JP3507532B2 - Moving image encoding method and apparatus - Google Patents

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a moving picture coding technique for compressing and coding a moving picture signal, and more particularly to a moving picture coding method and apparatus suitable for an ATM network.

[0002]

2. Description of the Related Art Research and development of an ATM (asynchronous transfer mode) network in which multimedia information such as image, voice and data is transferred / exchanged by carrying fixed length packets called cells is under way. When transmitting information of a moving picture by an ATM network, a moving picture coding apparatus for compressing and coding a moving picture signal is used. As a compression coding method of a moving image signal, a prediction residual signal which is a difference between an input moving image signal and a predicted image signal obtained by motion compensation prediction is coded (interframe coding) and transmitted. A motion compensation adaptive predictive coding system is known.

By the way, in an ATM network, when an error occurs in a communication path or when cells are congested in the network, a phenomenon called cell loss occurs in which some of the transmitted cells are discarded in the exchange. There are various general measures against this cell loss, but when transmitting moving image information with the moving image encoding device as described above, the encoding device periodically performs intraframe encoding within the screen. There are a method of setting the pixel area to be performed and refreshing the screen, and a method of concealing by utilizing the correlation between frames.

The former refresh method is being studied in the standardization work of MPEG2, for example.
In SO-IEC / JTC1 / SC29 / WG11 MPEG / 225b Test Model 4,
Intra-frame coding and inter-frame coding can be switched in units of 16 pixels × 16 pixels called a macro block. Specifically, as shown in FIG.
In one frame composed of 0 macroblocks x 44 horizontal macroblocks, a region consisting of horizontal 2 rows of macroblocks (2 vertical macroblocks x 44 horizontal macroblocks) indicated by diagonal lines is intra-frame coded and Slide the coding area by 2 lines for each frame,
A refresh method called intra slice in which one cycle is one frame has been proposed.

In the intra slice method, as shown in FIG. 5, when interframe coding is performed in the area 1,
By limiting the search range of the motion vector not to start from the region 2 that has not been refreshed, even if an error occurs in a certain frame due to the occurrence of the above cell loss,
As shown in FIG. 4, at worst, it is guaranteed that the screen is completely refreshed by the intraframe coding in two cycles of the intra slice so that the effect of the error is removed and the refresh is performed.

On the other hand, the latter concealment method is very effective when the correlation between frames is high, but is not effective when the correlation between frames is low as in the case of a scene change.

Therefore, for example, when cell loss occurs in a region where the correlation between frames is low at the time of a scene change, the concealment method is not effective and the refresh method completely recovers from an error as shown in FIG. However, at worst, two cycles will take a long time.

Another problem with the cell loss in the scene change frame is that the rate of intra-frame coding (intra coding) of the scene change frame is high, so that the amount of codes generated by the moving picture coding apparatus is high. Is often twice or more than the inter-frame coded frame, and if the cell loss occurrence rate is the same, the cell loss occurrence probability of the frame immediately after the scene change is twice or more that of the inter-frame coded frame.

[0009]

As described above, the refresh method and the concealment method conventionally considered as a cell loss countermeasure when the moving picture coding apparatus is applied to the ATM network are between the frames at the scene change. For the cell loss occurring in the low correlation region, the former has a problem that it takes time to recover from an error, and the latter has a problem that effective cell loss countermeasures cannot be taken.

[0010] The present invention is to reduce the probability of cell loss in the intra-coding region generated due to low correlation between frames by a scene change, the moving picture coding method and can quickly recover from the error with respect to cell loss The purpose is to provide a device.

[0011]

To solve the above problems SUMMARY OF THE INVENTION The present invention is a method of encoding by the selected coding mode from any of the input video signal intra coding mode or an inter coding mode Yes, the input
The intra-coding mode is periodically displayed in the screen of the moving image signal.
Area to be set and the area to be moved for each screen.
And the motion in the inter coding mode.
The refresh operation has already been performed for the vector search range.
In the moving picture coding method for limiting the area to a limited area, it is determined whether or not a scene change has occurred for each frame of the input moving picture signal based on the selected state of the coding mode. when it is determined that there, and changes the period of the refresh operation to perform the refresh operation from the upper end of the frame immediately following the frame determined as the scene change has occurred. Here, the intra coding means the intra-frame coding or the intra-field coding, and the inter coding means the inter-frame coding or the inter-field coding. Further, the moving picture coding apparatus according to the present invention selects a coding means having a coding mode for intra-coding an input moving picture signal and a coding mode for inter-coding, and a coding mode of the coding means. And a scene change determination means for determining whether or not a scene change has occurred for each frame of the input moving image signal based on the mode selection state by the mode selection means. Is the intra code periodically in the frame of the input video signal.
Performs a refresh operation to move the screen each of the region by setting the space of the modes, when it is determined to be occurring a scene change by the scene change determination means, it is determined that the scene change has occurred and it is intended to change the period of the refresh operation to perform the refresh operation from the end of the frame immediately after the frame, before
The coding means operates in the inter coding mode.
The search range of the vector
It is characterized in that it is limited to a limited area .

[0012]

[0013]

[0014]

In the present invention, the scene change is determined by the selected state of the coding mode. When changing scenes,
Since the ratio of intra-coded areas is large, in a frame in which the intra-coded areas are equal to or greater than a predetermined ratio,
It can be determined that a scene change has occurred . When it is thus determined that a scene change has occurred , the refresh operation cycle is changed, and, for example, the scene change operation cycle is initialized, so that the scene change is started from the end of the frame. By doing this, even if a cell loss occurs in the scene change frame, that is, in the intra-coded area, by limiting the search range of the motion vector to the already refreshed area, the range in which the effect of the error due to the cell loss propagates is limited. At worst, one cycle is enough.

Thus, even if cell loss occurs in a scene change frame, it is possible to quickly recover from an error.

[0016]

[0017]

[0018]

[0019]

Embodiments of the present invention will be described below with reference to the drawings. (Embodiment 1) FIG. 1 is a block diagram showing the configuration of a moving picture coding apparatus according to an embodiment of the present invention. In the figure, an interlaced moving image signal, that is, a moving image signal in which one frame is composed of two fields is input to the input terminal 10. The input moving image signal is first combined into a frame by the field merge circuit 11 and then input to the encoding unit 12.

The moving image signal input to the encoding unit 12 is
First, the blocking circuit 13 divides into macroblocks. The input moving image signal divided into macroblocks is input to the subtracter 14, and the difference between the input moving image signal and the predicted image signal generated as described later is calculated to generate a prediction residual signal. Either one of the prediction residual signal and the input moving image signal from the blocking circuit 13 is selected by the mode selection switch 15 and is discrete cosine transformed by the DCT (discrete cosine transform) circuit 16. D
The DCT coefficient data obtained by the CT circuit 16 is quantized by the quantization circuit 17. The signal quantized by the quantizer 17 is branched into two, and one of them is guided to the variable length coding circuit 27 provided on the output side of the coding unit 12.

On the other hand, the other of the signals quantized by the quantization circuit 17 and branched into two is the inverse quantization circuit 18 and the IDCT.
The (inverse discrete cosine transform) circuit 19 allows the quantization circuit 1
7 and the processing of the DCT circuit 16 which is the reverse of the above processing, the adder 20 adds the predicted image signal input via the switch 23 to generate a locally decoded signal. This locally decoded signal is stored in the frame memory 2
It is input to the motion compensation circuit 22 via 1 and a predicted image signal is generated.

Here, as shown in FIG. 5, the motion compensating circuit 22 prohibits the motion compensation before the refreshing for the already refreshed region 1, and generates the predicted image signal.

The coding control circuit 26 controls the coding unit 12 based on the coding information from the coding unit 12 and the information of the buffer amount B in the output buffer 30 which will be described later. For example, the motion vector and the coding mode information are sent to the variable length coding circuit 28. The additional information coded by the variable length coding circuit 28 and the quantized DCT coefficient data coded by the variable length coding circuit 27 are
After being multiplexed by the multiplexing circuit 29, it is smoothed to a constant output rate by the output buffer 30 and further cell-ized circuit 31.
Then, the cells are assembled into a cell of a predetermined ATM cell format to which header information is added, and the cell is sent to the transmission line.

The encoding unit 12 is further provided with a mode selection circuit 24 and a scene change determination circuit 25.
The mode selection circuit 24 selects, for each macroblock, a macroblock for intraframe coding and a macroblock for interframe coding based on the prediction information P from the motion compensation circuit 22. When the intra-frame coding is performed, the mode selection switch information M is A and the switch information S is A. Also, when performing inter-frame coding,
The mode selection switch information M is B, and the switch information S is B.

The mode selection switch 15 is set to A when performing intra-frame coding according to the mode selection switch information M.
To the B side when performing interframe coding. The switch 23 is switched according to the switch information S to the A side when performing intra-frame coding, and to the B side when performing inter-frame coding, similarly to the mode selection switch 15.

The scene change judging circuit 25 determines, based on the mode selection information SM from the mode selecting circuit 24, that the number of macroblocks to be intra-coded (hereinafter referred to as intra macroblocks) is a predetermined value or more. In some cases, it is determined that the frame is a scene change, and the scene change determination information SD is turned on. The scene change determination information SD is input to the mode selection circuit 24.

FIG. 2 is a flow chart showing the operation of the mode selection circuit 24 and the scene change determination circuit 25 each time one frame of the input moving image signal is input in this embodiment. In the figure, the part surrounded by the broken line shows the operation of the scene change determination circuit 25, and the other parts show the operation of the mode selection circuit 24.

First, the mode selection circuit 24 initializes the counter value N for counting the refresh cycle to 0 (S101). The scene change determination circuit 25 initializes a counter value IM representing the number of intra macroblocks based on the prediction information P in the frame to 0 and sets the scene change determination information SD to OFF (S102).

In loop 1 and loop 2, mode determination and scene change determination of macroblocks in a frame are performed. Here, i and j represent the addresses of the horizontal and vertical macroblocks in the frame, respectively, and MB_Ro
w and MB_Column represent the numbers of macroblocks in the vertical and horizontal directions in the frame, respectively.

The prediction information P from the motion compensation circuit 22 is IN
It is determined whether or not it is TRA, that is, whether or not intraframe coding is performed (S105), and the result of this determination is INTRA.
Then, B is assigned to the mode selection switch information M and B is assigned to the switch information S regardless of refresh (S1).
09). At this time, the scene change determination circuit 25 displays the IM
Is incremented by 1 and the value is determined (S110-S11
1) If the value of IM is larger than the threshold value ε, the scene change determination information SD is turned on (S112).

On the other hand, the prediction information P is IN in S105.
If it is determined not to be TRA, that is, basically when interframe coding is performed, the remainder obtained by multiplying the counter value N by the number of slides (twice here) divided by the number of macroblocks MB_Column on the side of one frame , That is (N
X2) 1 if the value of modMB_Column matches the value of i or the counter value N is multiplied by the number of slides.
It is determined whether or not the remainder obtained by adding the value is divided by MB_Column, that is, (N × 2 + 1) modMB_Column matches i (S106). When they match, it is determined that the intraframe coding is forcibly performed, and A is assigned to the mode selection switch information M and A is assigned to the switch information S (S107). If they do not match, it is determined that interframe coding is to be performed, and B is assigned to the mode selection information M and B to the switch information S (S108).

After the above processing is completed for one frame of the input moving image signal, the scene change determination information SD becomes O.
It is determined whether or not it is N (S113). If it is ON, the value of N is initialized to 0 (S114). If not, N is incremented by 1 and the value of (N + 1) mod (MB-Row / 2) (N +
N is replaced by 1 (a remainder obtained by dividing the value by one refresh period) (S115).

FIG. 3 shows an example of how the refresh cycle is changed in this embodiment. In the case of intra slice, it is possible to limit the error range to a maximum of 1 cycle by changing the refresh cycle and refreshing from the top edge of the frame immediately after the scene change. It is clear from 3.

When the present embodiment is compared with FIG. 5 showing the conventional case, the presence of a scene change frame shortens the refresh cycle during that period and increases the ratio of intra macroblocks. However, even if cell loss occurs in a scene change frame for which the conventional concealment method is not effective, the range of error propagation is one cycle at the maximum, and the effectiveness of the present invention is clear.

In the above embodiment, the case where the encoding is performed in units of frames has been described, but it goes without saying that the present invention can be applied to the encoding in units of fields. In short, a moving picture coding apparatus capable of selecting a mode of intra-coding (intra-frame coding or intra-field coding) and an inter-coding (inter-frame coding or inter-field coding) of an input moving picture signal. If so, the present invention can be applied.

(Embodiment 2) FIG. 6 is a block diagram showing the structure of a moving picture coding apparatus according to another embodiment of the present invention. The same parts as those in FIG. 1 are designated by the same reference numerals, and description will be made centering on differences from the first embodiment.

In FIG. 6, the interlaced moving image signal input to the input terminal 10 is split into two, one is input to the frame / field switch 61 and the other is input to the scene change detection circuit 62. flame/
The field switch 61 is the scene change detection circuit 6
It is controlled by the scene change detection information SC output from No. 2, and it is determined based on this scene change detection information SC whether the input moving image signal is input in frame units or field units. That is, if the scene change detection information SC is ON, the frame / field switch 61 is connected to the A side, and the input moving image signal is input to the encoding unit 12 in field units. If the scene change detection information SC is OFF, the frame / field switch 61 is connected to the B side, and the input moving image signal is input to the field merge circuit 11 and is combined into a frame before being input to the encoding unit 12. .

The coding unit 12 performs coding in units of macroblocks, and selectively performs coding in units of frames and coding in units of fields according to the format of the input moving image signal. The basic configuration and operation of the encoding unit 12 are the same as in FIG.

The encoding control unit 26 is based on the encoding information from the encoding unit 12, the scene change detection information SC from the scene change detection circuit 62 and the buffer amount information B from the output buffer 30. And sends the additional information to the variable length coding circuit 28.

The operation of this embodiment will be described with reference to FIG. In the case of normal encoding, (a) → (b), (c) →
As shown in (d), (d) → (e), the even field and the odd field are merged by the field merging circuit 11 to form a frame, and interframe prediction is performed to perform coding in interframe units. On the other hand, for the frame (c) in which a scene change is detected by the scene change detection circuit 62, inter-field prediction is performed for each field as shown in (b) → (c), and coding is performed in field units. As a result, inter-field prediction is performed from the even field for the odd field in the scene change frame, and thus intra macroblocks are unlikely to occur.

The effect of this embodiment will be described with reference to FIG. The broken line in FIG. 8 shows the change in the generated code amount at the scene change in the conventional encoding only on a frame-by-frame basis. Since the scene change frame has a low correlation with the previous frame, many intra macroblocks are generated, and the code amount is more than double the usual amount. For this reason, the probability of cell loss occurring in a scene change frame is more than doubled.

On the other hand, in this embodiment, since the field unit is coded at the time of the scene change, only the even field has a low correlation with the previous frame, and a lot of intra macroblocks are generated and the code amount is increased. Grows only in even fields. Therefore, as shown by the solid line in FIG. 8, the number of intra macroblocks generated at the time of scene change is reduced as compared with the conventional case. For this reason, the probability of cell loss occurring in intra macroblocks due to the low correlation between frames due to a scene change is reduced, and the image quality is improved.

[0043]

As described above, according to the present invention, the probability of occurrence of cell loss in the intra-coded area caused by the relative correlation between frames due to the scene change is reduced, and the error is promptly corrected when the cell loss occurs. A moving image coding method suitable for an ATM network that can recover and secure good image quality
And a device can be provided.

[Brief description of drawings]

FIG. 1 is a block diagram showing a configuration of a moving picture coding apparatus according to an embodiment of the present invention.

FIG. 2 is a flowchart for explaining the operation of a mode selection circuit and a scene change determination circuit in FIG.

FIG. 3 is a schematic diagram for explaining a refresh operation in the same embodiment.

FIG. 4 is a schematic diagram for explaining a conventional refresh operation.

FIG. 5 is a schematic diagram for explaining limitation of a motion vector search range.

FIG. 6 is a block diagram showing the configuration of a moving picture coding apparatus according to another embodiment of the present invention.

FIG. 7 is a schematic diagram for explaining the operation of the embodiment.

FIG. 8 is a characteristic diagram showing a temporal change in the generated code amount for explaining the effect of the embodiment.

[Explanation of symbols]

DESCRIPTION OF SYMBOLS 10 ... Moving image signal input terminal 11 ... Field merge circuit 12 ... Encoding part 13 ... Blocking circuit 14 ... Subtractor 15 ... Mode selection switch 16 ... DCT circuit 17 ... Quantization circuit 18 ... Inverse quantization circuit 19 ... IDCT circuit 20 ... Adder 21 ... Frame memory 22 ... Motion compensation circuit 23 ... Switch 24 ... Mode selection circuit 25 ... Scene change determination circuit 27 ... Variable length coding circuit 28 ... Variable length coding circuit 29 ... Multiplexing circuit 30 ... Output buffer 31 ... Cellizing circuit 61 ... Frame / field switch 62 ... Scene change detection circuit

(72) Inventor Hideyuki Ueno 1 Komukai Toshiba-cho, Kouki-ku, Kawasaki-shi, Kanagawa Toshiba Research & Development Center Co., Ltd. (72) Inventor Kimio Seki 1 Komukai-shi, Cho, Kawasaki-shi, Kanagawa Address In Toshiba Research and Development Center (72) Inventor Yoshihiro Kikuchi No. 1 Komukai Toshiba Town, Kouki-ku, Kawasaki City, Kanagawa Inside Toshiba Research and Development Center (72) Inventor Takashi Ida Komukai Toshiba, Saiwai-ku, Kawasaki City, Kanagawa Prefecture 1-cho, Toshiba Research & Development Center (72) Inventor Susumu Kamiwa 1 Komukai, Komukai-ku, Kawasaki-shi, Kanagawa Toshiba-cho Research & Development Center, Ltd. (72) Masahiro Oshikiri Small, Kawasaki-shi, Kanagawa Muko Toshiba Town No. 1 Toshiba Research & Development Center (72) Inventor Noboru Yamaguchi No. 1 Komukai Toshiba Town No. 1 Kobayashi Toshiba Research & Development Center, Kawasaki City, Kanagawa Prefecture (72) Inventor Amata Kawasaki, Kanagawa Prefecture Komukai Toshiba-cho 1-ku, Toshiba Research Co., Ltd. Inside the center (72) Inventor Yoko Miyakawa 1 Komukai Toshiba Town, Saiwai-ku, Kawasaki City, Kanagawa Prefecture Toshiba Research and Development Center (72) Inventor Takehiko Kagoshima 1 Komukai Toshiba Town, Kawasaki City, Kanagawa Prefecture Stock Corporate Toshiba Research and Development Center (56) Reference JP-A-3-190482 (JP, A) JP-A-3-101490 (JP, A) JP-A-2-266783 (JP, A) JP-A-5-161130 (JP, A) JP-A-3-139083 (JP, A) JP-A 61-131986 (JP, A) JP-A 64-27379 (JP, A) (58) Fields investigated (Int. Cl. ⁷⁾ , DB name) H04N 7/24-7/68

Claims (2)

(57) [Claims] 1. A method for encoding an input moving image signal according to an encoding mode selected from either an intra encoding mode or an inter encoding mode , wherein said input
The intra-coding mode is periodically displayed in the screen of the moving image signal.
Area to be set and the area to be moved for each screen.
And the motion in the inter coding mode.
The refresh operation has already been performed for the vector search range.
In the moving picture coding method for limiting the area to a limited area, it is determined whether or not a scene change has occurred for each frame of the input moving picture signal based on the selected state of the coding mode. when it is determined that there, and changes the period of the refresh operation to perform a refresh operation immediately over <br/> end of the frame of the frame where it is determined that the scene change has occurred Video coding method. 2. An encoding unit which has an intra encoding mode and an inter encoding mode and which encodes an input moving image signal, a mode selecting unit which selects an encoding mode of the encoding unit, and the mode selecting unit. A scene change determination unit that determines whether or not a scene change has occurred for each frame of the input moving image signal based on a mode selection state by the unit, and the mode selection unit is a frame of the input moving image signal. set the periodic regions of the intra coding mode performs a refresh operation of moving the region for each screen within, when it is determined to be occurring a scene change by the scene change determination unit, the scene refresh to perform a refresh operation immediately after the frame change is determined to have occurred from the upper end of the frame It is intended to change the cycle of work, the encoding means, in the inter coding mode
The refresh operation has already been performed within the motion vector search range.
A moving picture coding apparatus, characterized in that it is limited to the created area .