JPH08340540A - Image frame coding of decoding method and device therefor - Google Patents

Abstract

PURPOSE: To encode an image frame in the order of display with a desired decoding frame rate by coding a 1st frame obtained through classification of an input image independently and coding a predictive frame and a droppable frame based on a timewise preceding frame. CONSTITUTION: An input signal 1 is given to a frame memory (FLM) 2 an when the frame (FL) is classified as a 1st frame (1PL) or a predictive frame (PFL), it is fed to a reference FL memory(FLM) 3. Furthermore, a signal in the FL2 is given to a block sampling circuit 4, in which the signal is classified into blocks. When the FL is the IFL, it is given to a DCT circuit 7 and when the FL is the PFL or DFL, it is given to a motion detection circuit 5. The PFL and DFL are prediction-coded by using the preceding IFL or PFL from the FLM 3 as the FL reference and the result is fed to the DCT circuit 7. On the other hand, the IFL from the circuit 4 is given to the DCT circuit 7. The DCT circuit 7 converts each block into a coefficient of the DCT area and the converted coefficient is outputted as a coded bit stream via a quantization device 8 and a coder 9.

Description

Detailed Description of the Invention

[0001]

FIELD OF THE INVENTION The present invention relates to a method and apparatus for image frame encoding or decoding which can be used in a video data compression system having a decoder capable of decoding below a maximum frame rate. .

[0002]

2. Description of the Related Art Flexible frame rate is a function required to realize a low-cost or software-only decoder. These decoders do not have the processing power necessary to decode and display the video sequence at maximum frame rate. In addition, the flexible frame rate can realize a trick mode such as fast forward.

ITU-T Recommendation H.261 (ITU-T H.261, "Co
dec for Audiovisual Service atnx384 kbit / s ", 1988
In coding schemes that use only prediction modes, such as (see), they must be decoded because all frames are needed for prediction of the next frame.

ISO / IEC MPEG (ISO / IEC 1117
2-2, "Information Technology-Generic Coding of Mov
ing Pictures and Associated Audio up to 1.5MBits /
s ", Nov. 1992) Bidirectional predictive frames can be eliminated in coding schemes that use bidirectional predictive modes such as the standard. However, bidirectional predictive frames require an additional frame buffer. , There is also an inherent delay.

[0005]

However, the conventional method as described above has a problem that the predicted frame must be decoded in the case of ITU-T Recommendation H.261.

In addition, ISO / IEC MPEG requires an additional frame buffer for bidirectional prediction,
In addition, there is a problem that an inherent delay occurs.

Therefore, the present invention (1) has a flexible decoding frame rate, (2) reduces delay (display order is the same as decoding order), and (3) 1
It is an object to provide a method and an apparatus for image frame encoding or decoding, which requires a memory for only one frame.

[0008]

The aforementioned problems can be solved by changing the reference frame for prediction. The present invention uses a new type of predictive frame, hereinafter referred to as droppable frame. The droppable frame uses a previous intra-coded frame or a prediction frame as a reference frame. This allows the frame to be reduced without affecting the prediction of subsequent frames.

Even in the bidirectional prediction frame of the MPEG standard,
Although predicted frames can be removed in a similar way, forward and backward predictions result in the frames being displayed in a different display order than the coding order. Also, this prediction requires memory for one or more frames in addition to encoding and decoding delays. Droppable frames do not have this problem as all predictions are only forward.

The present invention according to claim 1 classifies an input image frame sequence into an image group consisting of a first frame, a prediction frame and a droppable frame, and separates the first frame from any other frames. And predictively coding the predicted frame based on the first frame or the predicted frame, which is temporally previous, and the droppable frame is temporally previous, and And a step of performing predictive coding based on the first frame or the predicted frame.

The output by the above-described image frame encoding method determines whether each frame type included in the image group belongs to the first frame, the predicted frame, or the droppable frame. Frame type parameter information indicating, and a frame parameter indicating the number of droppable frames included between the first frame and the prediction frame immediately following the first frame or between two consecutive prediction frames. It may be a bitstream including the number information.

According to the present invention of claim 3, when the frame in the image frame sequence coded by the image frame coding method of claim 1 is the first frame, the first frame is Independently of any of the frames of the image frame, and if the frame in the encoded image frame sequence is the prediction frame, the prediction frame is temporally previous to the first frame or the prediction frame. Predictive decoding based on a frame, if the frame in the encoded image frame sequence is the droppable frame, selecting whether to decode the droppable frame, and by that step The droppable frame selected for decoding is the first frame or the temporally previous frame. A decoding method of image frames comprising the steps of predicting decoding based on the predicted frame.

It should be noted that the above-mentioned selecting step is Rd._i 
Is the i-th decoded frame rate, R_f Is the maximum frame
Rate, factor (M + 1)_i Is number i in (M + 1)
Eye coefficient, and M is the first frame and subsequent
Between the predicted frame and the predicted frame
The drop between the prediction frame and
If it is the number of possible frames, Rd_i= R_f/ {F
actor (M + 1) _i} Based on the
It is also possible to select a removable frame.

According to a fifth aspect of the present invention, a plurality of frames in the image frame sequence encoded by the image frame encoding method according to the first aspect are the droppable frames in a predetermined order. And means for encoding to generate an encoded frame sequence, and to generate a frame sequence for locally decoding the first frame and the predicted frame in the encoded frame sequence. Image frame coding, characterized in that it comprises means for decoding and means for storing each frame in the locally decoded frame sequence in order to predictively code the later frame. It is a device.

According to a sixth aspect of the present invention, the first frame and the prediction frame in the image frame sequence encoded by the image frame encoding method according to the first aspect are:
Means for decoding to produce a locally decoded frame sequence, Rd _i for the i th decoded frame rate, R _f for the maximum frame rate, factor (M +
1) _i is the i-th coefficient of (M + 1), and M is the droppable between the first frame and the subsequent prediction frame or between the prediction frame and the subsequent prediction frame If it is the number of frames, based on Rd _i = R _f / {factor (M + 1) _i }, means for selecting the droppable frame to be decoded, and decoding the droppable frame selected by the means Decoding an image frame, characterized in that it includes means for encoding and each frame in the locally decoded frame sequence is stored for predictive coding of a temporally subsequent frame. It is a device.

According to the present invention of claim 7, the step of classifying an input image frame sequence into an image group consisting of a first frame, a prediction frame, a droppable frame and a dependent frame, and the first frame is classified into any other type. Encoding independently of a frame, predictively encoding the predicted frame based on the first frame or the predicted frame, which is temporally previous, and temporally preceding the droppable frame. Of the first frame or the predicted frame, and predictively coding the dependent frame based on the droppable frame or the dependent frame that is temporally previous. A method for encoding an image frame, which is characterized by being provided.

It should be noted that the output by the above-described image frame encoding method is such that the type of each frame included in the image group is one of the first frame, the predicted frame, the droppable frame, and the dependent frame. Information of a frame type parameter indicating whether or not the droppable frame is included between the first frame and the prediction frame immediately following the first frame or between two consecutive prediction frames, and The bit stream may include information on the number of frame parameters indicating the number of dependent frames.

According to a ninth aspect of the present invention, when the frame in the image frame sequence encoded by the image frame encoding method according to the seventh aspect is the first frame, the first frame is Independently of any of the frames of the image frame, and if the frame in the encoded image frame sequence is the prediction frame, the prediction frame is temporally previous to the first frame or the prediction frame. Predicting and decoding based on a frame; if the frame in the encoded image frame sequence is the droppable frame, selecting whether to decode the droppable frame; and The droppable frame selected for decoding by the Or a step of performing predictive decoding based on the predicted frame; a step of selecting whether or not to decode the dependent frame when the frame in the encoded image frame sequence is the dependent frame; And a predictive decoding of a dependent frame selected for decoding according to the dropperable frame or the dependent frame that is temporally previous. is there.

It should be noted that the above-mentioned step of selecting is based on Rd _i
Is the i-th decoded frame rate, R _f is the maximum frame rate, factor (M + 1) _i is the i-th coefficient of (M + 1), and M is between the first frame and the following predicted frame, or Decoding based on Rd _i = R _f / {factor (M + 1) _i } if the number of droppable frames and the number of dependent frames between the prediction frame and the subsequent prediction frame The droppable frame or the subordinate frame may be selected.

According to an eleventh aspect of the present invention, a plurality of frames in the image frame sequence encoded by the image frame encoding method according to the seventh aspect are arranged in a predetermined order, and the plurality of frames are the droppable frame and the droppable frame. Means for encoding to generate an encoded frame sequence, which is a dependent frame, and locally decoding the first frame, the predicted frame, and the droppable frame in the encoded frame sequence , A means for generating a decoded frame sequence, and means for storing each frame in the locally decoded frame sequence for predictive encoding of a temporally subsequent frame, Image frame decoding device.

According to a twelfth aspect of the present invention, the first frame and the prediction frame in the image frame sequence encoded by the image frame encoding method according to the seventh aspect are locally decoded to obtain a decoded frame. Means for generating columns, Rd _i is the i-th decoded frame rate, R _f is the maximum frame rate, and factor (M + 1) _i is (M
+1) i-th coefficient, and M is the droppable frame and the dependent frame that are between the first frame and the subsequent prediction frame or between the prediction frame and the subsequent prediction frame. Rd _i = R _f / {factor (M +
1) _i }, means for selecting the droppable frame and the dependent frame to be decoded, and a second frame for locally decoding the droppable frame in the encoded image frame sequence. Means for decoding to generate a sequence, means for decoding the dependent frames in the encoded image frame sequence,
Means for storing each frame in the locally decoded frame sequence for predictive decoding of a temporally later frame; and each frame in the locally decoded second frame sequence. An image frame decoding apparatus, comprising: means for storing a dependent frame temporally later for predictive decoding.

[0022]

The present invention is used in predictive coding schemes based on the H.261 and MPEG standards. In motion compensated predictive coding, a previously decoded frame is used as a reference frame for the next frame to be decoded. The present invention uses a new reference frame and a prediction frame so that the image frame decoding apparatus can selectively decode the same coded bitstream at different frame rates.

The image frame sequence is classified into an image group consisting of a first frame, a prediction frame and a droppable frame. The first frame of the image group is coded without reference to any other frame. The prediction frame is predictively coded by using the first frame or prediction frame that is temporally previous as a reference. The plurality of droppable frames separate the first and subsequent predicted frames. Also, these frames are predictively coded by using the temporally previous first or predicted frame as a reference. These frames are not used as a reference for any future prediction. Therefore, these frames can be deleted without losing any importance to the prediction of future frames. Therefore, the name is called droppable frame.

A second method, which is a further extension of the above invention, allows for large temporal intervals between the reference frame and some droppable frames. As an aid in reducing this interval, some of the droppable frames are predictively coded by using the droppable frame in time as a reference. At this time, these frames are subordinate to the reference droppable frame. If the reference frame is reduced, then the dependent frames will also be reduced. Hereinafter, these frames are referred to as dependent frames.

[0025]

Embodiments of the present invention will be described below with reference to the drawings.

FIG. 1 is a block diagram of an embodiment according to the image frame encoding apparatus of the present invention. The present invention relates only to a reference frame selection and predictive coding device. Therefore,
The coding method is used only for the description of the embodiment.

The coding functional block diagram shows a coding device using DCT with motion detection and compensation techniques.

In this embodiment, the input image frame is
The image group is composed of a first frame, a prediction frame, and a droppable frame. The first frame, hereafter referred to as the I frame, is coded independently of all other frames. Predictive frames, hereinafter referred to as P-frames, are predictively coded by using the preceding I-frame or P-frame as the frame reference. Also, hereinafter, the droppable frame, referred to as the D frame, is predictively coded by using the preceding I frame or P frame as the frame reference. However, D-frames are not themselves used to predict other D-frames. The classification of frames forms a part of the present invention and will be described in detail in a later example.

The input signal 1 which is the input frame image sequence is placed in the frame memory 2. If this frame is classified as an I-frame or a P-frame, it is sent via line 14 to the reference frame memory 3 for use as a reference frame in the motion detection of the next frame which is predictively coded. To be This signal is also sent to the block sampling circuit 4 through the line 13. With this block sampling circuit 4,
This signal is divided into blocks where the pixel data does not spatially overlap for further processing.

If the frame is classified as an I-frame, the sampled block is sent on line 16 to the DCT circuit 7. If the frame is classified as a P-frame or a D-frame, the sampled block is sent on line 17 to the motion detection circuit 5. This motion detection circuit 5 uses the reference frame memory 3 to obtain the motion vector that gives the best match in prediction.
And use the information from the current block 17. The motion vector and the local reconstructed frame are sent to the motion compensation circuit 6 through the line 19 and the line 20, respectively. The difference image is formed by subtracting the motion compensated decoded frame 21 from the current input block 15. Therefore, this signal is transmitted through line 22 to DC
It is sent to the T circuit 7.

In the DCT circuit 7, each block is converted into a DCT domain coefficient. The transform coefficients are sent via line 23 to a quantizer 8 where they are quantized. The quantized coefficients are then sent via line 24 to the run length and variable length encoder 9. Here, the coefficients are run-length encoded and variable length encoded to form the output bitstream 25.

Also, if the current frame is classified as a P frame or an I frame, the quantized coefficients are sent to the dequantizer 10 via line 26. The output of the inverse quantizer 8 is then sent via line 27 to the inverse DCT circuit 11. If the current frame is an I frame,
The reconstructed frame is placed in the local decoded frame memory 12 via line 28. The current frame is P
In the case of a frame, the output 29 of the inverse DCT circuit 11 is
It is added to the output 21 of the motion compensation circuit 6 to form the reconstructed frame 30. The reconstructed frame 30 is then placed in the local decoded frame memory 12 for motion compensation of subsequent frames.

FIG. 2 is a block diagram of an embodiment according to the image frame encoding apparatus of the present invention. The input bitstream 31 is sent to a variable length run length decoder 32. The block and side information are extracted by the variable length / run length decoder 32. The output of the variable length run length decoder 32 is sent to the inverse quantizer 33 via line 37. If the current frame is a P frame or a D frame, the side information containing the motion vector is sent via line 45 to the motion compensation circuit 36. Motion compensation circuit 36 uses this information and the information in local decoded frame memory 35 to form motion compensated signal 44.

Next, the output of the dequantizer 33 is the line 3
8 to the inverse DCT circuit 34. The coefficient is
It is converted by the inverse DCT circuit 34 and returned to the pixel value. If the current frame is an I frame, the output of the inverse DCT circuit 34 is sent on line 39 and stored in the frame memory 42. If the current frame is a P frame or a D frame, the output of the inverse DCT circuit 34 is sent on line 40 and added to the output of the motion compensation circuit 36. It also forms the reconstructed frame 41 stored in the frame memory 42. If the current frame is an I or P frame, the frame memory 42 is transferred to the local decoding frame memory 35 via line 46. Therefore, this information is used for motion compensation of subsequent frames.

3 and 4 show H.2 as an example of the prior art.
61 shows the frame classification and arrangement for the 61 coding standard and the MPEG coding standard. The arrow indicates the reference frame for the predictive coding frame. FIG. 3 is a diagram showing a typical frame sequence of the H.261 coding method. The first frame is coded without reference to any other frame. Therefore, a plurality of P frames follows. In this example, none of the P frames can be reduced because each one is needed for prediction of the next P frame. Therefore, it is not possible to reduce the frame unless complete decoding is performed.

FIG. 4 is a diagram showing a typical frame sequence of the MPEG standard. I-frames and P-frames are currently separated by multiple bi-predictive frames, hereinafter referred to as B-frames. These B-frames can be removed because they will not be used in future predictions. However, due to the bidirectional prediction, the display and transmission order of frames is different. This causes longer delays in the encoding and decoding process.

FIG. 5 is a diagram showing frame classification which constitutes the present invention. Flexible decoding frame rate is
This is achieved by allowing some frames that are not needed at the lower temporal resolution to be reduced. As shown in FIG. 4, the I frame and the P frame are separated by a plurality of frames. However,
These frames are predicted in the forward direction only, so
It does not require different transmission and display order as in the past.
Therefore, it has a delay similar to the example of FIG.
Having a frame that can be reduced so that a flexible frame rate can be achieved.

FIG. 6 is a diagram showing frame classification for the second method of the present invention. Flexible decoding at different frame rates is achieved as before. The only addition in this method is the inclusion of dependent frames, hereafter referred to as d-frames. The d-frame functions similarly to the D-frame, except that it uses the latest decoded frame for motion compensated prediction. This is the second
This means that a locally decoded frame memory of may be needed for the decoder. The purpose of d-frames is to reduce the interval between temporal predictions.

FIG. 7 is a diagram showing examples of different frame rates that can be achieved by decoding a selected number of frames from the same sequence. Different frame rates can be achieved at the decoder by selectively decoding only a certain number of frames from the same coded bitstream. In FIG. 7, the maximum frame rate of 1
/ 2, 1/4 and 1/8 are shown.

FIG. 8 is similar to FIG. 7 but shows an example of using the second method of the present invention. The method of selecting the decoded frame and the achievable frame rate is similar to the first method.

In theory, any number of droppable frames can exist between I and P frames. However, as the number increases, the accuracy of motion estimation decreases. In general, the fraction of the possible decoding frame rate is given by (Equation 1).

[0042]

[Equation 1]

Here, the meaning of each variable is as follows.

Rd _i : i-th decoded frame rate R _f : maximum frame rate factor (M + 1) _i : i-th coefficient of (M + 1) M: D-frame and d-frame between I-frame and P-frame if any Is the number of.

Table 1 is a table showing the relationship between the number of D frames and the possible frame rates that can be decoded from the same bitstream using the present invention.

[0046]

[Table 1]

[0047]

As is clear from the above, the present invention has the effect that a very flexible method of decoding an image frame sequence at a desired frame rate can be achieved.

Furthermore, since only forward prediction is used, this scheme has the advantage that the delay can be kept low in the encoding and decoding processes and thus it can be applied to communication services.

[Brief description of drawings]

FIG. 1 is a functional block diagram of an embodiment of an image frame encoding device according to the present invention.

FIG. 2 is a functional block diagram of an embodiment according to the image frame decoding apparatus of the present invention.

FIG. 3 is a diagram showing a prediction mode used in ITU-T Recommendation H.261.

FIG. 4 is a diagram showing a prediction mode used in the ISO-IEC / JTC MPEG standard.

FIG. 5 is a diagram showing a prediction mode used in an embodiment of the present invention.

FIG. 6 is a diagram showing a prediction mode used in an embodiment of the present invention.

FIG. 7 is a diagram showing an example of flexible frame rate decoding in the decoding apparatus of this embodiment.

FIG. 8 shows an example in which an embodiment of the second method of the invention is used.

[Explanation of symbols]

 1 Input Signal 2 Frame Memory 3 Reference Frame Memory 4 Block Sampling Circuit 5 Motion Detection Circuit 6 Motion Compensation Circuit 7 DCT Circuit 8 Quantizer 9 Run Length / Variable Length Encoder 10 Inverse Quantizer 11 Inverse DCT Circuit 12 Local Decoding Frame memory 25 output bit stream 31 input bit stream 32 variable length / run length decoder 33 inverse quantizer 34 inverse DCT circuit 35 local decoding frame memory 36 motion compensation circuit 42 frame memory 43 output reconstructed image frame

Claims (12)

[Claims] 1. A step of classifying an input image frame sequence into an image group consisting of a first frame, a prediction frame and a droppable frame, and a step of encoding the first frame independently of any other frames. And predictively encoding the prediction frame based on the first frame or the prediction frame that is temporally previous, and the droppable frame is temporally previous to the first frame.
Frame or the step of performing predictive coding based on the predicted frame. 2. The output of the image frame encoding method indicates which of the first frame, the predicted frame, and the droppable frame the type of each frame included in the image group belongs to. Frame type parameter information and the number of frame parameters indicating the number of droppable frames included between the first frame and the prediction frame immediately following the first frame or between two consecutive prediction frames. 2. The method of encoding an image frame according to claim 1, wherein the image frame is a bit stream including the information of. 3. When the frame in the image frame sequence encoded by the image frame encoding method according to claim 1 is the first frame, the first frame is independent of any other frame. And when the frame in the encoded image frame sequence is the predicted frame, the predicted frame is predicted and decoded based on the first frame or the predicted frame that is temporally previous. If the frame in the encoded image frame sequence is the droppable frame, the step of selecting whether to decode the droppable frame, and the step of selecting to decode the droppable frame The droppable frame, the first frame or the prediction frame that is temporally previous. And a step of performing predictive decoding based on the above. 4. Rd _i is the i-th decoded frame rate, R _f is the maximum frame rate, and factor (M +
1) _i is the i-th coefficient of (M + 1), and M is the droppable frame between the first frame and the subsequent prediction frame or between the prediction frame and the subsequent prediction frame. If it is a number, the selecting step is Rd _i = R _f / {fact
4. The image frame decoding method according to claim 3, wherein the droppable frame to be decoded is selected based on or (M + 1) _i }. 5. An encoded frame in which each frame in the image frame sequence encoded by the image frame encoding method according to claim 1 is a droppable frame in a predetermined order. Means for encoding to generate a sequence, and means for decoding the first frame and the predicted frame in the encoded frame sequence to generate a locally decoded frame sequence, An image frame encoding device, comprising means for storing each frame in the locally decoded frame sequence for predictive encoding of a frame that is temporally later. 6. The first image in the image frame sequence encoded by the image frame encoding method according to claim 1.
Means for decoding the frame and the predicted frame so as to generate a locally decoded frame sequence, Rd _i is the i-th decoded frame rate, R _f is the maximum frame rate, and factor (M + 1) _i Is the i-th coefficient of (M + 1), and M is the number of droppable frames between the first frame and the subsequent prediction frame or between the prediction frame and the subsequent prediction frame. Then Rd _i = R _f
Means for selecting the droppable frame to be decoded based on / {factor (M + 1) _i }, means for decoding the droppable frame selected by the means, and the locally decoded frame sequence And a means for storing each frame in order to predictively encode a frame temporally later. 7. A step of classifying an input image frame sequence into an image group consisting of a first frame, a prediction frame, a droppable frame and a dependent frame, and the first frame is coded independently of any other frame. Encoding the prediction frame based on the first frame or the prediction frame that is temporally previous, and the droppable frame is temporally previous to the first frame.
Of the frame or the predicted frame, and the step of predictively coding the dependent frame based on the droppable frame or the dependent frame that is previous in time. Image frame encoding method. 8. The output according to the encoding method of the image frame is such that the type of each frame included in the image group is any one of the first frame, the prediction frame, the droppable frame, and the dependent frame. Information of a frame type parameter indicating whether the frame belongs, the droppable frame and the subordinate frame included between the first frame and the prediction frame immediately following the first frame or between two consecutive prediction frames. 8. The image frame encoding method according to claim 7, wherein the bit stream includes a frame stream number information indicating the number of frames. 9. If the frame in the image frame sequence encoded by the image frame encoding method according to claim 7 is the first frame, the first frame is independent of any other frame. And when the frame in the encoded image frame sequence is the predicted frame, the predicted frame is predicted and decoded based on the first frame or the predicted frame that is temporally previous. And a frame in the encoded image frame sequence,
In the case of the droppable frame, the step of selecting whether or not to decode the droppable frame, and the droppable frame selected to be decoded by the step are the first frames that are previous in time. Or a step of performing predictive decoding based on the predicted frame; a step of selecting whether or not to decode the dependent frame when the frame in the encoded image frame sequence is the dependent frame; And a predictive decoding of a dependent frame selected to be decoded by the method based on the droppable frame or the dependent frame that is previous in time. 10. Rd _i is the i-th decoded frame rate, R _f is the maximum frame rate, and factor (M +
1) _i is the i-th coefficient of (M + 1), and M is the droppable between the first frame and the subsequent prediction frame or between the prediction frame and the subsequent prediction frame If it is the number of frames and the dependent frames, the selecting step selects the droppable frame or the dependent frame to be decoded based on Rd _i = R _f / {factor (M + 1) _i }. The image frame decoding method according to claim 9, wherein 11. A code for each frame in the image frame sequence encoded by the image frame encoding method according to claim 7, wherein a plurality of frames are the droppable frame and the dependent frame in a predetermined order. Means for encoding so as to generate an encoded frame sequence, and locally decoding the first frame, the prediction frame and the droppable frame in the encoded frame sequence to generate a decoded frame sequence. An image frame decoding apparatus comprising: a generating unit; and a unit that stores each frame in the locally decoded frame sequence so as to predictively code a temporally subsequent frame. . 12. Means for locally decoding the first frame and the predicted frame in the image frame sequence encoded by the image frame encoding method according to claim 7 to generate a decoded frame sequence. , Rd _i is the i-th decoded frame rate, R _f is the maximum frame rate, factor (M + 1) _i is the (M + 1) -th i-th coefficient, and M is the first frame followed by the predicted frame. , Or the number of dependent frames between the predicted frame and the subsequent predicted frame, based on Rd _i = R _f / {factor (M + 1) _i }, Means for selecting the droppable frame and the dependent frame to be decoded, and the drop in the encoded image frame sequence. Means for decoding a pable frame to generate a second frame sequence that is locally decoded; means for decoding the dependent frames in the encoded image frame sequence; Means for storing each frame in the decoded frame sequence for predictive decoding of a temporally later frame; and for each frame in the locally decoded second frame sequence, And a means for storing the subsequent dependent frame for predictive decoding.