JPH01243790A - Forecasting and encoding device between motion compensating frames - Google Patents

Abstract

(57) [Summary] This bulletin contains application data before electronic filing, so abstract data is not recorded.

Description

[Detailed description of the invention] [Industrial application field] The present invention relates to high efficiency encoding of television signals.

In particular, the present invention relates to a motion-compensated inter-frame predictive coding device that can reduce the device scale and improve prediction efficiency in a motion-compensated inter-frame prediction method.

[Conventional technology]

Since a television (hereinafter abbreviated as TV) signal has a wide frequency band, converting it into a digital signal and transmitting it usually requires a high transmission speed. Many schemes are known to reduce this speed.

One of the typical methods is 1-motion compensated interframe prediction, which consists of 1) sampling the 'I'V signal into pixels, and 2) using multiple pixels (for example, a square of 8 pixels each in the vertical and horizontal directions). 64 pixels) to 1
3) Find the block that is most similar to the above block while shifting its position little by little from the transmitted previous TV frame, and 4) calculate the above ``positional shift amount'' and ``difference between the two blocks. It is something that transmits ``°.

The prediction error signal after the above motion-compensated interframe prediction is variable-length coded and then stored in a buffer memory (hereinafter referred to as IQP).
IFO: First in, first out
st In First 0nt) and is sent to the transmission line at a constant speed. Here, as the absolute value of the above prediction error signal increases, many codewords with long codeword lengths are generated. Therefore, the amount of signals written to the FIFO increases, so the amount of accumulation in the FIFO increases, and eventually It overflows.

In order to prevent this overflow, the process is actually temporarily interrupted and the above writing is prohibited.

Conversely, if the absolute value of the prediction error signal is small and the code word length is short, the amount of storage in the 1-' IFO decreases and eventually becomes empty, making it impossible to send the signal to the transmission path. . To prevent this, a normal useless signal (dummy signal) is written into the FIFO.

The following two methods are known for searching for the most similar block while gradually shifting the position in the previous TV frame (hereinafter referred to as motion vector detection method).

1) Full search: Compare with all blocks at shifted positions within the range to be searched, and select the block with the smallest difference.

2) Multi-stage search: Comparing several predetermined types of blocks at different positions, it is assumed that an optimal block is found around the most similar block among them, and the surroundings are searched in more detail.

[Problem to be solved by the invention]

Each of the above full search/multistage search motion vector detection methods has the following problems. That is, the exhaustive search can select the most similar block within the range, but the problem is that the processing time increases because the number of blocks to be compared increases. In multi-stage search, the relationship is the opposite. Furthermore, in order to compare a large number of blocks within a short processing time, it is necessary to increase the circuit scale.

[Means to solve the problem]

The above problem is to increase the number of comparisons to improve the accuracy of detecting the optimal motion vector when processing time is available (for example, when the FIFO is about to overflow), and to improve the accuracy of detecting the optimal motion vector in other cases (for example, when the FIFO is about to overflow). This is solved by reducing the number of comparisons and improving processing speed.

[Effect]

When processing time is available, an optimal block can be detected by increasing the number of comparisons, so the codeword length can be shortened. Therefore, encoding efficiency can be improved. In this case, processing time will increase.

Since there is plenty of processing time, the number of comparisons can be increased without increasing the circuit scale.

Furthermore, when there is no processing time, the number of comparisons is reduced, which speeds up the processing, and there is no need to transmit wasteful dummy signals, so overall encoding efficiency can be improved. Therefore, in either case, high encoding efficiency can be achieved without increasing the circuit scale.

〔Example〕

Hereinafter, the present invention will be explained in detail using the drawings.

In order to facilitate the subsequent explanation, the block configuration of a conventionally known motion-compensated interframe predictive coding apparatus for TV signals will be briefly described using FIG.

A TV signal captured by a TV camera 1 is scanned in the scanning line direction, and becomes a pixel signal converted into a re-digital signal by an analog/digital converter 2. Furthermore, the scanning line/
The block scan conversion circuit 3 collects the pixels into blocks, which are stored in the -time storage memory 4.

A digital signal read out from temporary storage memory 4;
The signal read from the frame memory 5 that stores the signal of the previous 1 V frame that has been transmitted is the signal that is read out from the motion compensation circuit 6.
are compared and the similarity is calculated. According to the result, the motion compensation circuit 6 changes the delay amount of the delay circuit 7 to change the position in the previous TV frame to be compared. and,
The position with the highest degree of similarity is taken as the amount of movement.

Then, pixels to be encoded are preliminarily determined 81q using the pixel signal of the previous TV frame at the position whose motion amount has been compensated, and the interframe prediction error after motion compensation is calculated by the subtraction circuit 8.

The above prediction error is quantized by the quantizer 10 either as is or after orthogonally transformed by the orthogonal transform circuit 9,
The variable length encoding circuit 11 assigns a code word with a length corresponding to the frequency of occurrence.

Generally, it is known that when the prediction error is small, the codeword length becomes short, and when it is large, the codeword length becomes long. The signal encoded by the variable length encoding circuit 11 is
PIFO 12 for smoothing the transmission speed along with the above movement amount
will be written to. The encoded signal is read out from the FTFO 12 at a constant speed.

The signal is sent to the transmission line 13. Here, when the storage amount decreases and approaches empty, the PIFO 12 outputs an identification signal to the variable length encoding circuit 11 to send out a dummy signal.

On the other hand, when the accumulated amount increases and it is about to overflow, another identification signal is output to interrupt the encoding. This prevents malfunctions caused by the FIFO 12 becoming empty or overflowing.

At the same time, the above-mentioned quantized signal or a signal obtained by inverse orthogonal transformation of the quantized signal by the inverse orthogonal transform circuit 14 and the motion compensation prediction signal are added together by the adder circuit 15, and then the frame is added to predict the next TV screen. It is written into memory 5.

The overall configuration of the motion-compensated interframe coding device for 'rv signals has been briefly described above.Next, the general configuration method of the motion compensation circuit 6 that is particularly relevant to the present invention will be explained in detail using FIG. .

The motion compensation circuit 6 initializes the counter 21 when it finishes processing the motion compensation interframe prediction i+Il of the immediately preceding block. The output signal of the counter 21 is given to a read-only memory 22 (ROM:).

A shift amount (hereinafter referred to as "movement amount") for shifting the read position from the previous 1v frame in accordance with the counter value of 1 and a signal for identifying whether or not to continue the comparison are read out from the ROM 22.

The amount of motion is applied to the delay circuit 7, the amount of delay is controlled, and the pixel signal of the previous 1'' frame at the position where the amount of motion has been compensated for is read out.

The pixel signal of the previous TV frame and the temporary storage memory 4
A subtraction circuit 25 calculates a difference between the pixel signals stored in the pixel signal, an absolute value circuit 26 converts the difference into an absolute value, and an accumulation adder 27 calculates the similarity between two blocks in one block as a whole. As a result, the comparison circuit 28 compares the output of the storage adder 27 with the similarity in the previous most similar motion amount stored in the memory 29, and finds that the similarity is even higher than before. In this case, the degree of similarity is written into the memory 29, and at the same time, the amount of movement is written into the memory 30.

When the comparison for one amount of motion is completed, the counter 2] is incremented by 1, and an identification signal indicating whether to continue the comparison with the next amount of motion is read from the ROM 22. From then on, this process continues until the ROM 22 outputs an identification signal indicating that the comparison has ended.

Measure the amount of movement.

Then, the optimal amount of motion stored in the memory 30 is read out, and the signal of the previous TV frame at the position where the amount of motion has been compensated by controlling the amount of delay of the delay circuit 7 is read out, and the pixel to be encoded is predicted.

The conventional motion compensation circuit has been described above, and the method of varying the number of comparisons for motion amount search according to the present invention will be described with reference to FIG.

In the present invention, a total of three ROMs are prepared for identifying the amount of deviation and whether or not to continue the comparison. That is, the ROM 23 outputs an end signal only when the value of the counter 21 becomes large so as to perform multiple comparisons, and the ROM 24 outputs an end signal when the value of the counter 21 is small so as to perform a small number of comparisons. The ROM 22 outputs a signal to end the comparison of one block at an intermediate number of comparisons.

When the storage amount in the FIFO 12 increases to exceed the threshold value THI and is about to overflow, a signal indicating this is given to the switches 41 and 42 to select the ROM 231 and increase the number of comparisons. The accumulated amount of reverse FIFO 12 is the threshold T
When it becomes less than H2 and is about to become empty, similarly, that effect is given to the switches 41 and 42 to select the ROM 24 and reduce the number of comparisons. In other cases, the ROM 22 is selected and the normal number of comparisons is performed.

It is clear that one of the objects of the present invention is achieved through the above operations. Note that the following is also included in the present invention.

1) In the above embodiment, the number of comparisons for measuring the amount of motion was switched to three different numbers, but two or more arbitrary number of comparisons may be switched.

2) In the above embodiment, the FIFO storage amount was used as a means to measure the margin of processing time, but other means may also be used. For example, when the transmission speed becomes faster by changing the , since the processing time margin generally decreases,
The number of comparisons may be changed by means such as interlocking with a transmission speed changeover switch.

3) In the above embodiment, the case where the amount of motion is measured by a method called "block matching" was explained. is also valid. In other words, in the motion gradient method,
The rough amount of motion is measured from the difference between the corresponding ZTV frames and the gradient within the TV frame of the signal to be encoded, and then the rough amount of motion is compensated for.
'I' A more accurate amount of motion is detected from the difference between the V frames and the above gradient.

Generally, the greater the number of repetitions, the more accurate the measurement of the amount of motion becomes, but since the number of times of processing increases, the size of one circuit becomes larger. Therefore, according to the present invention, the number of repetitions described above may be increased only when there is sufficient processing time.

4) In addition to switching the number of comparisons as described above, the same effect can be obtained by simplifying the processing for one comparison (for example, comparing the difference between TV frames after subsampling).

〔Effect of the invention〕

As explained above, according to the present invention, the accuracy of measuring the amount of motion can be improved without increasing the scale of the apparatus, so that the encoding efficiency can be increased and the quality of one image can be improved. Therefore, the practical effect is great.

[Brief explanation of the drawing]

FIG. 1 is a block diagram of a circuit for varying the number of comparisons for motion amount search according to an embodiment of the present invention, FIG. 2 is a block diagram of a motion compensated interframe predictive coding device, and FIG. FIG. 2 is a block diagram of a motion compensation circuit. DESCRIPTION OF SYMBOLS 1...TV camera, 2...Analog/digital converter, 3...Scanning line/block scan conversion circuit, 4...
・-Time memory, 5...Frame memory, 6...
Motion compensation circuit, 7... Delay circuit, 8... Subtraction circuit,
9...Orthogonal transform circuit, 10...Quantizer, 11...
・Variable length encoding circuit, 12...FIFo, 13...
Transmission line, 14... Inverse orthogonal transform circuit, 15... Addition circuit, 21... Counter. 22.23,24...Motion amount generation ROM, 25...
Subtraction circuit, 26...Absolute value circuit, 27...Volume adder, 28...Comparison circuit, 29.30...Memory, 3
1゜1fIl Figure

Claims (1)

[Claims] 1. Means for measuring the amount of motion by comparing the TV signal to be encoded with the signal in the transmitted reference television frame, and the signal to be encoded by the signal in the reference television frame after compensating for the amount of motion. A motion compensation frame comprising means for predicting a television signal and obtaining a prediction error as a difference thereof, means for allocating the prediction error to a codeword, and means for sending the codeword to a transmission path through a smoothing buffer memory. The inter-predictive coding device includes means for detecting a margin of processing time based on the storage amount of the smoothing buffer memory and transmission speed, and means for switching the measurement accuracy of the amount of motion according to the margin. 1. A motion compensated interframe predictive coding device comprising: