JP2000201353A - Image encoding method, image encoder and storage medium in which image encoding program is stored - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide an image encoding method and image encoder to reduce hierarchizing loss while holding scalability when code quantity is controlled in the case of encoding a wavelet conversion coefficient. SOLUTION: An inputted moving image signal 1 is converted in a wavelet conversion part 2. The conversion coefficient is quantized at a specified quantization step in a quantization part 4 and encoded as the lowest order layer in an order from the conversion coefficient in a low band component toward the conversion coefficient in a high band component in a variable length encoding part 12. Furthermore, when the conversion coefficient is similarly quantized and encoded as a second layer at a quantization step smaller than the above quantization step, the quantization coefficient after quantization is reflectively encoded as reducing a quantization error in each layer until generated code quantity becomes specified code quantity by using an inverse quantization part 16 and a quantization part 19. In this case, the quantization steps of each layer in object frames to be encoded are decided with a quantization step calculating part 7 by the generated code quantity of the past frame for each layer calculated in a layer code quantity calculating part 15 and the quantization steps.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention belongs to the technical field of coding moving pictures and still pictures and controlling the amount of codes.

[0002]

2. Description of the Related Art In image compression coding, a discrete wavelet transform is used to remove a correlation between pixels. The discrete wavelet transform decomposes an image into frequency band components having different resolutions, similarly to hierarchical subband division. The total number of decomposed transform coefficients is the same as the number of pixels of the original image, but generally, signal energy is concentrated on lower frequency components.

[0003] Wavelet transform coefficients are obtained by a one-dimensional filter bank and sub-sampling processing. As shown in FIG. 2A, the outputs of the low-pass filter H _l (z) and the high-pass filter H _h (z) are obtained in the horizontal direction of the image X (z), and the outputs are １ ／ (↓ 2). ). Furthermore, by applying the same processing in the vertical direction, the image is decomposed into four components LL ₁ , LH ₁ , HL ₁ , and HH ₁ .
Among them, the lowest frequency component LL ₁ LL _2, L
Recomposes to H ₂ , HL ₂ , HH ₂ . By repeating this process for the required number of levels, the octave division of the image shown in FIG. 3 can be obtained.

In the reproducing process, as shown in FIG. 2A, a zero value is inserted between the coefficients in the vertical direction of the band components LL _{i + 1} and LH _{i + 1} (↑ 2), and the vertical direction is inserted. Are subjected to synthesis filters F _l (z) and F _h (z), respectively, and the sum of their outputs is obtained. The same processing is performed in the horizontal direction by HL _{i + 1} , HH
Perform on the _{i + 1} component. That is, 2 components obtained, LL _i components are synthesized by performing the same filtering process in the horizontal direction. By repeating this synthesizing process by the required number of levels, a reproduced value of the image is obtained. In order for the image to be completely reconstructed, the decomposition / synthesis filter must satisfy the following conditions.

H _l (z) H _h (−z) + H _l (−z) H _h (z) = c (1) F _l (z) = (2 / c) H _h (−z) ( 2) F _h (z) = (2 / c) H _l (−z) (3) Here, the tap length of the decomposition / synthesis filter is an odd number, and c is an arbitrary constant. FIG. 4 shows the relationship between the decomposition and synthesis filter coefficients.

[0006] Wavelet transform coefficients obtained by subjecting an input image signal to horizontal and vertical wavelets are quantized and coded. Generally, the same level of high frequency band (HL
_m , LH _m , HH _m , m∈natural number) are quantized in the same quantization step, and the quantized wavelet transform coefficients are Run-Length coded, Run-L
Variable length coding is performed using evel coding or the like to reduce the generated code amount. At this time, the wavelet transform coefficient values in the other high frequency bands except for the lowest frequency band LL can be approximated by a Laplace distribution concentrated near 0. Scalar quantization with center dead zone is used. Note that LL
If the components are roughly quantized, visually disturbing distortions are noticeable in the decoded image. Therefore, a general method of performing lossless encoding using DPCM without performing quantization is used.

[0007] Regarding the encoding method of wavelet transform coefficients, an encoding method utilizing correlation between coefficients at similar positions in each frequency band has been proposed.
In two-dimensional octave division using wavelets, 1
The level is decomposed into four components LL, LH, HL, and HH by the decomposition of the level.
From L _m, it is decomposed into (3m + 1) pieces of components up to _{LH 1, HL 1, HH 1} . FIG. 5 is an example of frequency band division when m = 3.

In order to eliminate the data redundancy due to the self-similarity described above, LH _i , H obtained by the i-level division are used.
The high frequency components of L _i and HH _i and the LH of (i−1) level
_i-1 , HL _i-1 , and HH _i-1 define a parent-child relationship at similar positions. The wavelet transform coefficients at similar positions are shifted from the lower resolution side to the higher frequency side with the higher resolution. Assuming a tree structure, if all the transform coefficients after a certain node after quantization are zero, one ZTR (Zero
By replacing it with a (Tree Root) symbol,
Perform further data compression.

FIG. 6 shows an example of a zero tree. In this case, the symbols that are actually subjected to the variable length coding are: a ZTR code; and a quantized coefficient value 0 (IZ code: Iso, not ZTR code).
rate Zero) Coefficient value after quantization (LEVEL). This wavelet coefficient coding method is called zero tree coding. At this time, four children correspond to one parent at the i level, but three children LH _m , HL _m , and HH _m correspond to the lowest frequency band LL _m exceptionally.

Further, a zero-tree encoding method is used,
As a still image encoding method, EZW (The Embe
dded Zerotree Wavelet Alg
(orthm) method has been proposed. FIG. 7 shows a block diagram of the EZW method.

In this method, the wavelet transform coefficient (33) is quantized (34) at a constant quantization step (Q ₁ = 2 ^Lmax-1 ) between frequency bands, and the quantized coefficient (35) is A zero tree is formed (36). And
The coefficients and ZTR codes of the calculated frequency bands from LLm to LH ₁ , HL ₁ , and HH ₁ are variable-length coded (3
7) Yes. The above processing steps (32 to 37) are the same as those in the zero tree encoding method.

At this time, between the coefficient (39) obtained by inversely quantizing (38) the quantized wavelet transform coefficient (35) actually encoded and the original wavelet transform coefficient (33) in the same quantization step. since the quantization error (40) occurs, the quantization error using the zerotree method similarly, encoded in order from the quantization error of LL _m to quantization error of _{LH 1, HL 1, HH 1} (41 To 43).

The processing steps for re-encoding the quantization error are shown by the shaded areas in FIG. When this processing is performed, the quantization error is similarly quantized (41). However, the quantization step at the time of this quantization error encoding is half (Q ₂ = Q) of the quantization step of the layer encoded immediately before. _1/2 = 2
^Lmax-2 ). Then, the processing of the shaded area is performed while halving the quantization step for each iteration.
Recursively, coding is repeated until the quantization step of the last layer (layer L _max ) becomes _{QL max} = 1, that is, there is no quantization error.

As a result, as shown in FIG.
, The quantization error decreases in order from the layer L _max , thereby realizing scalability in the image quality direction. Further, in each layer, as shown by arrows in the order of frequency band coefficient scanning in FIG. 8, LL _m , ｛HL, LH, H
H｝ _m−1 ... {HL, LH, HH} By performing coding using inter-frequency band scanning in the order of ₁ , spatial scalability in each layer is also realized.

In the EZW method, which is a still image coding method, control of the generated code amount is realized by utilizing these two scalabilities. If the target code amount is specified at the time of encoding, the code amount control that keeps the generated code amount within the target code amount by interrupting the encoding process immediately before the generated code amount matches or exceeds the target code amount. Has been realized. At this time, between each frequency band,
Since scalability is realized by the order of coefficient scanning between layers, visually disturbing distortion does not occur even if coding is interrupted at any place.

[0016]

[Table 1]

However, considering the coding efficiency of this frame, since the wavelet transform coefficient (23) is coded (developed on a bit plane) by dividing it into L _max layers, a loss called a layering loss is generated. Occurs. Specifically, when a certain wavelet transform coefficient “+75” is coded by the EZW method, zero-tree coding is performed in eight layers shown in Table 1. Table 1 shows an example in which the coefficient value “+75” is encoded. Each column of the table is from left to right. 1. Hierarchical level 2. Quantization step in the hierarchy 3. Input level to be quantized Wavelet transform coefficient in layer 1 Quantization error in layer (l-1) after layer l (l> 1) 4. Quantization index 5. A value obtained by inversely quantizing the coefficient after quantization. This is a decoding coefficient value when layers 1 to 1 are decoded. Each row indicates, from the top, the lowest layer to the highest layer. As shown in Table 1, input coefficient +75
To encode +1, -1,0, + 1, -1, +1,0, despite the fact that a single encoding is originally sufficient.
-1 and 8 times. This results in coding loss called layering loss.

Also, it has been assumed that all wavelet transform coefficients are quantized in the same quantization step or in the same quantization step for transform coefficients in the same frequency band. Is variable for each frequency band, and the moving image coding system ISO / IEC111
72-2 (MPEG-1) and ITU-T Recommendation H.264. 26
For example, a method of realizing code amount control using the quantization control adopted in Example 1 is also conceivable.

However, when transform coefficients are coded using a tree, such as in zero-tree coding, the quantized coefficients in the high frequency band corresponding to the transform coefficients to be coded must be determined at that time. This is obvious from FIG. Therefore, when the coefficients are encoded using the zero tree encoding and from the low frequency band to the high frequency band (FIG. 8), it is impossible to use the quantization control for each frequency band for the code amount control. .

As a unit for outputting a tree formed with coefficients in the lowest frequency band LL as a parent to encoded data,
A method is also conceivable in which the quantization step is fixed for each tree, and the amount of code is controlled by encoding while changing the quantization step for each tree.

However, in this case, a code indicating the quantization step of the tree itself is embedded for each tree, which is disadvantageous in terms of coding efficiency. Further, when the generated code amount approaches the target code amount and the coding of the frame is discontinued, there is a possibility that only the upper part of the screen is coded and the lower part is not coded. Further, when this coefficient scan is used, the spatial scalability functionality cannot be used for the encoded data, which is not desirable in this respect.

[0022]

When a moving image is encoded by using the wavelet transform and the zero-tree encoding, when the generated code amount per frame is controlled by using the EZW method, one transform coefficient is converted into a plurality of transform coefficients. Hierarchical loss due to encoding by dividing into layers occurs, and encoding efficiency is reduced.

Even when the wavelet transform coefficients are subjected to zero-tree coding, a method of describing the transform coefficients in the coded data for each tree as a scan order of the transform coefficients is used. For example, a method of changing the quantization step for each tree and controlling the code amount such that the quantization step is changed for each macroblock in H.261 or the like can be considered. A code indicating itself is embedded for each tree, which is not only disadvantageous in terms of coding efficiency, but also makes spatial scalability functionality unusable for coded data, which is not desirable.

[0024] An object of the present invention is to solve the above problems when encoding a wavelet transform coefficient obtained as a result of performing a wavelet transform on a moving image input signal, when controlling a generated code amount per frame. It is an object of the present invention to provide an image encoding method and apparatus that has a scalability function and has a very low layering loss by wavelet transform coefficient encoding and code amount control.

[0025]

SUMMARY OF THE INVENTION The present invention has been made to solve the above-mentioned problems, and employs the following means.

One of the means is to divide the frame to be encoded into a high band and a low band in the horizontal and vertical directions, and to re-band divide the low band components in both the horizontal and vertical directions in the same manner. The band division process is recursively repeated a designated number of times, transform coefficients belonging to a plurality of frequency bands obtained by the band division are quantized in a predetermined quantization step, and the transform coefficients in the low-frequency component are sequentially increased from the transform coefficients in the low-frequency component. Encode as the lowest layer toward the transform coefficients in the band component, further quantize the quantization error in a quantization step smaller than the quantization step, and encode the quantization error in the same way as the second layer In the moving picture coding method of performing coding while reducing the quantization error in each layer recursively until the generated code amount of the coding target frame reaches a predetermined code amount, In encoding a certain frame in the signal, a quantization step of each layer in the encoding target frame is performed by a generated code amount for each layer of a frame that has been encoded in the past and a quantization step used in each layer. Is determined.

Alternatively, in the above image encoding method, when the quantization step of the encoding target frame is determined based on the generated code amount and the quantization step for each layer of the previously encoded frame, The quantization step of the lowest layer of the encoding target frame is determined so that the generated code amount of the lowest layer of the frame is within a predetermined ratio with respect to the target code amount of the encoding target frame. Image encoding method.

Alternatively, in the above-described image encoding method, when the quantization step of the frame to be encoded is determined based on the generated code amount and the quantization step for each layer of the frame which has been encoded in the past. In the already-encoded frame, a quantization step of a layer that exceeds the target code amount of the encoding target frame and a layer immediately before exceeding the target code amount of the encoding target frame;
And a quantization step of the lowest layer of the encoding target frame is calculated and determined by using the respective generated code amounts up to the two layers.

Alternatively, in the above-described image coding method, when the quantization step of the frame to be encoded is determined based on the generated code amount and the quantization step for each layer of the frame which has been encoded in the past, the encoding step may be performed in the past. An image encoding method characterized in that a quantization step of a highest layer in a converted frame is determined as a quantization step of a lowest layer of an encoding target frame.

Alternatively, when encoding a still image,
When calculating an optimal quantization step while encoding the same input signal a plurality of times, any one of the above-described image encoding methods is applied, and each layer obtained when the encoding target still image is encoded immediately before And a quantization step for calculating an optimum quantization step for the lowest layer.

Alternatively, the frame to be encoded is divided into high and low bands in the horizontal and vertical directions, and the low band components in both the horizontal and vertical directions are similarly re-band divided, and Means for recursively specifying a specified number of times, means for quantizing transform coefficients belonging to a plurality of frequency bands obtained by the band division in a predetermined quantization step, and means for performing the transform in the low-frequency component. Means for encoding the transform coefficients in the high-frequency component in order from the coefficient as the lowest layer, and further quantizing the quantization error in a quantization step smaller than the quantization step, and Means for encoding as two layers, and until the generated code amount of the encoding target frame reaches a predetermined code amount, recursively, means for encoding while reducing the quantization error in each layer, Have In the image encoding apparatus, when encoding a certain frame in a moving image input signal, a unit that calculates a generated code amount for each layer of a frame that has been encoded in the past, and a quantization step used in each layer. Means for recording, and means for determining the quantization step of each layer in the encoding target frame from the generated code amount, the quantization step, and the target code amount of the encoding target frame, It is an image encoding device characterized by the following.

Alternatively, in the above-described image encoding apparatus, when a frame to be encoded is a still image and a certain frame in the moving image input signal is encoded, a generated code for each layer of a previously encoded frame is encoded. An image encoding apparatus characterized in that the means for calculating the amount is a means for calculating a generated code amount for each layer of a still image that has been encoded in the past when encoding a certain still image.

Further, a program for causing a computer to execute the process of the image encoding method described above is stored in a storage medium readable by the computer. Medium.

When a moving image is encoded using the conventional wavelet transform and zero-tree encoding, if the generated code amount per frame is simply controlled by using the EZW method, one transform coefficient can be assigned to many layers. In this case, there is a problem that a layering loss occurs due to the division and encoding, and the encoding efficiency is reduced. According to the present invention, it is possible to greatly reduce the hierarchization loss by optimizing the quantization step of the lowest layer of the encoding target frame.

[0035]

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

FIG. 1 is a diagram showing the flow of processing according to a moving picture coding method and a code amount control method according to an embodiment of the present invention, and also showing the structure of an encoder. In the figure,
The term “part” represents the XX part or the XX processing step.

In the present method and the present encoder, first, an input moving image signal 1 is input to a wavelet transform unit (or processing step) 2 and subjected to wavelet transform.
The wavelet transform coefficient 3 is quantized by the quantization unit (or processing step) 4 to become a transform coefficient 5 after wavelet quantization.

At this time, the quantization unit (or processing step)
In a quantization step 6 used in step 4, a quantization step calculation unit (or processing step) 7 calculates a quantization step 8 of each layer of the immediately preceding frame and a generated code amount 9 of each layer of the immediately preceding frame based on the following. It is calculated using the proposed method.

The wavelet quantized transform coefficient 5 is input to a zero tree forming unit (or processing step) 10 to form a zero tree, and the calculated zero tree symbol 11, ie, LL _m to LH ₁ , HL ₁ , HH. ₁
The coefficients and the ZTR code in each frequency band up to are encoded by the variable length encoding unit (or processing step) 12 and output as encoded data 13. At the same time, the generated code amount 14 at this time is stored in the layer code amount calculation unit (or processing step) 15 and is used for calculation of a quantization step in coding the next frame.

The coding of the lowest layer has been described above. If the generated code amount 14 is smaller than the target code amount, the inverse quantization unit (or processing step) 16 converts the transform coefficient 5 after wavelet quantization into inverse quantization. This is subtracted from the wavelet transform coefficient 3 by a subtractor (or subtraction processing step) 17 to obtain a wavelet transform coefficient quantization error 18.
Is calculated.

The wavelet transform coefficient quantization error 18 is quantized by a quantization unit (or processing step) 19 and outputs a coefficient 20 after quantization error quantization. put it here,
The quantization unit (or processing step) 19 is a quantizer for the second and higher layers, and is quantized by a quantization step smaller than the quantization step 6 used in the quantization unit (or processing step) 4. You. The coefficient 20 after the quantization error quantization is a zero tree forming unit (or processing step) 2
1. In the variable length coding unit (or processing step) 22, zero tree coding is performed in the same manner as in the lowest layer, and output as coded data 13.

Also, a quantization unit (or processing step) 1
In the quantization step 8 used in step 9 and the variable code length 14 in the variable length coding unit (or processing step) 21, the generated code amount 14 is the same as the coding amount of the lowest layer. It is stored in (or processing step) 15 and is used to calculate a quantization step when encoding the next frame.

Further, the generated code amount of the second layer shown by the inverse quantization unit (or processing step) 16 to the variable length coding unit (or processing step) 22 is added to the generated code amount of the lowest layer. If the added value, that is, the generated code amount of the encoding target frame is smaller than the target code amount, the processing of the inverse quantization unit (or processing step) 16 to the variable length coding unit (or processing step) 22 is recursively performed. And the process is continued until the target code amount becomes equal to the generated code amount of the frame. When zero-tree coding is performed recursively on the third layer and above, the selectors (or processing steps) 23 and 24 select the lower and left sides, respectively, to set the quantum error 18 and the quantization error coefficient 2 after quantization.
Select 0 as input.

In the embodiment of the present invention, when encoding a wavelet transform coefficient obtained as a result of performing a wavelet transform on a moving image input signal, a function of spatial scalability is used when controlling a generated code amount per frame. In order to realize a code amount control method that has a very low layering loss while having the same, the quantum of the encoding target frame is set so that the generated code amount of the lowest layer of the encoding target frame takes a value close to the target code amount. Determine the optimization step.

More specifically, the target code amount of the n-th frame is represented by T _n , and the l-th (l
<L _max ) When the generated code amount of the layer is S _n , _l and the quantization step of the l-th layer of the n-th frame is Q _n , _l ,

[0046]

(Equation 1)

The number of layers L satisfying the equation (4) is obtained, and
The quantization step Q _N , ₁ of the lowest layer of the encoding target frame is determined from _N−1 , _L−1 , Q _N−1 , _L , and _SN−1 , _l .
Here, the N-th frame is an encoding target frame.

From Q _N−1 , _L−1 , Q _N−1 , _L and S _N−1 , _l , Q _N , ₁
Various methods can be used to calculate the amount of generated code up to the (L-1) th layer of the immediately preceding frame and the (L-
1) Using the product of the quantization step of the layer, the product of the generated code amount up to the L-th layer of the immediately preceding frame and the quantization step of the L-th layer, and the target code amount of the current frame, ) Can be calculated by using the equation.

[0049]

(Equation 2)

When the generated code amount of the immediately preceding frame is very close to the target code amount of the encoding target frame,

[0051]

(Equation 3)

By substituting equation (6) into equation (5), L, L-1
To L _max , as shown in equation (7), the quantization step (Q _N−1 ,
_Lmax ) can be used as the quantization step (Q _N , ₁ ) of the lowest layer of the encoding target frame.

[0053]

(Equation 4)

As described above, when encoding each layer by the zero-tree encoding, it is assumed that the quantization step Q _n , _l is fixed irrespective of to which frequency band the transform coefficient to be encoded belongs. Although the specific method has been described, the same method is applied to the case where the quantization step is inclined by each divided frequency band in each layer, and the quantization of each frequency band of the lowest layer of the encoding target frame is performed. It is possible to calculate steps.

Further, even when encoding is performed using the quantization step of the lowest layer of the encoding target frame calculated using this method, it does not always coincide with the target code amount. When the generated code amount of the lowest layer is significantly smaller than the target code amount, the quantization step of the second layer is set to a value smaller than the quantization step of the lowest layer, and the quantization error is zero-tree coded. Avoid buffer underflow. Similarly, if the generated code amount of the lowest layer is much larger than the target code amount, the code amount is controlled by terminating the coding when the target code amount is exceeded, as in the EZW method. The scan between frequency bands in this method is shown in FIG.
(2), spatial scalability is achieved because of the use of a scan as shown in (1), so that visually disturbing distortion does not occur even when the coding is discontinued during the coding of the lowest layer.

Further, when the optimum quantization step is calculated while encoding the same input signal a plurality of times, the above-mentioned moving image code amount control method is similarly applied. It is also possible to calculate the optimal quantization step of the lowest layer by using the generated code amount of each layer and the quantization step when the encoding target still image is encoded. Specifically, in the equations (4) to (7), the same calculation can be performed by replacing N with the number of encoding passes in the repetitive encoding of the still image.

It is needless to say that some or all of the units described with reference to FIG. 1 can be made to function using a computer, or the processing steps described with reference to FIG. 1 can be executed with a computer.
A computer-readable storage medium such as a floppy disk (FD), an MO, a ROM, and a memory card may be used to store a program for causing a computer to function as each unit or a program for executing the processing steps on the computer. , CD, DVD,
It can be stored and provided on a removable disk or the like and distributed.

[0058]

As described above, according to the present invention,
Wavelet transform coefficient coding with very low layering loss while having a scalability function when coding moving images using wavelet transform and zero tree coding or recursively coding still images multiple times And code amount control can be realized.

[Brief description of the drawings]

FIG. 1 is a diagram showing a processing flow of a moving image encoding method and a code amount control method according to an embodiment of the present invention, and a configuration of an apparatus.

FIGS. 2A and 2B are diagrams illustrating a decomposition process in a wavelet transform, and FIG. 2B is a diagram illustrating a synthesis process.

FIGS. 3A, 3B, and 3C are diagrams illustrating octave division by decomposition processing in wavelet transform.

FIGS. 4A and 4B are diagrams showing a mutual relationship between decomposition / synthesis filter coefficients in a wavelet transform.

FIG. 5 is a diagram illustrating an example of frequency band division at the time of three-level division.

FIGS. 6A, 6B, and 6C are diagrams showing examples of a zero tree.

FIG. 7 is a diagram illustrating an EZW method which is a still image encoding method.

FIG. 8 is a diagram illustrating a case where a coefficient is encoded from a low frequency band to a high frequency band using zero tree encoding.

[Explanation of symbols]

 DESCRIPTION OF SYMBOLS 1 ... Input moving image signal 2 ... Wavelet transformation part (or processing step) 3 ... Wavelet transformation coefficient 4 ... Quantization part (or processing step) 5 ... Transform coefficient after wavelet quantization 6 ... Quantization step 7 ... Quantization processing step Calculating unit (or processing step) 8: Quantizing step of each layer of the immediately preceding frame 9: Generated code amount of each layer of the immediately preceding frame 10: Zero tree forming unit (or processing step) 11: Zero tree symbol 12: Variable length Encoding unit (or processing step) 13 Encoded data 14 Generated code amount 15 Layer code amount calculation unit (or processing step) 16 Inverse quantization unit (or processing step) 17 Subtractor 18 Wavelet transform coefficient Quantization error 19 Quantizer (or processing step) 20 Quantization error Coefficient after quantization 1 ... zerotree forming unit (or the process steps) 22 ... variable length coding unit (or the process steps) 23 ... selector 24 ... selector

 ──────────────────────────────────────────────────続 き Continuing on the front page (72) Inventor Jun Shimizu 3-19-2 Nishi Shinjuku, Shinjuku-ku, Tokyo Japan Telegraph and Telephone Corporation (72) Inventor Junichi Nakajima 3-192 Nishi-Shinjuku, Shinjuku-ku, Tokyo No. Nippon Telegraph and Telephone Corporation F term (reference) 5C059 KK27 MA24 MA32 MA34 MC11 ME01 PP01 PP04 SS20 TA46 TB04 TC18 TC38 TD11 UA02 UA15 UA39 5C078 BA53 BA64 CA21 DA00 DA01 DA07 DB05

Claims (8)

[Claims] 1. A band to be encoded is divided into high and low bands in the horizontal and vertical directions, and low band components in both the horizontal and vertical directions are similarly re-band divided, and the re-band dividing process is performed. Is recursively specified number of times, the transform coefficients belonging to a plurality of frequency bands obtained by the band division are quantized in a predetermined quantization step, and the transform coefficients in the low-frequency component are sequentially arranged in the high-frequency component. Encoding as the lowest layer toward the transform coefficient of, further quantizing the quantization error in a quantization step smaller than the quantization step, and encoding the quantization error as a second layer, A moving image coding method for recursively coding while reducing a quantization error in each layer until the generated code amount of the coding target frame reaches a predetermined code amount. In encoding a frame to be encoded, the quantization step of each layer in the encoding target frame is determined by the generated code amount of each layer of the previously encoded frame and the quantization step used in each layer. An image coding method characterized by the above-mentioned. 2. The image encoding method according to claim 1, wherein a quantization step of the encoding target frame is determined based on a generated code amount and a quantization step for each layer of a previously encoded frame. The quantization step of the lowest layer of the encoding target frame is determined so that the generated code amount of the lowest layer of the encoding target frame is within a predetermined ratio with respect to the target code amount of the encoding target frame. An image coding method characterized by the above-mentioned. 3. The image encoding method according to claim 1, wherein a quantization step of the encoding target frame is determined based on a generated code amount and a quantization step for each layer of a previously encoded frame. In frames that have been encoded in the past, a quantization step of a layer that has exceeded the target code amount of the encoding target frame and a layer that has just exceeded the target code amount of the encoding target frame, and respective generated codes up to the two layers. An image encoding method, wherein a quantization step of a lowest layer of an encoding target frame is calculated and determined using an amount. 4. The image encoding method according to claim 1, wherein a quantization step of the encoding target frame is determined based on a generated code amount and a quantization step for each layer of a previously encoded frame. An image coding method, wherein a quantization step of a highest layer in a frame which has been coded in the past is determined as a quantization step of a lowest layer of a frame to be coded. 5. The image encoding apparatus according to claim 1, wherein, when encoding a still image, calculating an optimal quantization step while encoding the same input signal a plurality of times. Applying the method and calculating the optimal quantization step of the lowest layer using the generated code amount and the quantization step of each layer when the encoding target still image was encoded immediately before; Encoding method. 6. A band to be encoded is divided into a high band and a low band in the horizontal and vertical directions, and a low band component in both the horizontal and vertical directions is similarly re-band-divided. Means for recursively specifying a specified number of times; means for quantizing transform coefficients belonging to a plurality of frequency bands obtained by the band division in a predetermined quantization step; and Means for encoding the transform coefficients in the high-frequency component in order from the coefficient as the lowest layer, and further quantizing the quantization error in a quantization step smaller than the quantization step, and Means for encoding as two layers, and until the generated code amount of the encoding target frame reaches a predetermined code amount, recursively, means for encoding while reducing the quantization error in each layer, Image mark In encoding a certain frame in a moving image input signal, the encoding apparatus calculates a code amount generated in each layer of a previously encoded frame, and records a quantization step used in each layer. Means for determining, and a means for determining a quantization step of each layer in the encoding target frame from the generated code amount, the quantization step, and a target code amount of the encoding target frame, Image encoding device. 7. The image encoding apparatus according to claim 6, wherein the encoding target frame is a still image, and in encoding a certain frame in the moving image input signal, each layer of the previously encoded frame is encoded. Wherein the means for calculating the amount of generated codes of the image coding apparatus is a means for calculating the amount of generated codes for each layer of a still image which has been coded in the past when coding a certain still image. . 8. Any one of claims 1 to 5
A storage medium storing an image encoding program, wherein a program for causing a computer to execute the procedure in the image encoding method described in the paragraph is stored in a storage medium readable by the computer.