JP2877906B2 - Transmission Code Error Detection for Interframe Coding - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION [Industrial Application Field] The present invention quickly recovers from damage to a reproduced image caused by a transmission code error in, for example, interframe coding suitable for compressing the amount of data of a moving image. The present invention relates to a transmission code error detection method used for processing.

[Conventional technology]

For high-efficiency coding of moving images, one screen memory (frame memory) having common contents is prepared on the transmission side and the reception side, and the transmission side compares the contents of the screen to be transmitted and the contents of the frame memory with each other. Inter-frame coding that transmits only the part that has occurred is effective.

In inter-frame coding, only the difference signal between the local decoded signal of the previous frame and the input image signal is transmitted, so if a code error occurs once during transmission, the contents of the receiving frame memory are replaced with the contents of the transmitting frame memory. Unlike that, the effect lasts for a long time.

Therefore, in the inter-frame coding, a technique for updating (refreshing) the contents of each frame memory for transmission and reception to compensate for the effect of the transmission code error is indispensable.

One of the periodic refreshes, for example, from top to bottom of the screen, for each part thereof, does not periodically perform inter-frame coding sequentially, transmits the input image signal as it is using predetermined coding, Refresh the contents of each frame memory for transmission and reception. Therefore, even if the reproduced image of the decoder is deteriorated due to a transmission code error, the effect can be settled if refreshing is completed for one screen.

However, in this method, the coding efficiency of the input image signal is low, and a large amount of code is required even when, for example, intra-frame coding is used, which hinders transmission of video at a low bit rate. That is, if the refresh cycle is shortened, the inter-frame coding characteristics deteriorate, and if the refresh cycle is lengthened to improve the effect, it is inevitable that the influence of the once generated code error is prolonged.

Another demand refresh is a technique for solving such a problem of the periodic refresh (see Japanese Patent Application Laid-Open No.
-41069).

That is, if the occurrence of a transmission code error is detected on the receiving side by any method, the occurrence of the transmission code error is notified to the transmitting side via a line in the direction opposite to the direction of transmission of the video signal. On the transmitting side, in response to this notification, the entire screen is transmitted at once using a predetermined encoding and refreshed. Therefore, according to this method, when a transmission code error occurs, its effect can be quickly cleared.

To detect a transmission code error in demand refresh, for example, as described in JP-A-60-29068, the theoretical "1" of the transmission side frame memory is integrated for each bit plane, and a predetermined number is calculated. The remainder is sent to the receiving side, and the receiving side determines the same remainder. If there is a mismatch, it is determined that a code error has occurred. The simplest and most effective method is to compare parity with a predetermined number of two.

[Problems to be solved by the invention]

By the way, with recent advances in coding technology, a technology for transmitting moving images using a low-speed (for example, 64 kbit / s) line has been developed with higher efficiency than conventional inter-frame coding.

The main technology is hybrid coding that combines transform coding with conventional inter-frame coding, and performs a discrete cosine transform (DCT), which is one of orthogonal transforms, on the inter-frame prediction error. This is a system in which coefficients are quantized and variable-length coded and transmitted (CCITT Recommendation H.261).

In addition, after performing the inverse quantization and inverse DCT (IDCT), the quantization result is summed with the local decoded signal of the immediately preceding frame, and is used as the local decoded signal for the processing of the next frame. To store.

Here, inverse particleization and inverse DCT are processes performed on both the transmitting side and the receiving side.If exactly the same inverse DCT circuit is used on the transmitting side and the receiving side, the local decoding signal on the transmitting side and the decoder on the receiving side are used. The outputs completely match, and the conventional transmission code error detection method can be applied for the demand refresh.

However, the CCITT (International Telegraph and Telephone Consultative Committee) has recommended H.26 for standardization of video conferencing / telephone coding systems.
In 1, a variety of designs are allowed for the inverse DCT circuit including irrational operations. That is, since an allowable error is defined in the inverse DCT calculation accuracy, different inverse DCT circuits may be used on the transmission side and the reception side.

In that case, even if there is no transmission code error, if there is a difference in the calculation method or the way of rounding the lower bits, the local decoded signal on the transmitting side and the decoder output on the receiving side will not match. Therefore, the conventional parity for each bit plane cannot be used for detecting a code error, and it has been difficult to detect a transmission code error for demand refresh.

Thus, if demand refresh becomes difficult,
To compensate for the effects of transmission code errors, as well as to compensate for the accumulation of inverse DCT computation errors on the transmitting and receiving sides, one had to rely on time-consuming periodic refreshes.

The present invention provides a transmission code error detection method in inter-frame coding that reliably detects a transmission code error and enables application of demand refresh even when an inverse DCT circuit that allows various designs is used. The purpose is to:

[Means for solving the problem]

According to the first aspect, an encoded image signal is generated by orthogonally transforming a difference signal between an input image signal and a local decoded signal of a previous frame output from a frame memory, and performing quantization and variable length encoding. And an encoding unit that performs an inverse quantization and an inverse orthogonal transform on the quantization result, and then sums the result with the locally decoded signal of the previous frame and stores the sum in the frame memory as a locally decoded signal to be used for processing of the next frame. , Facing the encoding unit via a communication line, captures the transmitted encoded image signal, performs variable-length decoding, inverse quantization, and inverse orthogonal transformation in the reverse process of the encoding unit, and further performs frame processing. A decoding unit that takes the sum of the decoded signal of the previous frame output from the memory and outputs the sum as an output image signal, and accumulates in the frame memory as a decoded signal to be used for processing of the next frame. In frame-time coding scheme,
An encoder including an encoding unit receives a local decoded signal from the encoding unit and calculates a sum of pixel levels in a predetermined unit, and a multiplexing unit that multiplexes the transmitted sum to an encoded image signal. A decoding circuit including a decoding unit, a separation circuit for separating a coded image signal transmitted from a communication line and a sum value, and a decoded signal from the decoding unit. A summation circuit for calculating the sum value of the pixel level of the unit, and a difference between the sum value obtained by the decoder and the sum value sent from the encoder and separated by the separation circuit, and an absolute value of the difference is defined as a predetermined value. And a comparator for determining that there is a transmission code error in the coded image signal when the transmission code error is exceeded.

According to a second aspect of the present invention, in the transmission code error detection system in the inter-frame encoding according to the first aspect, the summation circuit of the encoder includes the sum of the local decoded signal and the local decoded signal of the previous frame instead of the local decoded signal. And outputs the inverse orthogonal transform output before taking the sum with the decoded signal of the previous frame into the summation circuit of the decoder instead of the decoded signal.

According to a third aspect of the present invention, in the transmission code error detection system in the inter-frame encoding according to the first aspect, the summation circuit of the encoder takes in the inversely quantized output instead of the locally decoded signal, The summation circuit of the decoder is configured to take in the dequantized output instead of the decoded signal.

According to a fourth aspect of the present invention, in the transmission code error detection system in the inter-frame encoding according to any one of the first to third aspects, each of the summation circuits includes: Output only the predetermined middle bits,
This configuration is used for multiplex transmission and comparison processing.

(Operation)

Even if various designs are allowed for the inverse DCT circuit, it is not unlimited.

If the inverse DCT circuits of the transmission-side encoder and the reception-side decoder are significantly different, the error (mismatch error) is generated and accumulated each time the inverse DCT operates, and finally the transmission side The discrepancy between the contents of the frame memories on the receiving side and the receiving side becomes apparent at a glance.

Therefore, CCITT Recommendation H.261 defines a reference inverse DCT, which is used when the same test data is input.
It is determined so that the difference between the circuit output and the output of the inverse DCT circuit used is sufficiently small. For example, the maximum mismatch error (absolute value) is 1 for any pixel, and the average (absolute value) of the mismatch error obtained for 10,000 blocks of 8 × 8 pixels is 0.
It does not exceed 0015. Therefore, the reverse
The DCT mismatch error has an amplitude of 1 and a substantially positive / negative symmetric distribution.

On the other hand, if the contents of the frame memory on the decoder side are damaged by a transmission code error and there is something that is visible,
The error between the contents of each frame memory on the encoder side and the decoder side is larger in amplitude than the mismatch error of the inverse DCT,
In addition, the positive and negative symmetry is low.

The present invention focuses on the difference between the characteristics of the mismatch error of the inverse DCT and the error caused by the transmission code error.

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

In the figure, the sum of the pixel levels of the same region is calculated by sum circuits 33 and 63 provided on the encoder 10 side and the decoder 50 side, respectively, and the sum value obtained by the encoder 10 is transmitted to the decoder 50.
By comparing the sum values with the comparator 65, a transmission code error can be identified and its occurrence can be detected.

That is, one frame per horizontal to the luminance Y is 352 pixels, the dye C _b, C _r is the 176 pixels, the luminance Y is 288 pixels in the vertical direction, the color difference C _b, C _r is the 144 pixels, the image from the total 152,064 pixels For the signal (common intermediate format), the sum of the pixel levels of the entire screen (0 to 255 because each pixel level is represented by 8 bits) is obtained.

Here, since the mismatch error of the inverse DCT is substantially symmetrical in the positive and negative directions as described above, the sum of the local decoded signals (inverse orthogonal transform output, inverse quantization output) on the encoder 10 side and the decoder 50
The sums of the decoded signals (inverse orthogonal transform output, inverse quantized output) on the side are almost equal, and the absolute value of the difference is 0.0015 × 152064 = 228 at the maximum.

On the other hand, the error due to the transmission code error, depending on which part of the variable length code to be transmitted has an error, may cause a slight deterioration in the restored image, or may cause large damage that may be annoying, Practically, it is sufficient if the latter can be detected. For example, if 64 pixels having a difference of 64 levels are generated, the difference becomes a maximum of 4096. However, errors due to transmission code errors cancel out positive and negative, and the difference of 4096 is rare.

However, there is a high probability that the mismatch error is larger than 228, which is the maximum value of the mismatch error of the inverse DCT. For example, based on 256 (2 ⁸ ), the local decoded signal (inverse orthogonal transform output, inverse quantization output) on the encoder 10 side is used. If the difference between the sum of the above and the sum of the decoded signals (inverse orthogonal transform output, inverse quantization output) on the decoder 50 side is larger than this reference value, it can be determined that a transmission code error has occurred.

〔Example〕

Hereinafter, embodiments of the present invention will be described in detail with reference to the drawings.

FIG. 2 is a block diagram showing a configuration of an embodiment according to the first aspect of the present invention.

The present embodiment is applied to a hybrid coding scheme in which inter-frame coding and transform coding are combined, and shows a configuration example of a transmitting-side encoder and a receiving-side decoder.

In the figure, an input image signal a is subtracted by a subtracter 11 using a local decoded signal b one frame before output from a frame memory 13 as a prediction signal, and is converted into a prediction error signal c.
The prediction error signal c is converted through a DCT circuit 15 into a DCT coefficient d indicating a frequency component, and the DCT coefficient d is quantized through a quantization circuit 17 to obtain a quantization index e.
Is output as

The quantization index e is returned to the quantization representative value f of the DCT coefficient area via the inverse quantization circuit 19, and the quantization representative value f is again returned to the quantization prediction value of the pixel area via the inverse DCT circuit 21. It is converted into an error signal g. The quantized prediction error signal g is output by the adder 23 from the frame memory 13 to 1
The result is added to the locally decoded signal b before the frame, and is stored in the frame memory 13 as a locally decoded signal h used for processing of the next frame.

On the other hand, the quantization index e is
, And is transmitted to the communication line 41 via the multiplexing / demultiplexing circuit 27 as an encoded image signal i.

The basic configuration of the encoder 10 has been described above. As a refreshing means, the transmission code error detection signal j transmitted from the receiving-side decoder to the transmitting-side encoder and separated by the multiplexing / demultiplexing circuit 27 generates a common signal. A switch circuit 2 that switches the connection destination of the contact from the break contact α to the make contact β and inputs the input image signal a directly to the DCT circuit 15 without passing through the subtractor 11
9 and a switch circuit 31 for stopping the input of the local decoded signal b output from the frame memory 13 to the adder 23.

 Next, the basic configuration of the decoder 50 will be described.

The multiplexing / separating circuit 51 separates a coded image signal k (← i) from a signal received from the communication line 41. The coded image signal k is restored as a quantization index 1 via a variable length decoding circuit 53, and further, the quantization index l
Is returned to the quantization representative value m in the DCT coefficient area via the inverse quantization circuit 55. The representative quantization value m is the inverse DCT of the encoder 10.
The signal is converted into a quantized prediction error signal n of the pixel region through an inverse DCT circuit 57 similar to the circuit 21. This quantized prediction error signal n is added by an adder 59 to the decoded signal o of the previous frame output from the frame memory 61, output as an output image signal p, and used by the frame memory 61 for processing of the next frame. It is stored as a decoded signal.

Here, the configuration of the present invention is characterized in that, in the present embodiment, the encoder 10 takes in the local decoded signal h output from the adder 23 and, for example, clears it to zero at the beginning of the screen, And a sum circuit 33 for sequentially summing the pixels of the entire screen and sending the sum q to the communication line 41 via the multiplexing / demultiplexing circuit 27.

Further, in the decoder 50, the decoded signal p output from the adder 59 is fetched, and after being cleared to, for example, zero at the beginning of the screen in synchronization with the operation of the encoder 10, the pixels of the entire screen to be coded are processed. A summation circuit 63 for taking a successive sum and outputting the sum r, and a summation r and a multiplexing / demultiplexing circuit
The sum s (←) notified from the encoder 10 separated at 51
q), and a multiplexing / demultiplexing circuit 51 for transmitting the transmission code error detection signal t (→ j) output from the comparator 65.
Via the communication line 41.

The sum circuits 33 and 63 need only have a length of 26 bits in the case of an encoder or a decoder that handles a common intermediate format. When all the pixels of the target screen have been encoded, the summation circuit 33 outputs the sum value q of 26 bits, and receives it via the communication line 41 together with the encoded image signal i via the multiplexing / demultiplexing circuit 27. To the decoder 50 on the side.

By the way, the inverse DCT circuit 21 of the encoder 10 and the inverse DCT circuit of the decoder 50
The circuit 57 usually has a difference in its characteristics as described above,
The prediction error signals g and n output from each circuit include an inverse DCT mismatch error.

Therefore, the sum value q output by the sum circuit 33 which takes in the local decoded signal h and sums the pixels of the entire screen to be coded, and the sum circuit 63 which takes in the decoded signal p and sums the pixels of the entire screen in the same manner. This mismatch error exists between the output sum value r and an error based on a transmission code error sometimes being superimposed.

The comparator 65 calculates a sum value s and a sum value r corresponding to the sum value q.
If the absolute value exceeds a predetermined threshold, it is determined that a transmission code error has occurred on the communication line 41 based on the above-described principle, and the encoder is determined by the transmission code error detection signal t.
Notify the 10 side.

In the encoder 10, the transmission circuit error detection signal j corresponding to the transmission code error detection signal t is used to switch the switch circuit on the next screen.
By switching between 29 and 31, an image signal encoded by intra-frame encoding instead of inter-frame encoding is sent, and the contents of the frame memories 12 and 61 on the transmission side and the reception side are refreshed.

FIG. 3 is a diagram for explaining a timing relationship for detecting a transmission code error according to the method of the present invention.

For the input image signal a of the screens A, B, C, D,.
Since the sum value q _A ,... Corresponding to the screen A is known by the encoder 10 after the coding processing, the sum value q _A , for example, is inserted into a portion following the header of the coded image signal i of the next screen.

The decoder 50 obtains the total value r _{A at the} time T _A1 at which the decoding process of the encoded image signal k corresponding to the screen A ends, and the time at which the head of the encoded image signal k corresponding to the subsequent screen B is received.
At T _A2 , the sum S _A (q _A ) sent from the encoder 10 can be known.

Therefore, the comparator 65 can notify the encoder 10 side of the transmission code error detection result based on the comparison result after the time _TA2 .

 Next, another embodiment will be described.

According to the second aspect of the present invention, in the configuration of the embodiment shown in FIG. 2, the input of the summing circuit 33 of the encoder 10 is not the local decoded signal h but the output of the inverse DCT circuit 21. g, and the input of the summation circuit 63 of the decoder 50 is also the quantization prediction error signal n which is the output of the inverse DCT circuit 57.

In this configuration, only the pixels corresponding to the codes transmitted on the communication line 41, that is, the pixels that have changed from the previous screen are checked, but the prediction error signal has a distribution centered on zero, and The bit lengths of the values q (s) and r can be reduced as compared with the first embodiment. That is, by reducing the number of transmission bits, it is possible to suppress a decrease in coding efficiency due to transmission of the sum value.

According to the third aspect of the present invention, in the configuration of the embodiment shown in FIG. 2, the input of the summing circuit 33 of the encoder 10 is not the local decoded signal h but the DCT domain quantum which is the output of the inverse quantization circuit 19. Similarly, the input of the summation circuit 63 of the decoder 50 is also set to the quantization representative value m of the DCT coefficient area which is the output of the inverse quantization circuit 55.

According to this configuration, if the object of the sum is a DC component, an effect equivalent to the configuration of the invention according to claim 2 can be obtained. In addition, the target of the summation may be a plurality of appropriate quantized representative values (AC components).

By the way, in the embodiment of the invention described in claim 1, when the coding screen is a common intermediate format, the sum of the pixel levels of the entire one screen is 26 bits. There is a case where an excessive load is imposed in terms of coding efficiency.

Therefore, the minimum necessary bits for expressing the sum values q and r output by the sum circuits 33 and 63 will be described with reference to FIG. The same applies to the other embodiments described above.

Of the 26 bits, the lower bits are uncertain due to the inverse DCT mismatch error even if there is no transmission code error. Since the variation is about 200 as described above, at least the lower 7 bits (128) can be invalidated.

In this way, the lower u bits of the total value are truncated, and (26
-U) If there is a difference of two or more compared with the sum represented in bits, it can be determined that "there was a transmission code error". That is, if the original sum exceeds 2 ^u +1 to 2 ^{u + 2} −1, it can be determined that “a transmission code error has occurred”.

Now, assuming that u = 8, the sum value and the result of truncation of the sum value (sum value expressed by the upper 18 bits) are as shown in FIG.
As shown in the figure, it becomes step-like, and "there was a transmission code error."
Is determined when the difference of the original sum exceeds 257 to 1023. That is, if the difference between the sum value s (q) and the sum value r is 256 or less, it is determined that there is no transmission code error, and if the difference is 1024 or more, it is reliably determined that there is a transmission code error. .

If the difference between the original sum values is in the range of 257 to 1023, the judgment result becomes indeterminate if u bits are truncated.

Even if the upper bits of the total value are not sent, there is no problem in practice.

That is, when the high-order bits are truncated, only the low-order w bits of the (26-u) bits resulting from the truncation of the sum are transmitted from the encoder, and the corresponding w-bits are used by the decoder. If N is a natural number, then at full scale, just N × ^2w × ^2u or N ×
Even if there is a difference of ( ^2w ± 1) × ^2u , the difference is not detected because the upper bits are truncated, and the original “transmission coding error” is determined to be “transmission code error”. Is to be done.

However, the probability of occurrence of such a situation is 3/2 ^w , and even if the determination is incorrectly made on one screen, the probability of detection is high on the next screen.

FIG. 4B shows the possible bit positions of the sum when u = 8 and w = 8, and FIG. 4C shows the principle of erroneous determination.

In FIG. 4 (3), for example, when the result of the truncation of the total value on the transmitting side is 0, the result of the truncation of the total value on the receiving side is 0, ± 1, respectively. 2 ^{8 + 8} , 2 × 2 ^{8 + 8} , ...
,... Are also determined to be “no transmission code error”. However, the probability is 3/256.

The case where the sum is obtained in units of one screen by simple interframe coding has been described above. However, in the case of frame coding with motion compensation, the unit of the sum is reduced to a fraction of one screen, or The effectiveness of the present invention does not change regardless of the case where the size is increased to several screens.

〔The invention's effect〕

As described above, the present invention can reliably detect a transmission code error on the decoder side even in an interframe coding method in which an inverse DCT mismatch error exists, and can apply demand refresh.

Further, it is possible to efficiently and effectively take measures against transmission code errors regardless of the encoding target image, the used bit rate, and the like.

[Brief description of the drawings]

FIG. 1 is a block diagram showing a basic configuration of the present invention. FIG. 2 is a block diagram showing the configuration of an embodiment according to the first aspect of the present invention. FIG. 3 is a diagram for explaining a timing relationship for detecting a transmission code error according to the method of the present invention. FIG. 4 is a view for explaining the minimum necessary bits for expressing the sum value output by the sum circuit. 10 ... Encoder, 11 ... Subtractor, 13 ... Frame memory,
15: DCT circuit, 17: Quantization circuit, 19: Inverse quantization circuit, 21: Inverse DCT circuit, 23: Adder, 25: Variable length encoding circuit, 27: Multiplexing / demultiplexing Circuit, 29, 31 Switch circuit, 33 Sum circuit, 41 Communication line, 50 Decoder, 51 Multiplexing / demultiplexing circuit, 53 Variable length decoding circuit, 55 Reverse Quantization circuit, 57 ... Inverse DCT circuit, 59 ... Adder, 61 ... Frame memory, 63 ... Summation circuit, 65 ...
Comparator.

──────────────────────────────────────────────────続 き Continued on the front page (58) Field surveyed (Int.Cl. ⁶ , DB name) H04N 7/24-7/68 H04N 1/41-1/419

Claims (4)

(57) [Claims] An orthogonal transformation of a difference signal between an input image signal and a local decoded signal of a previous frame output from a frame memory, quantization and variable length encoding to generate an encoded image signal, and a quantization result After performing inverse quantization and inverse orthogonal transform on the encoding unit, the encoding unit takes the sum of the local decoding signal of the previous frame and stores the sum in the frame memory as a local decoding signal to be used for processing of the next frame. Through the encoder, captures the transmitted encoded image signal, performs variable-length decoding, inverse quantization, and inverse orthogonal transform in the reverse process of the encoder, and further outputs from the frame memory. And a decoding unit that takes the sum of the decoded signal of the previous frame and outputs the sum as an output image signal, and stores the decoded signal in the frame memory as a decoded signal to be processed for the next frame. In the inter-frame encoding method, an encoder including the encoding unit includes a summation circuit that fetches the local decoded signal from the encoding unit and obtains a sum value of a pixel level of a predetermined unit; A multiplexing circuit for multiplexing the sum to an encoded image signal, wherein a decoder including the decoding unit separates the encoded image signal and the sum from the encoded image signal transmitted via the communication line. A summation circuit that fetches the decoded signal from the decoding unit and obtains a sum of pixel levels in a predetermined unit; and a sum obtained by the decoder, A comparator that takes a difference from the sum total value separated by the circuit and, when the absolute value exceeds a predetermined value, determines that there is a transmission code error in the encoded image signal. Transmission in interframe coding Code error detection method. 2. A transmission code error detection method in inter-frame coding according to claim 1, wherein the summation circuit of the encoder has a configuration in which the sum with the local decoded signal of the previous frame is obtained instead of the local decoded signal. A configuration in which the inverse orthogonal transform output is taken in, and the summation circuit of the decoder takes in the inverse orthogonal transform output before taking the sum with the decoded signal of the previous frame in place of the decoded signal. A transmission code error detection method in encoding. 3. A transmission code error detection system according to claim 1, wherein the summation circuit of the encoder is configured to receive an inversely quantized output instead of the local decoded signal. A transmission code error detection method in inter-frame coding, wherein a circuit takes in an inversely quantized output instead of a decoded signal. 4. A transmission code error detection system in inter-frame coding according to claim 1, wherein each of said summation circuits comprises a predetermined middle bit among all bits representing a sum value to be detected. A transmission code error detection method in inter-frame coding, characterized in that only bits are output and the multiplex transmission and comparison processing are performed.