JPH11215504A - Dynamic image coder - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a dynamic image coder that reduces a code quantity, depending on a magnitude of a motion vector for improving the coding efficiency. SOLUTION: A coding means 100 is provided with a motion compensation means 44 that generates a motion vector from each image block, based on an input image 10 and a predicted image 10A that is coed in advance to compensate a motion of the predicted image 10A. The coder 100 uses the motion compensation means 10 to encode the input image 10. In the process of coding, a code-length designation means 50 designates a code length of a motion vector in a prescribed image unit as a unit with respect to the coding means 100. Furthermore, a code length update means 60 updates the code length designated by the code length designation means 50 based on the magnitude of the motion vector generated in a current image unit, by the motion compensation means 44 when the coding means 100 encodes the input image with respect to a succeeding image unit.

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 apparatus for coding a moving picture with motion compensation.

[0002]

2. Description of the Related Art Conventionally, there has been known a moving picture coding technique for reducing the code amount (information amount) of a moving picture by motion-compensating and predictively coding the moving picture. For example, Japanese Unexamined Patent Publication
No. 105299 discloses that one screen of an input image composed of a moving image is divided into predetermined image blocks, a plurality of candidate motion vectors are prepared for each image block, and encoding efficiency is most improved when performing inter-frame prediction. A moving image compression apparatus for selecting a motion vector to be executed is disclosed.

[0003] Specifically, the configuration is shown in FIG. In the figure, 430 is a controller, 431 is an image memory,
432 is a subtractor, and 433 is DCT (Discrete Cosine Tra
nsform) operation unit, 434 is a quantization unit, 435 is a variable length coding unit (VLC), 436 is a buffer, 438 is an inverse quantization unit, 439 is an inverse DCT operation unit (IDCT), 440 is an adder, 441 to 441 443 is a switch, and 444 and 445 are predictors. In addition, the above-described inverse quantization unit 43
8, inverse DCT operation section 439, adder 440, switch 4
41 to 443 and the predictors 444 and 445 constitute a motion-compensated inter-frame predictor (no code).

Hereinafter, four types of operations (first to fourth operations) of the moving picture compression apparatus shown in FIG. 6 will be described with reference to a flowchart shown in FIG. In the description of the first to fourth operations, each step of the flowchart shown in FIG. 7 is referred to, but steps not referred to in the description of each operation are deleted from the flowchart shown in FIG. 7 as far as the operation is concerned. Shall be.

First, a first operation of the moving picture compression apparatus will be described. A series of operations described below are performed by the controller 43.
It is performed under the control of 0. The image memory 431 stores external image data. The subtractor 432 outputs a difference between the image data read from the image memory 431 and a predicted image from a predictor 444 or 445 described later. The DCT calculation unit 433 performs orthogonal transform on the difference data from the subtractor 432 to convert the difference data into a frequency axis. Quantization unit 43
4 and the variable length coding unit 435
Are quantized and variable-length coded, and output to the outside as coded data via a buffer 436.

Here, the motion-compensated inter-frame prediction unit (without sign) performs an inter-frame prediction process for compression in the time axis direction under the control of the controller 430. That is, the data from the quantization unit 434 is
And the predictor 4 which is decoded by the inverse DCT
It is added to the predicted image from 43 or 444. Based on the predicted image, a plurality of candidate motion vectors are detected and prepared for each predetermined image block (partial image) obtained by dividing one screen (step S1).

Then, a motion vector which gives the best coding efficiency when actually coding is determined from a plurality of candidate motion vectors (step S5), and the already coded image (predicted image) is The motion is compensated by moving the image by the distance, and the inter-frame prediction is performed. In addition,
The switches 441 to 443 are switched by the controller 430 according to the image mode and the prediction mode.

Next, a second operation of the moving picture compression apparatus will be described. Inter-frame prediction is performed in the same manner as in the above-described first operation, and a plurality of candidate motion vectors are detected and prepared for each image block (step S1). A prediction error is detected for the plurality of detected motion vectors,
An error coefficient for this prediction error with respect to the minimum prediction error is calculated (step S2). Then, based on the calculated error coefficient, a motion vector that gives the best coding efficiency when actually coding is determined from a plurality of candidate motion vectors (step S5).

Next, a third operation of the moving picture compression apparatus will be described. When the prediction errors of the motion vectors of a plurality of candidates are detected in the same manner as in the above-described second operation (steps S1 to S1).
In step 2), the value of the motion vector to be actually encoded is calculated for the plurality of candidate motion vectors (step S3). Based on the calculated value of the motion vector, the code length at the time of encoding is calculated as a vector encoding coefficient illustrated in FIG. 8 (step S4). Then, based on the coding coefficients, a motion vector that gives the best coding efficiency when actually performing coding is determined (step S5).

Next, a fourth operation of the moving picture compression apparatus will be described. When the prediction errors of the motion vectors of the plurality of candidates are detected in the same manner as the above-described third operation (steps S1 to S1).
In step S2), based on the motion vectors of the plurality of candidates, the code length at the time of encoding is calculated as a code length vector encoding coefficient (step S4). Then, based on the coding coefficient, a motion vector that gives the best coding efficiency when actually performing coding is determined (step S).
5).

[0011]

The moving picture compression apparatus according to the prior art described above selects a motion vector that gives the best coding efficiency from among a plurality of motion vectors detected in advance and performs coding. The code length of the motion vector is set regardless of the size of the motion vector. For this reason, depending on the motion vector, the code length may become redundant, and the code amount cannot be effectively reduced.

More specifically, according to the above-mentioned prior art, a motion vector is determined based on the calculated coding coefficient. The coding coefficient is uniquely determined regardless of the size of the motion vector. The rules for coding motion vectors are fixed. Therefore, even in the case of a motion vector that provides the best coding efficiency, the code length is fixed, and there is a case where the motion vector is redundant.

The present invention has been made in view of such a problem, and provides a moving picture coding apparatus capable of reducing a coding amount according to the magnitude of a motion vector and improving coding efficiency. The task is to

[0014]

Means for Solving the Problems The present invention has the following arrangement to achieve the above object. That is, the present invention divides an input image into predetermined image blocks, generates a motion vector for each of the image blocks from the input image and a previously encoded prediction image, and generates a motion vector for the prediction image based on the motion vector. Encoding means for compensating motion, encoding the input image by performing motion compensation by the motion compensating means, and encoding the motion vector with a designated code length; and In units of a predetermined image unit consisting of: a code length designating unit that designates the code length of the motion vector to the encoding unit, and when the encoding unit encodes the input image for the next image unit, The code length specified by the code length specifying means is determined by the motion compensation means based on the magnitude of the motion vector generated in the current image unit. And it has a configuration in which a new code length updating means.

Further, the code length updating means of the present invention comprises:
If the magnitude of all motion vectors generated in the current image unit is expressed using a code length smaller than the code length specified by the code length designating means for the current image unit, the code length The code length specified by the specifying means is updated to a minimum code length sufficient to represent the magnitudes of all the motion vectors.

Further, the code length updating means of the present invention is arranged so that the magnitude of a predetermined percentage of motion vectors, which is the majority of motion vectors generated in the current image unit, is larger than that of the current image unit. When expressed using a code length smaller than the code length specified by the code length specifying means, the code length specified by the code length specifying means is
It has a configuration in which the code length is updated to a minimum code length sufficient to represent the magnitude of a predetermined ratio of motion vectors that is the majority of the motion vectors.

Still further, the code length updating means of the present invention expresses the magnitude of a motion vector generated in the current image unit by using the code length designated by the code length designating means, and converts the input image. The input image is represented by expressing a motion vector generated in the current image unit using a first code amount generated when encoding and a minimum code length sufficient to represent the size of the motion vector. And a second code amount generated when encoding is performed. When the increment of the second code amount with respect to the first code amount is equal to or smaller than a predetermined threshold, the code specified by the code length specifying means is determined. It has a configuration to update the length.

Further, the code length updating means of the present invention expresses the magnitude of a motion vector generated in the current image unit by using a code length designated by the code length designating means, and The input image is expressed by expressing a motion vector generated in the current image unit using a first code amount generated when encoding and a code length smaller than the code length specified by the code length specifying unit. A second code amount generated when encoding is performed, and a code length specified by the code length specifying means when an increment of the second code amount with respect to the first code amount is equal to or less than a predetermined threshold value. Is updated to a code length smaller than the code length specified by the code length specifying means.

Furthermore, the code length updating means of the present invention encodes the input image by the encoding means when an increment of the second code amount with respect to the first code amount exceeds a predetermined threshold. The code length is updated in a direction in which the code length specified by the code length specifying means is increased within a range in which the code amount generated when the code length does not exceed the predetermined threshold value.

Further, the present invention further comprises a code length input means for inputting a code length specified externally,
The code length updating unit selects a code length generated by the code length updating unit or a code length input to the code length input unit according to a predetermined condition, and sets a code length specified by the code length specifying unit. It has a configuration to update.

Further, the code length updating means of the present invention generates code lengths of a plurality of candidates within a motion search range allowed for the image block, and uses the code lengths of the plurality of candidates. A code length generated when each of the input images is encoded by expressing the motion vector is compared, and a code length that minimizes the code amount is selected from the code lengths of the plurality of candidates.

Further, the code length updating means of the present invention further comprises a code amount obtained by expressing the motion vector using the code length input to the code length input means, and a code length updating means for generating the code length. The code length obtained by expressing the motion vector using the calculated code length is weighted and compared, and the code length is selected based on the result of the comparison.

Further, the present invention has a configuration in which the image unit is composed of two or more frames, and the number of frames constituting the image unit is variable.

[0024]

Embodiments of the present invention will be described below with reference to the drawings. In the drawings, common elements are denoted by the same reference numerals, and redundant description will be omitted.

Embodiment 1 FIG. 1 shows a configuration of a video encoding device according to the first embodiment of the present invention. In FIG. 1, reference numeral 100 denotes an input image 10 composed of a moving image, which is divided into predetermined image blocks, and a motion vector is generated for each of the image blocks from the input image 10 and a prediction image 10A which has been encoded in advance. It has a motion compensation prediction unit 44 (motion compensation means) for compensating the motion of the prediction image 10A based on the vector. The motion compensation unit 44 performs motion compensation on the input image 10 and encodes the input image 10 with a designated code length. An encoding unit (encoding means) for encoding the motion vector.

Reference numeral 50 denotes a code length designation unit (code length designation means) for designating a code length of a motion vector to the encoding unit 100 in units of a predetermined image unit composed of the above image blocks. When the encoding unit 100 encodes the input image for the next image unit to be encoded, the motion compensation prediction unit 44 uses the code based on the magnitude of the motion vector generated in the image unit to be currently encoded. A code length updating unit (code length updating unit) for updating the code length specified by the length specifying unit 50.

Further, 31 is an image memory for storing a pre-encoded image, 32 is a difference calculator for calculating the difference between the input image 10 and the predicted image 10A from the image memory 31, and 33 is a difference calculator. An encoder that encodes the difference data from the encoder 32; a variable length encoder (VLC) that obtains encoded data by encoding the data from the encoder 33 in a variable length manner; A buffer for temporarily storing encoded data from the encoder 35; a decoder 38 for locally decoding data from the encoder 33;
Is the predicted image 10A from the image memory 31 and the decoder 38
And an adder for adding the data from.

Next, the operation of the moving picture coding apparatus according to the first embodiment will be described by taking as an example the case of performing inter-frame predictive coding. In the first embodiment, the above-described “image unit” is referred to as a “frame”, and encoding is performed for each image block obtained by dividing each frame into some predetermined image blocks.

First, from the image memory 31, the image data at the same position as the image block of the input image 10 to be currently encoded or the image data specified by the motion vector generated by the motion compensation prediction unit 44 is predicted. The image is read for each image block as the image 10A. Difference calculator 32
Is the predicted image 10A from the image memory 31 and the input image 1
The difference from 0 is calculated to obtain the distortion evaluation value 46 and the difference data.

Here, the motion compensation prediction unit 44 calculates the distortion evaluation value 4 within the motion vector search range specified in advance.
6 (for example, the sum of absolute differences in block units), an image closest to the input image 10 or image data that matches the input image 10 is extracted as a predicted image 10a from the image memory 31 to detect a motion vector.

The difference data from the difference calculator 32 when the distortion evaluation value 46 indicates the minimum is encoded by the encoder 33 and is variable-length encoded by the variable-length encoder 35. At this time, the motion vector detected by the motion compensation prediction unit 44 is encoded with a code length specified by a code length specification unit 50 described later. The data that has been variable-length coded by the variable-length coder 35 is temporarily stored in a buffer 36, and is coded on a transmission path (not shown) according to a transmission rate.
Output as 0.

The data encoded by the encoder 33 is locally decoded by a decoding unit 38,
The predicted image 10A from 1 is decoded and added by the adder 40 and written into the image memory 31. Thus, the image data in the image memory 31 is updated. As described above, the inter-frame prediction encoding is performed on the input image 10.

In the above-described series of operations of the inter-frame predictive coding operation, the code length specifying unit 50 sets the coding unit 100 in units of one frame (predetermined image unit).
Is specified to the variable-length encoder 35 constituting. This code length may be a part or the whole of the encoded code of the motion vector. The code length designated by the code length designation unit 50 is also given to the motion compensation prediction unit 44, and determines the search range of the motion vector in the motion compensation prediction unit 44.

Further, the code length updating unit 60 includes the encoding unit 1
00 encodes the input image 10 for the next frame, the motion compensation prediction unit 44 determines the code length designation unit 50 based on the magnitude of the motion vector generated in the current frame.
Update the code length specified by.

Hereinafter, the operations of the code length specifying unit 50 and the code length updating unit 60 will be described in detail by taking as an example a case where a code length of a part of the encoded code of the horizontal motion vector is specified (updated). .

First, the code length of a part of the encoded code of the motion vector is defined as psize, and the code length psize is represented by an integer R that can represent a search range of the motion vector. For example, as an example, the integer R is represented by the following equation (1). However, in Expression (1), the symbol “≪” represents a bit shift operator.

R = 1≪psize (1)

In this case, the integer R is set to “1” for psize [bit] by MSB (Most Significant Bit: most significant bit).
It is represented as a binary number bit-shifted to the side, and is equivalent to 2 ^{ps ize in} decimal.

When the code length psize is represented by an integer R, the unit of motion compensation is an M × N pixel block (M,
(N is a natural number), the number of horizontal pixels is “M”,
Highest value max and lowest value min of possible motion vector
Is represented by the following equation (2), and the range of possible motion vectors is represented by the following equation (3).

Max = (M × R) −1, min = {(− M) × R} (2)

Range = (2M × R) (3)

Here, as an example, if psize = 3 and M = 8, from the above equations (2) and (3), R = 1≪3 = 8, max = (8 ×
8) Obtain -1 = 63, min = ((-8) × 8) =-64, range = 128.

Next, an integer mccod satisfying -M≤mccode≤M
The value MV of the magnitude of the motion vector is represented by a combination of the integer value mccode and the integer value mcrec, using e and the integer value mcres represented by the code length of psize [bit] as parameters.

As an example, when the condition of R = 1 and mccode = 0 is satisfied, the value MV of the magnitude of the motion vector is expressed by the following equation (4), and the above condition of R = 1 and mccode = 0 is satisfied. If not satisfied, the value MV of the magnitude of the motion vector is represented by the following equation (5). However, in equation (5), abs (mccod
e) represents the absolute value of the integer value mccode, and when mccode <0,
MV = -MV.

MV = mccode (4)

MV = [{abs (mccode) -1} × R] + mcres + 1 (5)

Thus, the value M of the magnitude of the motion vector
When V is expressed, integer values mccode and mcres can be set from a quotient {abs (mccode) -1} obtained by dividing the value MV of the magnitude of the motion vector by the integer R and the remainder (mcres + 1).

As an example, the value M of the magnitude of the motion vector
When expressing “53” as V, the quotient and remainder obtained by dividing “53” by M (= 8) are “6” and “5”, respectively, so mccode = 6
+ 1 = 7, mcres = 5-1 = 4. When the integer values mccode and mcres are set in this way, MV = ((abs (7)
-1) × 8) + 4 + 1 = 53 is obtained.

As another example, when “40” is expressed as the value MV of the magnitude of the motion vector, the quotient and remainder obtained by dividing “40” by M (= 8) are “5” and “0”, respectively. It is. in this case,
Psize [bit] that gives the code length of mcres when the remainder is "0"
Becomes “0”, and the magnitude of the motion vector cannot be expressed.

Therefore, in this case, a part of the quotient is excessively converted, and the quotient and the remainder are respectively set to “4” and “8”.
= 5, mcres = 7 ("111" in binary notation). When the integer values mccode and mcres are set in this way, MV = ((abs (5) -1) × 8) + 7 + 1 = 40 is obtained from the above equation (5). From the above example, the value MV of the magnitude of the motion vector is an integer value
It can be seen that mccode and mcres are represented as parameters.

If the motion vector magnitude value MV is generated with a higher probability near “0”, the integer value mccode
The encoding efficiency is better when represented by a variable length code. Also, when the integer value mcres is represented by a variable length code,
Since the code length of the motion vector of a value before and after a multiple of the integer R is reversed, the encoding efficiency is better when the integer value mcres is represented by a fixed-length code.

By expressing the magnitude MV of the motion vector using the above equations (4) and (5), the search range of the motion vector is determined in units of an integer R derived from the code length psize. The numerical value mcres fills a gap in the search range that cannot be represented by the integer R. Therefore, the magnitude MV of the motion vector is represented by the integer values mccode and mcres, and the code length psize can be determined based on the magnitude of the motion vector.

Here, the description returns to FIG. The code length specifying unit 50 specifies the code length psize determined based on the size of the motion vector to the encoding unit 100 as described above. The code length psize affects the code length of the motion vector generated in the current frame. If the code length psize of the motion vector is large, the search range (vector distribution range) of the motion vector is wide, and if it is small, it is narrow. Become.

On the other hand, based on the magnitude of the motion vector generated in the current frame under the influence of the code length psize specified by the code length specifying means 50, the code length update unit 60 The code length ps specified by the code length specifying unit 50 so that the code amount is reduced.
Update ize. As described above, the code length of the motion vector is updated based on the magnitude of the motion vector generated in the current frame, and the code amount in the next frame is reduced.

In the above-described first embodiment, the value MV is used as the magnitude of the motion vector. However, the “distribution” of the magnitude of the motion vector generated in the current frame, or the “distribution” of the magnitude of the motion vector generated in the current frame. The difference may be used, and does not limit the content of the present invention.

Although the magnitude of the motion vector is represented by using the integer values mccode and mcres as parameters, the present invention is not limited to this representation method. Code length specification section 50
Should be expressed so as to change according to the specified code length psize. Further, although the sum of absolute differences is used as the distortion evaluation value 46, for example, the sum of squared differences may be used.

Further, in the first embodiment, the inter-frame predictive coding is performed. However, it is configured to operate as a typical moving picture coding apparatus, for example, as in intra-frame coding. Is also good. In this case, the input image 10
May be directly encoded by the encoder 33 according to the condition.

Embodiment 2 Next, a second embodiment of the present invention will be described with reference to FIG. In the second embodiment,
The code length updating unit 50 (code length updating unit) determines whether the magnitude of all motion vectors generated in the current frame (current image unit) is the code length specified by the code length specifying unit 50 for the current frame. When expressed using a code length smaller than psize, the code length psize specified by the code length specifying unit 50 is updated to the minimum code length qsize sufficient to express the sizes of all the motion vectors. Be composed.

For example, in the current frame, psiz
If e = 3 and M = 8, the magnitude (vector) of all motion vectors generated in the current frame is -32 ≦ vector ≦ 3
When 2 is satisfied, the minimum code length sufficient to represent the size of all motion vectors is “2”, and all motion vectors can be represented even with psize = 2.

Because, when psize = 2, the maximum value max of the motion vector is "31" and the minimum value min of the motion vector is "-32" from the above equations (1) and (2). Because. Therefore, the code length updating unit 60
Assuming that qsize is “2”, the code length psize specified by the code length specifying unit 50 is updated to the code length qsize.

In this case, since the integer value mcres can be represented by 2 bits, the code length per motion vector is reduced by 1 bit. Therefore, if the code length psize is set to “2” and the motion vector of the current frame is re-encoded, the amount of code generated in the current frame can be reduced.

However, in practice, there is no time margin for re-encoding with the current frame due to the limitation of the arithmetic processing speed of the device. Even if there is a time margin, it is difficult to re-encode the current frame in an environment where a buffer capacity enough to accumulate the encoded motion vector for one frame cannot be secured. Becomes Therefore, the code length update unit 60 sets the code length psize to “2” in the next frame.
Update to (qsize) and encode.

Here, in general, in a moving image, when there is no rapid change in motion, the motion of adjacent image frames tends to be similar, and the motion vector of the next frame is similar to the motion vector of the current frame. It is predicted that a motion vector will be determined. Therefore, even if the code length psize is updated to “2” (qsize) in the next frame and coding is performed, the code amount generated in the next frame can be effectively reduced.

It should be noted that it has an operation processing speed sufficient for re-encoding, has a buffer capacity enough to accumulate a motion vector encoded in frame units, and has an encoding output from the buffer to the transmission path. If the specified transmission rate can be maintained without data interruption, the code length may be updated in the current frame and re-encoded. In this case, the code amount can be suppressed lower than when the code length is updated in the next frame.

Embodiment 3 Next, a third embodiment of the present invention will be described with reference to FIG. In the above-described second embodiment, the code length is updated when “all” motion vectors generated in the current frame can be expressed using the code length qsize. In the third embodiment, the code length updating unit is used. Reference numeral 50 denotes a code length designating unit 5 that determines the magnitude of a predetermined percentage of motion vectors that are “majority” of motion vectors generated in the current frame with respect to the current image unit.
When 0 is expressed using a code length smaller than the specified code length psize, the code length psize specified by the code length specifying unit 50 is used to express the size of the predetermined ratio of the majority of the motion vectors. Minimum code length qsiz
It is configured to update to e.

In the following, the validity of updating the code length by focusing on the above-described “majority” of the predetermined proportion of motion vectors will be described. If the motion tendencies of adjacent moving images are similar, if all the motion vectors of the current frame can be represented based on the code length qsize, the majority of motion vectors in the next frame will be based on the code length qsize. It is expected that it can be expressed. Therefore, if the code length qsize is specified in the next frame, the generated code amount can be reduced with an extremely high probability. In the second embodiment, the code length is updated from this viewpoint.

However, even if “all” motion vectors can be represented based on the code length qsize in the current frame, “all” motion vectors can be represented based on the code length qsize in the next frame. Is not always the case. Also, when the “majority” motion vectors are vectors that can be represented based on the code length qsize in the current frame, the “all” motion vectors are vectors that can be represented based on the code length qsize in the next frame. Sometimes.

That is, if it can be predicted that the “majority” of motion vectors can be expressed based on the code length qsize in the next frame, it is not always necessary to pay attention to “all” of the motion vectors generated in the current frame. Instead, it suffices to focus on a predetermined percentage of motion vectors that are "majority." In the third embodiment, from this viewpoint, the code length updating unit 60 updates the code length by focusing on a predetermined percentage of motion vectors that are “majority” in the current frame.

Embodiment 4 Next, a fourth embodiment of the present invention will be described with reference to FIG. In the above-described second and third embodiments, when the “all” and “majority” motion vectors are vectors that can be expressed based on the code length qsize, the code length psize is updated to the code length qsize. In the fourth embodiment, a condition is added, and the code length is updated when the increment of the code amount is equal to or smaller than a predetermined threshold.

That is, the moving picture coding apparatus according to the fourth embodiment is different from the second or third embodiment shown in FIG.
In the above configuration, a code length updating unit 60A is provided instead of the code length updating unit 60.

The code length updating unit 60A generates a code length when the input image 10 is encoded by expressing the magnitude of the motion vector generated in the current frame using the code length psize specified by the code length specifying unit 50. The input image 10 is encoded by expressing a motion vector generated in the current frame by using a first code amount to be used and a minimum code length qsize sufficient to express the size of the motion vector in the current frame. The code length psiz specified by the code length specifying unit 50 when the increment of the second code amount with respect to the first code amount becomes equal to or smaller than a predetermined threshold TH.
It is configured to update e to the code length qsize.

The operation of the code length updating unit 60A (method of updating the code length) according to the fourth embodiment will be specifically described below. Comparing the code amount of the motion vector generated when coding based on the code length psize and coding based on the code length qsize shows that the code length qs is smaller.
The coding amount of the motion vector itself is smaller when coding based on ize. However, it is necessary to determine whether the total code amount decreases in consideration of the code amount of the difference data generated in the encoder 33.

Hereinafter, a method of updating the code length in consideration of the code amount of the difference data generated in the encoder 33 will be specifically described. The motion vector mvp having the smallest distortion evaluation value 46 in the current frame can be encoded based on the code length psize, but if it cannot be encoded based on the code length qsize, the motion vector mvp cannot be adopted. Therefore, in this case, coding is performed by performing motion compensation using the motion vector mvq with the smallest distortion evaluation value 46 within the motion search range based on the code length qsize.

Here, the code amount per frame when coding with motion compensation using the motion vector mvp is fr
Assuming that the code amount per frame when performing motion compensation using cap (first code amount) and motion vector mvq is frcaq (second code amount), the code length in the encoding unit 33 The code amount that increases by updating to qsize is (frcaq-frcap). The number of motion vectors to be encoded in the current frame is large enough and the code amount increment (frcaq
-frcap) is small enough, the code length qsi in the next frame
If ze is specified, the code amount of the motion vector itself generated in the next frame is reduced.

However, the code amount ommv of the motion vector reduced by updating to the code length qsize is equal to the code length qsiz
The code amount which increases conversely by updating to e (frcaq-frcap)
If it is smaller, specifying the code length qsize in the next frame increases the total code amount, and lowers the coding efficiency. Therefore, the code length updating unit 60A not only reduces the code amount ommv that can be reduced, but also increases the code amount (frcaq-frca
p) is also evaluated to determine whether to specify the code length qsize.

If the number of motion vectors encoded per frame is afrmv, and the number of motion vectors that can be represented based on the code length qsize is afrmvq, the code length psize is updated to the code length qsize. The code amount ommv reduced per frame by ommv = (ps
ize-qsize) × afrmvq [bit].

On the other hand, by updating the code length psize to the code length qsize, the code amount (frc
aq-frcap) is frcaq-frcap = (bcaq-bcap) × (afrmv-afrmv
q) It is [bit]. Here, bcaq is a code amount per one coded block (image block) when coding is performed by performing motion compensation using the motion vector mvq, and bcap is a code amount obtained by performing motion compensation using the motion vector mvp. 1 when it became
This is the code amount per coded block.

The code length updating unit 50 determines the amount of code (frcaq-frcap) that increases per frame by comparing it with a threshold value TH, and the determination condition of TH ≧ (frcaq-frcap) is satisfied in the current frame. If the code length in the next frame
Update to qsize. This threshold value TH is set to the code amount ommv reduced per frame. Here, the code amount is set in order to set a margin in consideration of the fact that the motion vector generation characteristics change between the current frame and the next frame.
A value obtained by weighting the value of ommv (w × ommv: for example, w = 0.7) may be set as the threshold value TH.

As an example, when the range of the magnitude of the motion vector gradually decreases or conversely increases over a plurality of continuous frames, similar characteristics are expected in the next frame. However, if the motion of the subject changes abruptly and the generation characteristics of the motion vector change, this expectation may be deviated and the prediction of the entire generated code amount may be largely deviated. Therefore, in consideration of the possibility that the prediction of the generated code amount is disturbed due to the change in the generation characteristic of the motion vector, the weighting coefficient w is set to a large value (a small number close to 1).

Embodiment 5 Next, a fifth embodiment of the present invention will be described with reference to FIG. In the above-described second to fourth embodiments, the code length is updated by focusing on the magnitude of the “all” or “majority” motion vectors generated in the current frame. Then
The code length in the next frame is designated (updated) by focusing on the code amount generated in the current frame without being limited to the “all” or “majority” motion vectors.

That is, the code length updating unit 60 according to the fifth embodiment shown in FIG.
0 represents the size of the motion vector generated in the current frame using the code length psize specified by 0.
The motion vector generated in the current frame is expressed using a first code amount generated when encoding is performed and a code length qsize smaller than the code length psize specified by the code length specifying unit 50. And a second code amount generated when encoding is performed. When the increment of the second code amount with respect to the first code amount becomes equal to or smaller than a predetermined threshold TH, the code length specified by the code length specifying unit 50 is determined. It is configured to update psize to a code length qsize smaller than this code length psize.

The operation of code length updating section 60 according to the fifth embodiment will be described below. As described in the fourth embodiment, the code amount ommv reduced per frame by updating the code length psize to the code length qsize is ommv = (ps
ize-qsize) × afrmvq [bit], while the increasing code amount (frcaq-frcap) is frcaq-frcap = (bcaq-bcap) × (afrmv
-afrmvq) [bit].

Therefore, if the “all” or “majority” motion vectors in the current frame are equal to the code length qsize
If it can be expressed by, the number of motion vectors afrmvq that can be expressed based on the code length qsize becomes sufficiently large, and ommv >> (frcaq-frcap).

However, the code amount ommv is equal to the code amount (frcaq-frca
If the value does not exceed p), the coding efficiency is improved, and the number of motion vectors afrmvq does not necessarily need to be a sufficiently large value. That is, as long as the condition of ommv ≧ (frcaq-frcap) is satisfied, the code amount in the next frame can be reduced by specifying the code length qsize based on the code amounts frcap and frcaq.

Embodiment 6 FIG. Next, a sixth embodiment of the present invention will be described with reference to FIG. In the above-described fifth embodiment, the code length specified by the code-length specifying unit 50 is updated based on the code amount in the current frame. However, the code-length updating unit 60A according to the sixth embodiment includes: When the increment of the second code amount with respect to the first code amount exceeds a predetermined threshold value TH, the code amount generated when the input image 10 is encoded by the encoding unit 100 is within a range where the code amount is equal to or less than the predetermined threshold value. , The code length psize specified by the code length specifying unit 50
Is configured to be updated in the direction of increasing.

Here, as described above, the first code amount is:
The code amount generated when the input image 10 is encoded by expressing the size of the motion vector generated in the current frame using the code length psize specified by the code length specifying unit 50, and the second code amount Is a code amount generated when the input image 10 is encoded by expressing a motion vector generated in the current frame using a code length qsize smaller than the code length psize specified by the code length specifying unit 50.

The operation of code length updating section 60A according to the sixth embodiment will be described below. In Embodiments 2 to 4 described above, the number of motion vectors afrmvq is a sufficiently large value because attention is paid to “all” or “majority” motion vectors. However, as in Embodiment 5 described above, all"
Alternatively, without paying attention to the “majority” motion vector,
The code amount ommv reduced per frame is ommv ≧ (frcaq
-frcap).

Therefore, in the fifth embodiment, the threshold value T
H is set to the code amount ommv, and TH ≧
When the judgment condition (frcaq-frcap) is satisfied, the code length is updated to psize. As a result, the amount of generated codes can be reduced without paying attention to "all" or "majority" motion vectors.

In some cases, the increase of the total code amount exceeds the decrease of the motion vector code amount, and TH <(frcaq-frcap) may be satisfied. However, this threshold TH is equal to the code amount ommv
Alternatively, a value (w × ommv) obtained by multiplying the code amount ommv by a weighting coefficient w is set. Therefore, in the sixth embodiment, if the increase in the total code amount exceeds the decrease in the code amount of the motion vector, if the motion search range is allowed in a wider range, the code length is increased in the next frame. To specify.

More specifically, the code length psiz specified by the code length specifying unit 50 is within a range in which the code amount generated when the input image 10 is coded is equal to or less than a predetermined threshold value TS.
Update in the direction of increasing e, and update to a code length rsize larger than the code length psize. As a result, the amount of generated codes can be kept low, and a code length with good coding efficiency can be specified within the motion searchable range.

Here, as described in the first embodiment, the search range of the motion vector is range = (2M × R), and the physical upper limit of the code length rsize is the bit width of the integer R. However, the upper limit of the code length rsize is determined by introducing a predetermined threshold value TS. That is, the upper limit of the code length rsize is selected within a range in which the code amount generated when encoding the input image 10 is equal to or less than the predetermined threshold value TS. This threshold value TS is appropriately set within a range in which the code amount can be effectively suppressed, and may be the above-described threshold value TH.

Embodiment 7 FIG. Next, a seventh embodiment of the present invention will be described with reference to FIG. In the first to sixth embodiments, the code length is updated by the code updating unit 60. However, the moving picture coding apparatus according to the seventh embodiment shown in FIG. Embodiment 1
In the above configuration, a code length input unit 70 for inputting a code length specified from the outside is further provided, and the code length updating unit 6
0 selects the code length qsize generated by the code length updating unit 60 or the code length exsize input to the code length input unit 70 according to a predetermined condition, and updates the code length qsize specified by the code length specifying unit 50. It is configured to

The operation of the moving picture coding apparatus according to the seventh embodiment will be described below. The code length updating unit 60 determines that the code length exsize input to the code length input unit 70 and the code length qsize generated by the code length updating unit 60 are smaller than the code length psize specified by the code length specifying unit 50. When the code amount generated is smaller than when coding based on the code length psize, regardless of which code length exsize or qsize is adopted, either one is adopted in accordance with predetermined conditions described later.

On the other hand, when the code length exsize and the code length qsize are larger than the code length psize, the coding is performed based on the code length psize regardless of which code length exsize and code length qsize is adopted. Similarly, when the generated code amount is reduced, one of them is adopted according to a predetermined condition.

Here, as the above-mentioned predetermined condition, encoding is performed based on the code amount generated when encoding is performed based on the code length qsize generated by the code length updating unit 60 and the code length exsize specified from the outside. It is set to compare the amount of code generated when the encoding is performed and to select a code length that reduces the amount of code generated when the input image is encoded by the encoding unit 100.

Further, as the above-mentioned predetermined condition, it may be set that the code length exsize from the outside is set as the upper limit of the code length. According to this predetermined condition, the code length is actually
Even if the motion search can be performed in a wider range than the search range given by exsize, if the code length qsize does not exceed the code length exsize, the code length qsize is adopted and the code length qsi
If ze exceeds the code length exsize, the code length exsize is adopted.

Further, as the above-mentioned predetermined condition, the code length exsize may be set as the lower limit of the code length. According to this predetermined condition, for example, in a situation where it is desired to encode a motion vector with a small code length due to a restriction on the capacity of the buffer 36 for storing encoded data,
Code length for the next frame when performing normal encoding
It is possible to update to a code length q-size that is larger than exsize or a code length exsize, and forcibly specify an external code length exsize when performing non-normal encoding. As described above, by introducing the code length exsize from the outside, it is possible to use the code length properly according to the encoding method.

In the seventh embodiment, code length updating section 60 adopts either code length exsize or code length qsize. However, the present invention is not limited to this. May be assigned. In the seventh embodiment, the code length input unit 70 is added to the configuration of the first embodiment.
Has been described as an example, but may be applied to other embodiments 2 to 6.

Embodiment 8 FIG. Next, an eighth embodiment of the present invention will be described with reference to FIG. In the above-described seventh embodiment, the configuration is such that the code length exsize from the outside is selected and designated. However, the moving picture coding apparatus according to the eighth embodiment has the configuration of the seventh embodiment shown in FIG. Includes a code length updating unit 60B instead of the code length updating unit 60.

The code length updating unit 60B generates code lengths of a plurality of candidates within a motion search range allowed for an image block, and expresses a motion vector using the code lengths of the plurality of candidates. The code amount generated when each of the input images 10 is encoded is compared, and a code length with the smallest code amount is selected from a plurality of candidate code lengths.

The operation of the moving picture coding apparatus according to the eighth embodiment will be described below. According to the above-described seventh embodiment, actually, the code length specified by the code length specifying unit 50 is specified.
A code length smaller than psize may provide good coding efficiency, and a code length larger than psize may provide good coding efficiency. Therefore, in the eighth embodiment, a plurality of code lengths including a code length larger than and smaller than the code length psize specified by the code length specifying unit 50 are prepared.
Select the code length that gives the least code amount.

That is, the code length updating unit 60 B
As shown in the example, the code lengths psize1 to psize4 of a plurality of candidates that can be taken within the motion search range allowed for each image block in the current frame are generated and prepared. Here, psize1 is the minimum code length and gives the minimum motion search range in the frame currently being encoded. Also, psize4 is the maximum code length and gives the maximum motion search range.

The code length updating unit 60B has a code length psize1
For each of the code lengths of psize4 to psize4, the amount of code generated in the encoder 33 when encoding is performed based on each predicted image block with the smallest distortion evaluation value 46 is obtained and integrated. When the encoding for one frame is completed, the code amount for one frame generated by the encoding unit 33 for each code length is obtained.

Next, the total code amount is obtained by combining the code amount generated by the encoder 33 for each code length and the increase / decrease of the code amount of the motion vector caused by changing the code length psize. Here, if the total code amount generated when encoding based on the code length exsize is minimized, the code length ex
Select size.

If the total code amount generated when encoding based on the code length exsize is not minimum, select the code length with the minimum total code amount from the code lengths psize1 to psize4. I do. Code length update unit 60
B is the code length psize specified by the code length specifying unit 50
Is updated to the selected code length.

In the eighth embodiment, the code length updating unit 60B is responsible for the generation and selection of a plurality of candidate code lengths. However, the present invention is not limited to this. It does not limit the essence of the present invention. Further, in FIG. 5, four types of code lengths are illustrated as the code lengths of the plurality of candidates. However, the code lengths are not limited to the four types, and the code lengths of the plurality of candidates may be determined as needed.

Embodiment 9 FIG. Next, a ninth embodiment of the present invention will be described with reference to FIG. In the seventh embodiment, the code length specified from the outside is selected according to the predetermined condition. However, the code length updating unit 60 of the moving picture coding apparatus according to the ninth embodiment is configured to select the code length. A code amount obtained by expressing a motion vector using a code length exsize externally input to the input unit 70, and a motion vector obtained by expressing the motion vector using a code length qsize generated by the code length updating unit 60. The code amount is weighted and compared, and the code length is selected based on the result of the comparison.

Hereinafter, the operation of the moving picture coding apparatus according to the ninth embodiment will be described. In Embodiment 7 described above, for example, when the code length that gives the minimum total code amount is the code length exsize, the value of the code length qsize (psize) that is internally updated (specified) is also the code length exsize. It has the same value. However, in this case, the code length exsize specified externally is not preferentially selected and specified to the encoding unit 100, and the code length qsize (psize) that gives the minimum total code amount is , Just equal to the value of the code length exsize.

Therefore, the code length updating section 60 of the moving picture coding apparatus according to the ninth embodiment weights the code amount generated when coding is performed based on the internally generated code length qsize (psize). The weighted code amount and the code amount generated when encoding based on the code length exsize specified from the outside are weighted and compared, and the code length qsize (psize) or the code length e that gives a smaller weighted code amount
Select xsize.

As described above, by weighting and comparing the code amount generated when encoding based on each code length, it is determined whether or not the code amount (unweighted code amount) is minimized. Code length ex specified from outside
It becomes possible to select with priority given to size.

In the ninth embodiment, code length updating section 60 weights and compares code amounts to select a code length. However, the present invention is not limited to this. , And does not limit the essence of the present invention.

Embodiment 10 FIG. Next, a tenth embodiment of the present invention will be described. In the first to ninth embodiments,
Although the “predetermined image unit” is set to “one frame”, the moving image encoding apparatus according to the tenth embodiment differs from the configuration of the first embodiment shown in FIG. The image unit is composed of the above frames, and the number of frames constituting the image unit is variable.

For example, if the number of frames constituting the "predetermined image unit" is variably set in accordance with the magnitude of the motion vector generated in the current frame, the code may be changed in accordance with the magnitude of the image motion. The operation relating to designation (update) of the length can be simplified.

More specifically, if it is predicted from the magnitude of the motion vector in the current frame that a motion vector having the same magnitude will occur in the next several frames, the code length designation unit 50 The update of the code length by the update unit 60 is rejected, and the code length specified for the current frame is maintained over several frames. This allows
The number of updates of the code length is reduced, and the update operation is simplified.

In the tenth embodiment, the first embodiment is used.
Although the number of frames is made variable in the configuration described above, the present invention may be applied to other embodiments 2 to 9. In Embodiments 1 to 10 described above, an image unit is defined in units of frames. However, a part of one frame may be defined as an image unit, and the size of the image unit may be defined as needed. Good. Further, the image block is defined by dividing the image unit into several parts. However, the size of the image block is not limited as long as it does not exceed the size of the image unit.

[0116]

As is clear from the above description, according to the present invention, the following effects can be obtained. That is, according to the present invention, an input image is divided into predetermined image blocks, a motion vector is generated for each of the image blocks from the input image and a previously encoded prediction image, and the prediction is performed based on the motion vector. Encoding means for compensating for motion of an image, encoding the input image by motion compensation with the motion compensating means, and encoding the motion vector with a designated code length; A code length designating unit for designating the code length of the motion vector to the encoding unit in units of a predetermined image unit composed of image blocks, and the encoding unit encodes the input image for the next image unit At this time, the code length designation means designates the motion compensation means based on the magnitude of the motion vector generated in the current image unit. Since a code length updating means for updating the over de length reduces the code amount according to the magnitude of the motion vector, it is possible to improve the coding efficiency.

Further, the code length updating means may include a code length updating means for setting the size of all motion vectors generated in the current image unit smaller than the code length specified by the code length specifying means for the current image unit. When expressed using lengths, the code length specified by the code length specifying means is updated to the minimum code length sufficient to represent the magnitudes of all the motion vectors, so that the entire next image unit is updated. Can be suppressed low.

Further, the code length updating means may determine that the magnitude of a predetermined percentage of motion vectors, which is a majority of the motion vectors generated in the current image unit, is equal to the code length designation for the current image unit. When the code length specified by the code length specifying means is expressed using a code length smaller than the code length specified by the means, the code length specified by the code length specifying means is expressed by a magnitude of a predetermined ratio of motion vectors which is a majority of the motion vectors. Since the code length is updated to the minimum code length, the code amount of the entire next image unit can be substantially reduced.

Further, the code length updating means encodes the input image by expressing the magnitude of the motion vector generated in the current image unit using the code length designated by the code length designating means. The input image is encoded by expressing the motion vector generated in the current image unit using the first code amount generated when and the minimum code length sufficient to express the size of the motion vector. And a second code amount generated when the second code amount is generated, and when the increment of the second code amount with respect to the first code amount becomes equal to or smaller than a predetermined threshold, the code length specified by the code length specifying means is updated. Therefore, the code amount of the entire next image unit can be effectively reduced.

Further, the code length updating means encodes the input image by expressing the magnitude of the motion vector generated in the current image unit using the code length specified by the code length specifying means. The input image is coded by expressing a motion vector generated in the current image unit using a first code amount generated at the time and a code length smaller than the code length specified by the code length specifying means. And a second code amount generated when the code length is specified by the code length specifying means when an increment of the second code amount with respect to the first code amount is equal to or smaller than a predetermined threshold value. Since the code length is updated to a code length smaller than the code length specified by the length specifying means, the code amount of the entire next image unit can be suppressed low.

Further, the code length updating means, when the increment of the second code amount with respect to the first code amount exceeds a predetermined threshold, when the input image is encoded by the encoding means, Since the code length to be generated is updated in a direction in which the code length specified by the code length specifying means is increased within a range in which the generated code amount is equal to or less than the predetermined threshold value, the coding efficiency is improved within the motion searchable range so that Of the entire image unit can be effectively reduced.

Further, the apparatus further comprises code length input means for inputting a code length specified from the outside, wherein the code length updating means generates the code length generated by the code length updating means or the code length in accordance with a predetermined condition. The code length input to the input means is selected, and the code length specified by the code length specifying means is updated, so that the encoding efficiency is improved under predetermined conditions, so that the next image unit as a whole is improved. Can be effectively reduced.

Further, the code length updating means generates code lengths of a plurality of candidates that can be taken within a motion search range allowed for the image block, and uses the code lengths of the plurality of candidates to generate the motion vector. Is expressed, and the amount of code generated when each of the input images is encoded is compared, and the code length that minimizes the code amount is selected from the code lengths of the plurality of candidates. Thus, the code amount of the entire next image unit can be effectively reduced so that the coding efficiency is improved in the search range to be searched.

Further, the code length updating means includes a code length obtained by expressing the motion vector using the code length input to the code length input means, and a code length generated by the code length updating means. Is used to weight and compare the code amount obtained by expressing the motion vector, and the code length is selected based on the result of the comparison. The code length specified from the outside can be specified through the code length input means.

Further, since the image unit is composed of two or more frames and the number of frames constituting the image unit is variable, the code amount of the next image unit as a whole is changed according to the magnitude of the change of the image. It can be kept low.

[Brief description of the drawings]

FIG. 1 is a block diagram illustrating a configuration of a video encoding device according to Embodiment 1 of the present invention.

FIG. 2 is a block diagram illustrating a configuration of a video encoding device according to Embodiment 4 of the present invention.

FIG. 3 is a block diagram illustrating a configuration of a video encoding device according to Embodiment 7 of the present invention.

FIG. 4 is a block diagram illustrating a configuration of a video encoding device according to Embodiment 8 of the present invention.

FIG. 5 is a diagram for explaining a relationship between a code length generated by a code length updating unit of a video encoding device according to Embodiment 8 of the present invention and a motion search range.

FIG. 6 is a block diagram of a moving image compression apparatus according to the related art.

FIG. 7 is a flowchart showing a flow of an operation of the moving picture compression apparatus according to the conventional technique.

FIG. 8 is a diagram illustrating coefficients internally generated by a moving image compression apparatus according to the related art.

[Explanation of symbols]

Reference Signs List 10 input image, 10A predicted image, 31 image memory, 32 difference calculator, 33 encoder, 35 variable encoder, 36 buffer, 38 decoder, 40 adder,
44 motion compensation prediction unit, 50 code length designation unit, 60,
60A, 60B code length updating unit, 70 code length input unit, 90 encoded data, 100 encoding unit.

Claims (10)

[Claims] 1. An input image is divided into predetermined image blocks, a motion vector is generated for each of the image blocks from the input image and a previously encoded predicted image, and a motion of the predicted image is calculated based on the motion vector. Encoding means for encoding the input image with motion compensation by the motion compensating means, and encoding the motion vector with a designated code length; and A code length designating unit for designating the code length of the motion vector to the encoding unit in units of a predetermined image unit, when the encoding unit encodes the input image for the next image unit, The motion compensation means updates the code length designated by the code length designation means based on the magnitude of the motion vector generated in the current image unit. Video encoding apparatus characterized by comprising a code length updating means for. 2. The code length updating unit according to claim 1, wherein the magnitude of all motion vectors generated in the current image unit is smaller than the code length specified by the code length specification unit for the current image unit. 2. The method according to claim 1, wherein when represented using a length, the code length specified by the code length specifying means is updated to a minimum code length sufficient to represent the magnitudes of all the motion vectors. 5. The video encoding device according to item 1. 3. The code length updating means according to claim 1, wherein a magnitude of a predetermined ratio of motion vectors, which is a majority of motion vectors generated in said current image unit, is set to said code length designation for said current image unit. When the code length specified by the code length specifying means is expressed using a code length smaller than the code length specified by the means, the code length specified by the code length specifying means is expressed by a magnitude of a predetermined ratio of motion vectors which is a majority of the motion vectors. 2. The moving picture coding apparatus according to claim 1, wherein the moving picture coding apparatus updates the minimum code length to the minimum code length. 4. The method according to claim 1, wherein the code length updating means encodes the input image by expressing a magnitude of a motion vector generated in the current image unit using a code length designated by the code length designating means. The first occurring in
And a minimum code length sufficient to represent the size of the motion vector, and expressing the motion vector generated in the current image unit and encoding the input image. 2 and obtain the first code amount.
4. The moving picture code according to claim 2, wherein the code length designated by the code length designation means is updated when an increment of the second code quantity with respect to the code quantity is equal to or less than a predetermined threshold value. Device. 5. The apparatus according to claim 1, wherein said code length updating means encodes said input image by expressing a magnitude of a motion vector generated in said current image unit using a code length designated by said code length designating means. The first occurring in
And a second code generated when the input image is encoded by expressing a motion vector generated in the current image unit using a code amount smaller than the code length specified by the code length specifying means. And the code amount of the first
When the increment of the second code amount with respect to the code amount is equal to or smaller than a predetermined threshold value, the code length specified by the code length specifying unit is updated to a code length smaller than the code length specified by the code length specifying unit. The moving picture coding apparatus according to claim 1, wherein: 6. The code length updating means, when an increment of the second code amount with respect to the first code amount exceeds a predetermined threshold value, the code length updating means generates when the input image is encoded by the encoding means. The moving picture encoding apparatus according to claim 5, wherein the code length is updated in a direction to increase the code length specified by the code length specifying means within a range in which the code amount to be performed is equal to or less than the predetermined threshold. 7. Code length input means for inputting a code length specified from outside, wherein said code length updating means generates the code length generated by said code length updating means or said code length input according to a predetermined condition. 7. The moving picture coding apparatus according to claim 1, wherein the code length input to the means is selected, and the code length specified by the code length specifying means is updated. 8. The code length updating means generates code lengths of a plurality of candidates within a motion search range allowed for the image block, and converts the motion vector using the code lengths of the plurality of candidates. 8. The method according to claim 7, wherein code amounts generated when the input image is expressed and encoded are compared, and a code length that minimizes the code amount is selected from the code lengths of the plurality of candidates. 5. The video encoding device according to item 1. 9. The code length updating means, comprising: a code amount obtained by expressing the motion vector using the code length input to the code length input means; and a code length generated by the code length updating means. 8. The moving picture coding apparatus according to claim 7, wherein the coding amount obtained by expressing the motion vector is compared with each other by weighting, and a code length is selected based on a result of the comparison. 10. The moving picture coding apparatus according to claim 1, wherein said picture unit comprises two or more frames, and the number of frames forming said picture unit is variable.