JPH04280190A - Picture decoding device - Google Patents

Abstract

PURPOSE:To make the sequence of a picture frame outputted from a decoder coincident with a frame displaying the sequence when a picture signal applying movement compensation inter-frame prediction coding is decoded. CONSTITUTION:A selector 13 is thrown to the position (g) with a selector control signal 111 outputted from a prediction mode control section 12 to use a picture signal stored in a frame memory 4 as the output of the device in order to the prediction signal when a picture frame subject to in-frame coding and a picture frame applied with movement compensation prediction from one preceding decoded frame are decoded. Moreover, when a picture frame implementing both forward and backward movement compensation prediction is decoded, the selector control signal 111 is used to thrown the selector 13 to the position of (f) and the picture signal during decoding is used as the output of the device.

Description

[Detailed description of the invention]

0001

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an image decoding device for reproducing a highly efficiently encoded image signal.

0002

2. Description of the Related Art In order to reduce the data amount of a transmitted image signal, high-efficiency image encoding processing is performed. In the present invention, a forward motion compensated interframe predictive encoding method is used. The target is a system in which there is a mixture of frames for prediction, frames for backward motion-compensated inter-frame prediction, or frames for forward-backward motion-compensated inter-frame prediction. Motion compensation prediction, as described in Japanese Patent Publication No. 63-20075, predicts the image signal of the current input frame from the image signal of the frame that has already been encoded while compensating for the motion of each part. This is a technology that reduces the amount of information contained in the prediction error signal. In motion compensation, forward motion compensation is used when the frame used for prediction is a frame temporally past the current input frame, and backward compensation is used when the frame used for prediction is a frame temporally future than the current input frame. The case where prediction is performed from both the forward and future frames is called forward-backward bidirectional motion compensation.

FIG. 4 shows, as described above, a frame for which motion compensation prediction is performed from other frames and an intra-frame encoded frame for which encoding is performed using only the current input frame without using other frames. This shows how it is done at intervals. In the figure, 1 to 11 are frame numbers. in,
I, P, and B each represent a prediction method, where I is an intraframe coded frame, P is a forward motion compensation predicted frame, and B
is a forward-backward motion-compensated prediction frame. Image frame no. 1 is an intra-frame encoded frame, which performs encoding independently without using other frames. Next, image frame No. From the encoding result of No. 1, image frame No. 1 is extracted using forward motion compensation prediction. 4 is encoded. And image frame no. 1 and no. From the encoding result of No. 4, image frame No. 2.No. In step 3, encoding is performed using bidirectional motion compensation prediction. Furthermore, image frame No. Using the encoding result of No. 4, image frame No. 7 is encoded using forward motion compensated prediction. And image frame no. 4 and N
o. 7, image frame No. 7 is used. 5, No. 6 is encoded. Similarly, No. 10, No. 8, No.
9 are encoded in sequence. Image frames are input to the decoding device according to the order of encoded frames as described above, so in the example of FIG. 4, the order of image frames input to the decoding device is 1, 4. , 2, 3, 7
, 5, 6, 10, 8, 9, and so on.

[0004] In actual encoding, intra-frame encoding is performed in motion compensation block units (one frame divided by a predetermined number of pixels), but forward motion compensation prediction In the corresponding frame, intraframe coding and forward motion compensation predictive coding are used while being switched as appropriate for each block of motion compensation. Also,
In a frame compatible with bidirectional motion compensation prediction, intraframe coding, forward, backward, and bidirectional encoding are switched and used. This is done to improve coding efficiency, and this relationship is shown in FIG. 6 as a frame prediction mode and a block prediction mode.

FIG. 3 shows, for example, Nikkei Electronics 199
0.10.15 (No.511) P124-P129
FIG. 7 is a block diagram of the conventional image decoding device shown in FIG. 6 of the second special feature entitled "Single high-efficiency image encoding method." In FIG. 3, reference numeral 1 denotes a variable length decoding unit, which inputs an image encoded signal from a transmission path and encodes a prediction error signal into a signal 101.
(hereinafter signal 101), intraframe encoding/
Forward motion compensation predictive coding/bidirectional motion compensation predictive coding identification signal 102 (hereinafter referred to as signal 102), intraframe coding/forward motion compensation predictive coding for each motion compensation block,
An identification signal 103 (hereinafter referred to as signal 103) for backward motion compensation predictive encoding/bidirectional motion compensation predictive encoding and a motion vector 104 (hereinafter referred to as signal 104) for motion compensation are output. 2 is an inverse quantization unit which inputs a signal 101 and outputs a DCT (discrete cosine transform) coefficient 105 (hereinafter referred to as signal 105). 3 is an inverse DCT unit which inputs the signal 105 and outputs a prediction error signal 106 (hereinafter referred to as signal 106). A frame memory A 4 stores signals from selectors B 11 in order to perform interframe prediction. A motion compensation prediction unit A 5 reads out the signal stored in the frame memory A 4 and outputs a predicted signal 107c (hereinafter referred to as signal 107c). A frame memory B 6 stores the signal from the frame memory A 4 in order to perform interframe prediction. 7 is a motion compensation prediction unit B which reads out the signal stored in frame B of 6 and outputs a predicted signal 107e (hereinafter referred to as signal 107e). 8 is a motion-compensated forward-backward bidirectional prediction unit;
The signal stored in frame memory A of No. 4 and frame memory B of No. 6 is read out and a predicted signal 107d (hereinafter referred to as signal 107d) is output. 9 is selector A, each contact c,
Signals 107c, 107d, and 107e are input to d and e, respectively, and their outputs are connected to the adder 10. The adder 10 converts the signal from the selector A of 9 and the inverse DCT section 3
A decoded image signal 108 (hereinafter referred to as signal 108) is created by adding the signals 106 from . 11 is selector B,
A signal 106 from the inverse DCT unit 3 is input to the contact a, and a signal 108 from the adder 10 is input to the contact b. The output of selector B of No. 11 becomes the output of this decoding device, and is also connected to Frame A of No. 4 at the same time. 12
is a prediction mode control unit which inputs signals 102 and 103, and depending on the contents, selector setting signal 109 (hereinafter signal 1).
09) and switches selector B of 11,
The selector A of 9 is switched by outputting a selector setting signal 110 (hereinafter referred to as signal 110). frame prediction mode,
Switching of each contact point of selector A of 9 and selector B of 11 for the block prediction mode is shown in FIG.

Next, the operation will be explained. In FIG. 3, the variable length decoding unit 1 receives the image frame number of FIG. When an image encoded signal corresponding to 1 is input, the variable length decoding unit 1 decodes it according to a predetermined procedure and outputs it as a fixed length code (signals 101, 102, 103, 104). Signal 101 is input to inverse quantization section 2 . In this example, a method called DCT+quantization is used on the transmitting side to encode the prediction error signal, so in decoding, the inverse quantization unit 2
The signal inversely quantified is subjected to inverse DCT with a DCT coefficient of 105.
The prediction error signal 106 is inputted to the prediction error signal 106 and subjected to inverse DCT.
is played as. Regarding DCT, see Hashimoto: "Image Coding Algorithm II-Transform Coding", Journal of the Society of Television Engineers, vol. 43, No. 10 (1989) pp1145
1152, this is a type of orthogonal transformation that is attracting attention as a high-efficiency encoding technique. Now, image frame No. Since 1 is an intra-frame encoded frame, all motion compensation prediction blocks are intra-frame encoded as shown in FIG. Therefore, the signal 109 always sets the 11 selector B to the a side. As a result, the prediction error signal reproduced as the signal 106 is output from the decoding device to the outside as a decoded image signal, and at the same time is written into the frame memory A of No. 4 in preparation for the decoding of the next frame. In this way, image frame no.
When the decoding of No. 1 is completed, the image frame No. 4 is stored in the frame memory A of No. 4. 1 decoded image signal is stored.

Next, image frame No. 4 is input to the variable length decoding section 1. Image frame no. Since frame 4 is a forward motion compensation predictive frame, intraframe encoding and forward motion compensation predictive encoding are performed on a block-by-block basis for motion compensation, as described above. First, image frame No. Consider the case where a motion compensation block of 4 is forward motion compensation predictive encoded. At this time, as shown in FIG.
0 sets the selector A of 9 to the c side, and the signal 109
Accordingly, selector B of 11 is set to the b side. The signal 101 output from the variable length decoding section 1 is processed by the inverse quantization section 2,
It is processed by the inverse DCT unit 3 and becomes the signal 106, which is sent to the adder 1.
It is input to 0. On the other hand, the motion compensation prediction unit A of No. 5 selects the image frame No. 4 stored in the frame memory A of No. 4 according to the signal 104. Necessary information is read out from the decoded image signal 1 and output as a predicted signal 107C. At the same time, in preparation for decoding the next frame, image frame No. The decoded image signal of 1 is transferred to frame B of 6. Since the selector A of 9 is set to the c side as described above, the signal 107c is sent to the adder 10 through the selector A of 9.
is added to signal 106 described above to produce signal 108
is given to selector B of 11. Since the selector B of No. 11 is set to the b side, the signal 108 is sent to the outside as an output of the decoding device, and is also written to the frame memory A of No. 4 in preparation for decoding the next frame.

Next, image frame No. Consider the case where a certain motion compensation block of 4 is intra-frame coded. At this time, the signal 109 sets the 11 selector B to the a side as shown in FIG. The process by which the signal 106 is decoded is the same as that for the forward motion compensation block, but the signal 106 is not added to the predicted signal and is sent out as a decoded image signal and simultaneously written into the frame memory A of No. 4. Also, the image frame No. 4 is transferred from frame memory A of 4 to frame memory B of 6. Transfer of the decoded image signal No. 1 is performed in exactly the same manner. In this way, image frame no. When the decoding of No. 4 is completed, the image frame No. 4 is stored in the frame memory A of No. 4. The decoded image signal No. 4 is stored in the frame memory B No. 6, and the decoded image signal No. 4 is stored in the frame memory B No. 6. One decoded image signal is stored respectively.

Image frame No. When the decoding of image frame No. 4 is completed, image frame No. 2 is input to the variable length decoding section 1. Image frame no. 2 is a bidirectional motion compensation prediction frame, the prediction mode is switched to intraframe coding, forward prediction, backward prediction, and bidirectional prediction for each motion compensation block as shown in FIG. Now, image frame N
o. Consider the case where a certain motion compensation block No. 2 is bidirectionally predicted. At this time, as shown in FIG. 6, selector A of 9 is d
On the side, 11 selectors B are set on the b side, respectively. Signal 106 is processed in the same manner as described above and provided to adder 10. On the other hand, the bidirectional prediction unit 8 detects the image frame No. according to the signal 104. At the same time, necessary signals are read out from frame memory B No. 6 in which the decoded image signal No. 1 is stored, and at the same time, the decoded image signal no. Necessary signals are also read out from frame memory A of No. 4 in which decoded image signals of No. 4 are stored, and a predicted signal 107d is generated using both of these signals. As a method of generating this predicted signal, for example, image frame No. 1 side prediction signal and image frame No. An arithmetic average of four side prediction signals is often used. signal 10
7d passes through selector A of 9, is added to signal 106 by adder 10, and is sent as an output of the decoding device. However, at this time, writing to frame memory A of No. 4 and transfer from frame memory A of No. 4 to frame memory B of No. 6 are not performed. In this way, during decoding of a bidirectional motion compensation predicted frame, frame memory A 4 and frame memory B 6 are not updated, regardless of the motion compensation block-by-block prediction mode.

Now, image frame No. Consider the case where a certain motion compensation block No. 2 is forward predicted. At this time, as shown in FIG. 6, selector A 9 is set to the e side, and selector B 11 is set to the b side. Then, image frame No. 6 in frame memory B. A predicted signal 107e is read from the decoded image signal of 1, and the adder 10 outputs a decoded image signal obtained by adding the signal 106 and the signal 107e. At this time as well, frame memory A of 4
and 6, frame memory B is not updated.

Next, image frame No. When a certain motion compensation block 2 is backward predicted, selector A 9 is set to the c side and selector B 11 is set to the b side, as shown in FIG. Then, image frame No. 4 in frame memory A. Signal 107 from the decoded image signal of 4
c is read out, and the adder 10 adds the signal 106 and the signal 1
The decoded image signal to which 07c has been added is sent out as an output. Also at this time, frame memory A of 4 and frame memory B of 6 are not updated.

Furthermore, image frame No. When a certain motion compensation block 2 is intra-frame encoded, as shown in FIG. 6, selector B 11 is set to the a side, so the prediction error signal 106 is sent out as output as is. . Also at this time, frame memory A of 4 and frame memory B of 6 are not updated.

Image frame No. When the decoding of image frame No. 2 is completed, image frame No. 2 is completed. 3 starts decoding. Image frame no. 3 is also a bidirectional motion compensation predicted frame, and the contents of frame memory A of 4 and frame memory B of 6 are also the same as image frame No. The state is exactly the same as when decoding started in step 2, so
Image frame no. The decoding of 3 is No. This is done in the same way as 2.

Next, image frame No. 7 is input. Image frame no. Image frame No. 7 is a forward motion compensation prediction frame. It is decoded in the same way as in step 4. In this process, image frame No. The reproduced decoded image signal of No. 7 is written to the frame memory A of No. 4, and the image frame No. 7 which was in the frame memory A of No. 4 is written. The decoded image signal No. 4 is transferred to the frame memory B No. 6.

As described above, each time an intra-frame encoded frame or a forward motion compensation predicted frame is input, decoding is continued while updating frame memory A 4 and frame memory B 6. Each input frame No. and the writing and reading states of frame memories A and B at that time, the states of selectors A and B, and the frame No. output from the decoding device. are summarized in Figure 5. Output frame No. in FIG. As you can see, the order is input frame No.
．． Similarly, it can be seen that the numbers are 1, 4, 2, 3, 7, and so on.

[0016]

[Problems to be Solved by the Invention] Since the conventional image decoding device is configured as described above, the order of the frames output from this device is the same as the order of the input frames, and the It's not in order. For this reason, in order to perform display, it is necessary to separately provide a plurality of frame memories, store them in one frame memory, and rearrange them, which poses a problem of high cost.

The present invention has been made to solve the above-mentioned problems, and aims to obtain an image decoding device in which the order of frames output from the decoding device matches the order of frames to be displayed. The purpose is

[0018]

[Means for Solving the Problems] The image decoding device according to the present invention includes a new selector provided at the output of the device, one input of the selector is connected to the output of the conventional device, and the other one is connected to the output of the conventional device.
Connect the input to the conventional frame memory B to one input,
One of these is selected and output from the image decoding device. As a selection criterion, when decoding an intra-frame encoded or forward motion compensation predicted frame, the input to frame memory B is selected, and when decoding a bidirectional motion compensation predicted frame, the conventional output is selected.

[0019]

[Operation] When decoding an intra-frame coded frame or a forward motion compensation predicted frame, the image decoding device according to the present invention outputs the contents of a frame memory provided in the device as part of the prediction unit, and performs bidirectional motion compensation. When decoding a predicted frame, the reproduced image signal being decoded is output, so the order of frames output from the decoding device matches the order of frames to be displayed.

[0020]

[Example] Example 1. Hereinafter, one embodiment of the present invention will be described based on the drawings. FIG. 1 is a block diagram of an image decoding device showing an embodiment of the present invention.
performs the same functions as those with the same reference numerals in the conventional image decoding device shown in FIG. The prediction mode control section 12 has the same function as the conventional example shown in FIG. ing. In addition, the selector C of 13 switches the output of this decoding device, and the contact f is connected to the output of the selector B.
The contact g is connected to the input of the frame memory B 6, and this switching is performed by the signal 111 output from the prediction mode control section 12.

Next, the operation will be explained. Image frame No. input to this image decoding device. The prediction mode is as shown in FIG. The operations of the units 1 to 12 corresponding to each input image frame in FIG. 1 are the same as in the conventional system. Further, FIG. 6 shows the operations of the 13 selectors C corresponding to the frame prediction mode and block prediction mode of each input frame.

In FIG. 1, image frame No. When the image encoded signal corresponding to No. 1 is input to the variable length decoding section No. 1, it is processed in the same manner as before to generate a prediction error signal 106. And image frame no. Since frame 1 is intra-frame encoded, selector B 11 is set to the a side and selector C 13 is set to the g side, as shown in FIG. Therefore, the signal 106 is written into the frame memory A of No. 4. Image frame no. Since frame 1 is the first frame input to this device, the transfer from frame memory A of 4 to frame memory B of 6 is not performed. As a result, no signal is output from the 13 selectors C set on the g side. In this way, image frame No. When the decoding of 1 is completed, frame memory A of 4 is stored as before.
Image frame No. One decoded image signal is stored.

Next, image frame No. 4 is input. Image frame no. Since 4 is a forward motion compensation predicted frame, the selector C of 13 is set to the g side as shown in FIG. Further, the forward motion compensation predictive frame is subjected to intra-frame encoding and forward motion compensation predictive encoding in units of motion compensation blocks, but in any case, No. The decoded image signal No. 4 passes through the selector B No. 11 and is written into the frame memory A No. 4 in preparation for decoding the next frame. At the same time, the No. 4 frame memory A that had been stored up until now. Image frame 1 is transferred to frame memory B 6, passes through selector C 13, and is sent out as an output of the decoding device. With the above operations, image frame No. When the decoding of No. 4 is completed, the image frame No. 6 is stored in the frame memory B of No. 6. The decoded image signal No. 1 is stored in the frame memory A No. 4, and the image frame No. 4 is stored in the frame memory A No. 4. 4
decoded image signals are stored.

Next, image frame No. 2 is input. Image frame no. Since frame 2 is a bidirectional motion compensation frame, selector C 13 is set to the f side as shown in FIG. Further, the bidirectional motion compensation predictive frame is subjected to intraframe encoding, forward motion compensation predictive encoding, backward motion compensation predictive encoding, and bidirectional motion compensation predictive encoding in units of motion compensation blocks, but in any case, No. The decoded image signal No. 2 passes through the selector B No. 11, and is sent out as an output of the decoding device via the f contact of the selector C No. 13. However, since the signal that has passed through selector B 11 is not written to frame memory A 4 as in the conventional example, both memories 4 and 6 are not updated.

Next, image frame No. 3 is input, but image frame No. Image frame No. 3 is a bidirectional motion compensation predicted frame. It is decoded in the same manner as 2 and sent as is as the output of the decoding device. Also, both memories are not updated, and the image frame No. 6 is stored in frame memory B. Image frame numbers 1 and 4 are stored in frame memories A. 4 remains stored.

Image frame No. Following image frame No. 3. 7 is input. Since this frame is a forward motion compensation predicted frame, selector C of 13 is set to the g side as shown in FIG. And image frame no.
Similarly to 4, the decoded image signal that has passed through selector B 11 is written to frame memory A 4 in preparation for decoding the next frame. At the same time, the No. 4 frame memory A that had been stored up until now. The 4th image frame is
The signal is transferred to the frame memory B of No. 6, passes through the selector C of No. 13, and is sent out as an output of the decoding device. In this way, image frame No. When the decoding of 7 is completed,
Image frame No. 6 is stored in frame memory B of No. 6. The decoded image signal of No. 4 is stored in the frame memory A of No. 4, and the image frame No. 4 is stored in the frame memory A of No. 4. 7 decoded image signals are stored. As described above, the order of image frames output from this decoding device so far is 1, 2, 3, 4.

Next, image frame No. 5, No. 6 is input next, but since this frame is a bidirectional motion compensation frame, it is image frame No. 6. 2.No. 3, it is sent out as the output of this device. After that, image frame no. When 10 is input, the contents of frame memory A of 4 and frame memory B of 6 are updated, and the image frame No. 1 stored in frame memory A of 4 is updated. 7 decoded image signals are output.

The above operating state is shown in FIG. In this way, when the prediction mode for each frame is intraframe coding or forward motion compensation predictive coding, the decoded image signal is transferred from frame memory A to frame memory B,
At the same time, the signal is used as the output of the decoding device. Further, when the prediction mode for each frame is bidirectional motion compensation predictive coding, the data is used as it is as an output of the decoding device without being transferred between frame memories. As a result, Figure 2
Output frame No. As shown in , the order of frames output from the decoding device is 1, 2, 3, 4, 5, 6, 7.
..., which matches the order in which they should be displayed.

[0029]

As described above, the image decoding device according to the present invention outputs a previously decoded image signal when decoding an intraframe coded frame and a forward motion compensation predicted frame, and outputs a previously decoded image signal, When decoding,
Since the device is configured to output the decoded image signal as is, the order of the frames output from this device matches the order of the frames to be displayed, making it unnecessary to use multiple expensive memories to rearrange the frames. This has the effect of eliminating

[Brief explanation of the drawing]

FIG. 1 is a configuration diagram of an image decoding device according to an embodiment of the present invention.

FIG. 2 is a diagram showing the operating state of an image decoding device according to an embodiment of the present invention.

FIG. 3 is a configuration diagram of a conventional image decoding device.

FIG. 4 is a diagram showing types of frames.

FIG. 5 is a diagram showing the operating state of a conventional image decoding device.

FIG. 6 is a diagram showing the operating state of each selector for frame prediction modes.

[Explanation of symbols]

1 Variable length decoding unit 2 Inverse quantization section 3 Inverse DCT section 4 Frame memory A 5 Motion compensation prediction unit A 6 Frame memory B 7 Motion compensation prediction unit B 8 Motion compensated bidirectional prediction unit 9 Selector A 10 Adder 11 Selector B 12 Prediction mode control unit 13 Selector C

Claims (1)

[Claims] 1. Means for generating a predicted signal from a frame memory storing previously decoded image signals, means for generating a decoded image signal by adding a decoded prediction error signal and the above prediction signal, the above prediction error signal and the above decoding In an image decoding device that decodes a motion-compensated interframe prediction signal and is equipped with a means for switching image signals, the image signal decoded in the past is not necessary to generate the prediction signal, or the image signal for one frame is required. When decoding, the image signal read from the frame memory is output, and when decoding an image frame that requires a plurality of previously decoded image signals to generate a predicted signal, the decoded image signal is output. image decoding device.