the DPCM of the input digital video signal (PCM) supplied from the line 1 for the input to the DPCM encoder 25 or usage of the locally decoded signal may be applied to the refresh signal. In the latter case, the DPCM decoder 26, the selector 27 and the DPCM decoder 35 are unnecessary. Further, in place of the DPCM coding, an orthogonal transform coding may be used. In this case, the component of the referential numeral 25 is an orthogonal transform encoder and the component of the referential numeral 35 is an orthogonal conversion decoder.

As described above in detail, according to the present invention, the error which is generated in the motion compensation inter-frame coding and the motion compensation inter-frame decoding can be removed quickly before it moves and enlarges. Accordingly, practical use of the present invention brings beneficial effects.