JP4181699B2 - Encoding apparatus and method, and storage medium - Google Patents

Description

[0001]
BACKGROUND OF THE INVENTION
The present invention relates to an encoding apparatus and method, and a storage medium.
[0002]
[Prior art]
Images, particularly multi-valued images, contain a great deal of information, and there is a problem that the amount of data becomes enormous when storing and transmitting the images. For this reason, when storing and transmitting images, high-efficiency coding is used to reduce the amount of data by removing the redundancy of the images and allowing some degradation in image quality.
[0003]
As one of these high-efficiency encodings, JPEG compression encoding is known. In this method, a multi-value image is converted into frequency components by DCT conversion for each block, and the obtained conversion coefficient is quantized and variable-length encoded.
[0004]
The encoding using the DCT transform has a problem that block distortion occurs in the decoded image when the compression rate is set high. In recent years, a new encoding method using wavelet transform has been proposed in order to eliminate such distortion.
[0005]
Also, variable length coding is often applied as part of these various high efficiency codings.
[0006]
[Problems to be solved by the invention]
However, a method for variable-length coding a multi-valued image with simple processing and a processing method for executing this variable-length coding method at high speed have not been established yet.
[0007]
The present invention has been made in view of such a situation, and an object of the present invention is to provide an encoding apparatus and method, and a storage medium capable of variable-length encoding a multi-valued image easily and at high speed.
[0008]
[Means for Solving the Problems]
  In order to solve this problem, for example, the encoding of the present invention.MethodHas the following configuration. That is,
  A symbol group of a predetermined unit to be encoded isBased on a certain parameterThe code amount is determined depending on the contents of the symbol group and the parameters.A variable-length code part;The code amount is determined only by the parameters.With fixed-length code partTo the encoded data representedIn an encoding method for encoding,Each obtained by the code amount estimation process for the symbol group of the predetermined unit,ParametersInAgainstAboveBy evaluating the difference value of the code amount of the variable length code part,The minimum code amount for the symbol group of the predetermined unitIt is characterized in that an optimum value of the parameter is determined.
[0009]
Further, according to a preferred embodiment of the present invention, multi-value image data is divided into a plurality of information groups (a collection of quantized values for one sub-block in the present embodiment), and the information groups are individually separated. A coding apparatus for performing variable length coding (Golomb Rice coding) based on a parameter (k parameter), and how much the code amount of a variable length code portion differs between adjacent k parameters, and a difference value thereof The optimum parameter k for each information group is determined by sequentially comparing all the difference values obtained within the allowable range of the k parameter with a predetermined value.
[0010]
DETAILED DESCRIPTION OF THE INVENTION
Hereinafter, embodiments according to the present invention will be described in detail with reference to the accompanying drawings.
[0011]
  [First Embodiment]
  Hereinafter, an example of encoding 8-bit monochrome image data will be described as the first embodiment.,The present invention is not limited to this, and is applied to a case where a monochrome image represented by 4 bits for each pixel or a color multi-value image representing a color component (RGB / Lab / YCrCb) of each pixel by 8 bits is encoded. Is also possible. When applied to these, each color component may be encoded as a monochrome image.
[0012]
FIG. 1 is a block diagram of an image encoding device according to the first embodiment.
[0013]
In the figure, 101 is an image data input unit, 102 is a data storage unit, 103 is a discrete wavelet transform unit, 104 is a coefficient quantization unit, 105 is an encoding parameter selection unit, 106 is a Golomb Rice encoding unit, and 107 is a bit. A plane scanning unit 108 is a buffer, and 109 is a code output unit.
[0014]
First, the image input unit 101 inputs pixel data to be encoded in raster scan order. The image input unit 101 includes, for example, image data input processing from an imaging device such as a scanner or a digital camera, an imaging device such as a CCD, or a network line interface.
[0015]
Further, the image input target may be a recording medium such as a RAM, a ROM, a hard disk, and a CD-ROM.
[0016]
The image data storage unit 102 stores the image data input by the image input unit 101.
[0017]
The discrete wavelet transform unit 103 performs known discrete wavelet transform on the image data for one screen stored in the image data storage unit, and decomposes the image data into a plurality of frequency bands. In the present embodiment, it is assumed that the discrete wavelet transform for the image data string x (n) is performed by the following equation.
[0018]
r (n) = floor {(x (2n) + x (2n + 1)) / 2}
d (n) = x (2n + 2) -x (2n + 3) + floor {(-r (n) + r (n + 2) +2) / 4}
r (n) and d (n) are conversion coefficients, r (n) is a low frequency component, and d (n) is a high frequency component.
[0019]
In the above formula, floor {X} represents the maximum integer value not exceeding X. Although this conversion formula is for one-dimensional data, it is possible to perform two-dimensional conversion by applying this conversion in the order of the horizontal direction and the vertical direction, and the LL, It can be divided into four frequency bands (sub-blocks) of HL, LH, and HH. These converted data are stored in the data storage unit 102 in order to output the data smoothly to the next-stage coefficient quantization unit 104 or to perform further wavelet transform.
[0020]
The generated LL component is subjected to a discrete wavelet transform in the same procedure as described above to be decomposed into seven frequency bands (sub-blocks) as shown in FIG. In the present embodiment, the discrete wavelet transform is repeated once more, so that 10 pieces of LL, HL3, LH3, HH3, HL2, LH2, HH2, HL1, LH1, and HH1 are obtained as shown in FIG. Are divided into frequency bands (sub-blocks) and stored in the data storage unit 102.
[0021]
The transform coefficients are output from the data storage unit 102 to the coefficient quantization unit 104 in the order of subblocks LL, HL3, LH3, HH3, HL2, LH2, HH2, HL1, LH1, and HH1, and in the raster scan order for each subblock. .
[0022]
The coefficient quantization unit 104 quantizes each of the wavelet transform coefficients in a quantization step determined for each frequency component, and outputs the quantized value to the encoding parameter selection unit 105 and the buffer 108.
[0023]
When the coefficient value is X and the quantization step value for the frequency component to which this coefficient belongs is q, the quantized quantized value Q (X) is obtained by the following equation.
[0024]
Q (X) = floor {(X / q) +0.5}
FIG. 3 shows the relationship between each frequency component and the quantization step in the present embodiment. As shown in the figure, the quantization step is larger for the high frequency components (HL1, LH1, HH1) and the like than for the low frequency components (LL, etc.). As a result, the information of high frequency components that are hardly visually noticeable is reduced to improve the compression.
[0025]
The quantized value is evaluated by the encoding parameter selection unit 105, and stored in the buffer 108 until the k parameter required for encoding the quantized value is determined by the subsequent Golomb Rice encoding unit 106. Retained.
[0026]
The encoding parameter selection unit 105 selects the k parameter used in the latter-stage Golomb Rice encoding unit 106 based on the quantized value.
[0027]
The k parameter is a value indicating the code length of the fixed length part when performing variable length coding called Golomb Rice coding. For convenience of explanation, before explaining the coding parameter selection unit 105, the operation of the Golomb Rice coding unit 106 will be explained first.
[0028]
Although the basic method of encoding performed by the Golomb Rice encoding unit 106 is known, the basic operation of encoding and the characteristic part of the present invention will be described below.
[0029]
First, data for one sub-block is sequentially read from the buffer 108 that stores the quantization value.
[0030]
Next, the Golomb Rice encoding unit 106 performs preprocessing and Golomb Rice encoding processing described below based on the k parameter corresponding to the sub-block.
[0031]
The Golomb Rice encoding unit 106 checks the positive / negative of each input quantization value and outputs a negative sign bit. Specifically, “0” is output when the quantized value is 0 or positive, and “1” is output when the quantized value is negative, and then the quantized value is converted into an absolute value. (This process is not strictly included in the Golomb Rice encoding process, but here, as the pre-processing, it is performed in the Golomb Rice encoding unit 106)
Next, the absolute value is Golomb Rice encoded. The Golomb Rice encoding process when the absolute value of the quantization value to be encoded is V and the value of the k parameter applied to the sub-block to be processed is k is performed according to the following procedure.
[0032]
First, V in binary representation is shifted to the right by k bits, then “0” is continuously arranged for the number of obtained values (remaining values), and then “1” is arranged as a delimiter bit. Then, a variable length code (Golomb Rice code) is generated by arranging the lower k bits of the original V.
[0033]
FIG. 4 shows an example of Golomb Rice code for V = 0 to 7 and k = 0, 1, and 2. For example, when V = 4 and k = 1, 4 is a binary number “100B (B indicates a binary number)”, and when k (= 1) bits are shifted to the right, the rest is “10B”. Since “2”, “0” is continued twice, followed by “1” as a delimiter, and “0010” is obtained by the lower-order k (= 1) bit of the value 4 (= 100 B).
[0034]
As can be seen, the code length of each variable length code obtained from the absolute value V and the parameter k is V >> k (the value remaining when V is shifted to the right by k bits) +1 (separator bit) + k (parameter Value) bits. If a negative sign bit indicating positive / negative is added, it is further increased by one bit.
[0035]
As can be seen from the figure, when the k parameter = 0, the Golomb Rice code length for the quantized value “0” is particularly shorter than when k = 1,2. This indicates that since the quantized value group to be encoded is biased to 0, it is more appropriate to perform encoding with a smaller k parameter.
[0036]
Further, the Golomb Rice encoding used here has an effect that encoding and decoding can be performed by a simple calculation without actually holding a code table as shown in FIG. Normally, in variable-length encoding such as Huffman encoding, a table indicating variable-length codes for encoding target values must be held. In particular, when a plurality of variable length codes are appropriately switched for a value to be encoded according to a related state, it is necessary to have a plurality of the tables.
[0037]
As described above, encoded data composed of a negative bit (representing +/−) and a variable length code (Golomb Rice code) for the input quantized value is generated and output to the bit plane scanning unit 107 at the subsequent stage. To do.
[0038]
Next, processing contents of the encoding parameter selection unit 150 that generates k parameters required by the Golomb Rice encoding unit 106 described above will be described.
[0039]
The total code amount T of the sub-block to be encoded is given by assuming that the value of the k parameter is p.
T_p= Σ (Vi >> p (value obtained by shifting Vi to the right by p bits) +2 (negative bit and separator bit) + p) (1)
It becomes.
[0040]
Assuming that the number of quantization values in the sub-block to be encoded is N, the above equation (1) is as follows.
[0041]
T_p= (P + 2) × N + ΣVi >> p (2)
In the above equation (2), the first term: (p + 2) × N is the code amount by the fixed-length code portion, and the second term: ΣVi >> p is the code amount by the variable-length code portion. Here only the second term
S_p= ΣVi >> p (3)
It expresses. This S_pIncreases as the value of p decreases. On the other hand, the code amount of the fixed-length code portion increases in proportion to the value of p, and when p increases by one, the code amount increases by N bits.
[0042]
Next, the number of “1” s in each bit plane when Vi is expressed in binary will be considered.
[0043]
The number of “1” present in the plane of the least significant bit of Vi >> p is B_{p + 1}And That is,
B_{p + 1}= Σ (Vi >> p) & 1 (4)
It is. Where B₁Represents the number of “1” s in the least significant bit plane of Vi, and B₂Represents the number of “1” s in the second plane from the least significant bit of Vi.
[0044]
ΣVi >> p = (2 × ΣVi >> (p + 1)) + (Σ (Vi >> p) & 1)
From the relationship of (3) and (4)
S_p= 2 x S_{p + 1}+ B_{p + 1}(5)
Is obtained. k parameter values differ by one_p-1And S_pThe difference between D_pThen,
D_p= S_p-1-S_p= (2 x S_p+ B_p-S_p
= S_p+ B_p(P ≧ 1) (6)
This D_pRepresents how much the variable-length code amount is increased by changing the value of the k parameter from p to p−1. If this value is smaller than N, the total code amount is ND_pTherefore, the value of the k parameter should be p-1 rather than p. Conversely, D_pIf ≧ N, the value of the k parameter should be p.
[0045]
Similarly, two different k parameter values_p-2And S_pDD difference between_pThat is, DD_p= S_p-2-S_pIf this value is smaller than 2N, the variable length code amount becomes 2N-DD by changing the value of the k parameter from p to p-2._pTherefore, the value of the k parameter should be p-2 rather than p. Conversely, DD_pIf ≧ 2N, it can be said that p is better for the value of the k parameter.
[0046]
When the above equation (6) is further modified,
D_p-1= S_p-1+ B_p-1
= (2 x S_p+ B_p) + B_p-1
= (S_p+ B_p) + S_p+ B_p-1
= D_p+ S_p+ B_p-1
S_p, B_p-1≧ 0
D_p-1≧ D_p(7)
The following can be said from the relationship of the above equation (7) and the above description.
[0047]
[Sequence: D₁, D₂, D_Three, ..., D_p-1, D_p, D_{p + 1}, ..., D_m, D1 ≧ D2 ≧ D3 ≧ ··· ≧ Dp-1 ≧ Dp ≧ Dp + 1 ≧ ·· ≧ Dm,_q≧ N ≧ D_{q + 1}Q is one of the optimal values of the k parameter. ]
The optimum value here means that the entire code amount when Golomb Rice encoding is minimized. M is not zero B_pIs the largest value of p, and S corresponding to this m_mIs zero. (Because the largest quantized absolute value is shifted to the right by m bits to 0)_m= B_m≦ N.
[0048]
When the k parameter is changed from q to q + 1, the code amount is ND_{q + 1}When the k parameter is changed from q to q-1, the code amount is D_qNOnly increase. However, D_{q + 1}= N or D_qWhen N = N, the code amount does not increase and remains the minimum. D_{q + 1}When N = N, q + 1 is also the optimum value, and D_qWhen N = N, q-1 is also an optimum value.
[0049]
There can be a maximum of three optimum values. The condition is D_q= N = D_{q + 1}In other words, S_q-1= 0, S_q= N, S_{q + 1}= 2N. At this time, q-1, q and q + 1 are optimum values, and the total code amount T takes an extreme value (minimum value). Therefore, the total code amount increases as the k parameter moves away from the optimum value.
[0050]
On the other hand, when q does not exist, N> D₁In this case, the k parameter value is 0 and the total code amount is minimum. The total code amount monotonically increases with respect to the change (increase) in the value of k.
[0051]
Here, in order to generalize the relationship between the sequence and the optimal value, D₀= D₁It is defined as + N. Then, "Number sequence: D₀, D₁, D₂, D_Three, ..., D_p-1, D_p, D_{p + 1}, ..., D_mD₀≧ D₁≧ D₂≧ D_Three≧ ・ ・ ≧ D_p-1≧ D_p≧ D_{p + 1}≧ ・ ・ ≧ D_mD_q≧ N ≧ D_{q + 1}Q exists, and this q is one of the optimum values of the k parameter. The existence of the optimum value q became clear.
[0052]
From the above description, it can be seen that the local minimum value of the total code amount Tk with respect to the change of the k parameter is also the global minimum value in the entire range of the k parameter. Therefore, the search for the optimum value of the k parameter does not need to evaluate whether or not the total code amount Tk is the minimum for all the values of k, and it is only necessary to find the value of k where Tk is the minimum value locally. It can be said. This simplifies the k parameter search method.
[0053]
From the various relationships described above, several methods (embodiments of the invention) for determining the optimum value of the k parameter are conceivable. An example of the first embodiment is shown in the flowchart of FIG.
[0054]
First, in step S501, the code amount S of the variable length code portion._p(P = 0, 1, 2,..., M−1), and in step S501a, S is not zero._p1 is added to the maximum value of p to obtain m.
[0055]
In step S502, m is set to i. In step 503, D is set._i= S_i-1-S_iCalculate
[0056]
In step S504, D_iCompare N with D_iIf <N, go to step S505, otherwise go to step S507.
[0057]
In step 505, i = i-1 is calculated and the candidate value of the k parameter is updated. In step 506, it is determined whether or not the updated value of i is 0. If i = 0, the process goes to step 507, and if i = 0, the process goes to step S503 to continue the loop processing.
[0058]
In step S507, the value of the k parameter is determined as i.
[0059]
Code amount S by variable length code portion in step S501_pFIG. 6 shows an example of the calculation program (described in C language). In this example, the addition is performed at least once even when V = 0, but the do to While statement may be executed only when V ≠ 0 is detected in the if statement.
[0060]
When the processing of FIG. 6 is performed for all (N) Vs in the sub-block, the upper limit candidate value m of the k parameter is obtained in the next step S501a. This value is obtained by calculating in advance the theoretical maximum value of the variable p in FIG. 6 from the theoretical maximum value of V and sequentially evaluating the value of S [].
[0061]
In step S502, m is set to i.
[0062]
Next, in step S503, D_m= S_m-1-S_mCalculate This D_mIs all D_iIt is the smallest among them, N at the maximum, and cannot be larger than N. This represents the number of “1” s in the bit plane of the most significant bit (non-zero bit) of the largest value V.
[0063]
Therefore, in step S504, D_{i (= m)}<N is determined, and the process proceeds to step S505.
[0064]
In step S505, the value of i is decremented by 1 from m to m−1. In the next step S506, it is determined whether i is 0 or not. Here, i ≠ 0 and the process returns to step S503.
[0065]
The above steps S503 to S506 are repeated, and D_iIf> N or i = 0, the code amount does not decrease even if the value of the k parameter is changed. Therefore, the process goes to step S507 and the value of the k parameter is determined as i.
[0066]
As described above, the processing is performed by the encoding parameter selection unit in FIG. 1, and the above-described Golomb Rice encoding processing unit 106 is executed based on the parameter. The encoded data is output to the bit plane scanning unit 107 at the next stage.
[0067]
The bit plane scanning unit 107 receives encoded data encoded and output for each quantized value, rearranges them in bit plane units, and outputs the rearranged data. This is performed for each frequency component (subblock) described above.
[0068]
First, the encoded data generated by the Golomb Rice encoding unit is separated into a fixed length code portion and a variable length code portion determined by the value of the k parameter. In addition, the sign has a fixed negative bit of 1 bit.
[0069]
The scanning order is from important information. First, the negative bit is scanned for all quantized values in the sub-block. Next, the variable length code portion having the information of the upper bits of the quantized value is scanned in order from the upper bit plane. Finally, the fixed-length code portion is scanned in order from the upper bit plane. FIG. 7 shows the Golomb Rice code for each quantized value (however, the variable length code portion and the fixed length code portion are separated) and the scanning order.
[0070]
A little more detailed explanation is added about the scanning of the variable length code part. As the name of the variable-length code is not constant, it is not uniquely determined how to align the short code and the long code, but here the first bit of the variable-length code is on the same plane. It shall be arranged as follows. That is, the variable length codes are arranged in order from the top of the variable length code area in FIG. When the bit plane information is scanned in order from the upper bit, the information (code bit) is lost from the plane in the middle of the short code, so the code without information skips scanning. Therefore, the information to be scanned (code bits) decreases as the lower bit plane is reached.
[0071]
FIG. 8 shows an arrangement of bit information scanned in the above order (negative sign, variable length code, fixed length code).
[0072]
This bit plane scanning is performed in the order of LL, HL3, LH3, HH3, HL2, LH2, HH2, HL1, LH1, and HH1 in units of sub-blocks.
[0073]
The bit string output from the bit plane scanning unit is sequentially transferred by the next code output unit 109. The transfer partner may be a recording medium such as a hard disk or a DVD, or an interface such as the Internet, a general public line, or a wireless line.
[0074]
The encoded data generated in the above embodiment includes various information necessary for decoding, that is, the image size, the number of bits per pixel, the quantization step for each frequency component, the value of the k parameter, and the like. Appropriately added to the encoded data as attached information.
[0075]
[Second Embodiment]
A method slightly different from the coding parameter selection method in the first embodiment, that is, the k parameter generation method will be described with reference to FIG. Note that the same steps as those in FIG. 5 are added to the processing steps having the same role.
[0076]
In the second embodiment, the formula (6): D_p= S_p+ B_pIs used. Therefore, processing is performed in step S903 instead of step S501 and step S903 instead of step S503.
[0077]
In step S901, the code amount S by the variable-length code unit._p(P = 0, 1, 2,..., M−1) and the number B of “1” in each bit plane B_p(P = 1, 2,..., M) is calculated. In step S903, D is calculated._i= S_i+ B_iCalculate
[0078]
S in step S901_pAnd B_pFIG. 10 shows an example of the calculation program (described in C language). Cumulative addition processing is performed on S0 and B1 in the first round of the do-while loop statement in the program, and cumulative addition processing is performed on Sp-1 and Bp in the pth round.
[0079]
The greatest feature of the second embodiment is D_iIs calculated by the equation (6). S required for that_iAnd B_iIs obtained in advance in step S901, and D is obtained in step S903._iCalculate Since other processes are the same as those in the first embodiment, description thereof will be omitted.
[0080]
[Third Embodiment]
In the third embodiment, the formula (5): S_p= 2 x S_p+ 1 + B_{p + 1}Use the relationship. That is, in the process of step S901 in the second embodiment, S_pAnd B_pThe part for calculating (p = 0, 1, 2,..., M) is replaced with the processing in steps S911 to S913 shown below.
[0081]
In step S911, B_p(P = 0, 1, 2,..., M) is calculated.
[0082]
In step S912, non-zero B_pThe maximum value m of p is obtained.
[0083]
In step S913, S_m= 0 and S_p-1= 2 x S_p+ B_p(P = m, m−1, m−2,..., 1) is calculated.
[0084]
The other processes are the same as those in the second embodiment, and a description thereof will be omitted. The process of step S913 is referred to as “S_m-1= B_mInitialized to S_p-1= 2 x S_p+ B_p(P = m−1, m−2,..., 1) is calculated. May be changed.
[0085]
[Fourth Embodiment]
In the fourth embodiment, the relationship of the above expression (6): D_p= S_p+ B_pCan be easily derived from D_p≦ S_pThis relationship is used. D_p≦ S_pS_p<N, D_pIt can be said that <N.
[0086]
Therefore, as shown in the flowchart of FIG._iStep S1101 for checking whether or not the relationship <N is satisfied is provided before step S903 in FIG. 9 showing the second embodiment, and if the relationship is satisfied, step S903 and step S504 are passed.
[0087]
Note that this embodiment can also be applied to the first embodiment. In that case, step S1101 is provided before step S503 in FIG. 5, and if this relationship is established, step S503 and step S504 are passed.
[0088]
[Fifth Embodiment]
Another method different from the encoding parameter selection method in the first and second embodiments will be described.
[0089]
In this embodiment, the difference between Sp-2 and Sp having two different k parameter values, that is, DD._p= S_p-2-S_pIs used.
[0090]
As already explained, DD_pIs smaller than 2N, the variable length code amount becomes 2N-DD by changing the value of the k parameter from p to p-2._pTherefore, the value of the k parameter should be p-2 rather than p. Conversely, DD_pIf ≧ 2N, it can be said that p is better for the value of the k parameter.
[0091]
Therefore, the DD_pCan be used to update k-parameter candidate values at high speed by 2 units. A flowchart showing the processing method of this embodiment is shown in FIG. The same steps as those in FIG. 5 are added to the processing steps having the same role.
[0092]
In the present embodiment, steps S1201 to S1004 for performing another loop process are provided between step S502 and step S503 in FIG. 5 in the first embodiment.
[0093]
In step S1201, DD, which is the most characteristic feature of the present embodiment._i= S_i-2-S_iCalculate In step S1202, DD_iAnd 2N are compared. In step S1203, i is updated by 2 units. In step S1204, it is determined whether i is 1 or less. The loop process in steps S1201 to S1204 is very similar to the loop process in steps S503 to S506 in the first embodiment.
[0094]
The difference is that the unit for updating i is 2. DD calculated in step S1201_iIs compared with 2N in step S1202, and it is determined whether or not two i can be updated. If it can be updated, the value of i is updated by i = i−2 in step S1203.
[0095]
If i ≦ 1, step S1201 cannot be executed._-1If i ≦ 1 is detected in step S1204, the process proceeds to the next stage (step S503). If i> 1, the process returns to step S1201 to continue the loop process.
[0096]
When the loop processing is exited, the code amount does not decrease even if the k parameter is changed from i to i-2. However, if the k parameter is changed from i to i-1, the code amount may be reduced.
[0097]
Therefore, in order to check whether or not the k parameter can be changed from i to i-1, D is determined in step S503._i= S_i-1-S_iAnd D in step S504_iCompare N with D_iIf <N, i can be changed from i to i-1, so that the process proceeds to step S505. Otherwise, i is continued as it is, and the process proceeds to step S507 to set the value of i as the value of the k parameter.
[0098]
In this embodiment, i is first updated by 2 units._i-4-S_iCan be updated in 4 units, and can be updated in another unit using other calculation formulas.
[0099]
When updating by 4 units, there are four possibilities of i to i-3 as the optimum k parameters when the first loop processing is exited. Therefore, in order to specify one of these four values in the next stage, it is possible to consider a processing method in which whether or not it can be updated in one unit is determined by a loop process at most three times and the value of i is updated.
[0100]
[Sixth Embodiment]
As a coding parameter selection method, a further different method is shown in FIG. 13 and described. Note that the same steps as those in FIG. 5 are added to the processing steps having the same role.
[0101]
In this embodiment, D_iWithout using SN_p= S_pIt is characterized by using a value of + N (p = 0, 1, 2, 3,..., M).
[0102]
In step S1301, the SN_pIn step S1302, S is calculated._i-1And SN_iCompare
[0103]
Other processes are the same as those in the first embodiment.
[0104]
S calculated in step S501_pA value SN obtained by adding N to the values (p = 0, 1, 2, 3,..., M)_pIs calculated in step S1301, and i is initialized in step S502.
[0105]
Thereafter, in order to check whether or not i, which is a candidate value of the k parameter, can be updated to i-1, in step S1302, S_i-1And SN_iCompare S_i-1<SN_iIf so, the code amount is reduced, so that the process goes to step S506 to update i to i-1. Otherwise, the process returns to step S504, and the value of the k parameter is fixed to i.
[0106]
After i is changed to i-1, it is determined in step S506 whether i = 0. When i = 0, the value of i cannot be updated any more, so the loop process is exited and the process goes to step 504. Otherwise, the process returns to step S1302 and i update processing is continued.
[0107]
[Seventh Embodiment]
In all of the embodiments described so far, the initial value of the loop control variable i is set to m. In the present embodiment, the initial value of the control variable i is set to 0.
[0108]
Basically, the present embodiment is applicable to all the embodiments described so far. In particular, the present embodiment will be described based on FIG. 5 which is the first embodiment. This embodiment in which the initial value of i is 0 is shown in FIG.
[0109]
In the figure, in step S1402, an initial value 0 is set to i, and in step S1404, D_iCompare N with D_iIf> N, go to step S1405, otherwise go to step S1407. In step S1405, 1 is added to i, and the process returns to step S503.
[0110]
In step S1407, the value of the k parameter is determined to be i-1.
[0111]
In the present embodiment, Steps S501a and S506 for determining m, which are necessary in the previous embodiments, are no longer necessary.
[0112]
D as mentioned before_mSince ≦ N, the value of i increases to m or before that,_iSince ≦ N, the branch destination in step S1404 is always step S1407, and the value of k is determined here.
[0113]
The reason why the value of the k parameter is set to i-1 instead of i in step S1407 is that the relationship at the time of branching from step S1404 to step S1407 is D_i-1> N ≧ D_iBecause. As described above, the optimal value of the k parameter in this relationship is i-1.
[0114]
The present embodiment can take the simplest processing form so far.
[0115]
[Eighth Embodiment]
In the present embodiment, the initial value of the control variable i is changed depending on the result of the comparison processing performed before that.
[0116]
The optimum value of the k parameter may exist near i = m or may exist near i = 0. It is reasonable to search from i = m for the former and from i = 0 for the latter.
[0117]
The determination must first be made before entering the loop process. D for h = floor {m / 2}_hAnd D_hAnd N are compared. D_hIf <N, the process is performed with i set to 0; otherwise, the process is performed with i set to m. This is the outline of this embodiment.
[0118]
This embodiment is shown in FIG. 15 and will be described in more detail. In the figure, in step S1502, h = floor {m / 2} is calculated. In step S1503, D corresponding to the above h._hCalculate
[0119]
In step 1504, D_hAnd N, and D_hIf <N, the process branches to step S1402. Otherwise, the process branches to step S502.
[0120]
The other processing steps are the same as the same number processing steps in FIG. 5 or FIG.
[0121]
The processing after step S1402 is exactly the same as that of the seventh embodiment, but the processing after step S502 is basically the same as that of the first embodiment, but there are some differences. The difference is that there is no determination (corresponding to step S506) whether i has reached a predetermined value in the loop processing.
[0122]
The reason why there is no determination is that when the value of i is decreased by 1 from the initial value m and reaches the above h in the loop processing, the determination at step S504 exits the loop processing and branches to step S507 without fail. Because I know. This is obvious from the condition (Dh ≧ N) when branching from step S1504 to step S502.
[0123]
Similarly, the upper limit of the value of i when exiting the loop processing in steps S503, S1404, and S1405 is h. This can be seen from the condition (Dh <N) when branching from step S1504 to step S1402.
[0124]
In the present embodiment, the upper limit number of loop processes until the determination of the k parameter can be halved, so that the k parameter can be determined at a higher speed.
[0125]
Further, as an application of this embodiment, if the number of branches to the loop processing for determining the k parameter is increased, the upper limit number of loops of each loop processing can be further reduced, and the k parameter can be determined at a higher speed.
[0126]
[Ninth Embodiment]
In the initial value setting method as in the ninth embodiment, it is necessary to prepare two types of loop processing when the control variable i increases and loop processing when the control variable i decreases. Therefore, a processing form when the two loop processes are shared will be described with reference to FIG.
[0127]
In this embodiment, in step S1601, the initial value of i is set to h. In step 1602, it is determined whether i is equal to h. In step 1603, the initial value of i is reset to 0, and h is set to -1.
[0128]
The other processing steps are the same as those in the eighth embodiment: the same number processing step in FIG.
[0129]
H is calculated by the same calculation as in the eighth embodiment (step S1502), and in step S1601, the h is set to i. Next, the process proceeds to step S503, where D_h= S_h-1-S_hCalculate In step S1404, D_hIs greater than N. The processing so far is the same as in the eighth embodiment, although the structure of the flowchart is different.
[0130]
D_hIf> N, i = h is set as an initial value, and the same processing as the loop processing when the control variable i increases is performed in steps S503, S1404, and S1405. Since the value of i when exiting the loop process passes through step S1405 (i = i + 1 is calculated) at least once, the value of i is larger than h, and through the determination of step S1602, k is determined in step S1407. The parameter value is determined as i-1.
[0131]
D_hIf not> N, the process branches from the first time to step S1602. Since i = h at this time, the process branches to step S1603, the initial value of i is reset to 0, and h is set to -1. The reason for setting h to −1 is to set a value that i cannot take so that the determination in step 1602 for the second time is not caught (i = h is not satisfied).
[0132]
The processing after resetting initial values of i to 0 is completely the same as the loop processing in steps S503, S1404, and S1405 in the eighth embodiment, and the value of i when the loop processing is exited is the maximum. floor {m / 2}.
[0133]
Since the value of the variable h is set to -1 in step S1603, i = h is never satisfied in step S1602. Therefore, the process branches to step S1407, and the value of the k parameter is determined to be i-1. The
[0134]
Compared to the eighth embodiment, the overhead due to the common processing is that the determination processing in step S1602 required when resetting the initial value of i to 0 is 1/2 (initialization) Sometimes the probability of performing this determination process is about 1/2) and once when exiting the loop process (because this determination process cannot be avoided), the processing routine is compact. It seems that the merit of is greater.
[0135]
In addition, although the structure in each said embodiment shall be based on FIG. 1, the process of the same figure is realizable also with a program, for example. Since a device for inputting an image may be a network, a video camera, an image scanner, or a storage device (floppy, etc.), the present invention is, for example, a system or a general-purpose information processing apparatus (personal computer, etc.) having any of these. It can also be applied to.
[0136]
Accordingly, a storage medium (or recording medium) that records the program code of the software that implements the functions of the above-described embodiments is supplied to the system or apparatus, and the computer (or CPU or MPU) of the system or apparatus stores the storage medium. Needless to say, this can also be achieved by reading and executing the stored program code. In this case, the program code itself read from the storage medium realizes the functions of the above-described embodiments, and the storage medium storing the program code constitutes the present invention. Further, by executing the program code read by the computer, not only the functions of the above-described embodiments are realized, but also an operating system (OS) running on the computer based on the instruction of the program code. It goes without saying that a case where the function of the above-described embodiment is realized by performing part or all of the actual processing and the processing is included.
[0137]
Furthermore, after the program code read from the storage medium is written into a memory provided in a function expansion card inserted into the computer or a function expansion unit connected to the computer, the function is determined based on the instruction of the program code. It goes without saying that the CPU or the like provided in the expansion card or the function expansion unit performs part or all of the actual processing and the functions of the above-described embodiments are realized by the processing.
[0138]
As described above, according to the present embodiment, when a symbol is encoded with a variable-length code part and a fixed-length code part based on a certain parameter, the code amount of the variable-length code part for each value of the parameter It is now possible to evaluate only the difference value and determine the optimal parameter at high speed.
[0139]
【The invention's effect】
As described above, according to the present invention, it is possible to perform variable-length encoding of a multilevel image easily and at high speed.
[Brief description of the drawings]
FIG. 1 is a block configuration diagram of an apparatus according to an embodiment.
FIG. 2 is a diagram illustrating a state where a discrete wavelet transform is performed.
FIG. 3 is a diagram illustrating quantization steps used for frequency components (sub-blocks).
FIG. 4 is a diagram illustrating an example of a code obtained by performing Golomb Rice encoding.
FIG. 5 is a flowchart showing processing contents according to the first embodiment;
FIG. 6 is a diagram illustrating an example of a program that calculates a code amount Sp of a variable length code for each value of a k parameter.
FIG. 7 is a diagram representing a Golomb Rice encoded code in a bit plane.
FIG. 8 is a diagram illustrating a state of a bitstream that is finally output.
FIG. 9 is a flowchart showing processing contents according to the second embodiment;
FIG. 10 is a diagram illustrating a program for calculating both the code amount Sp and the number Bp of 1 in each bit plane.
FIG. 11 is a flowchart illustrating processing contents according to a fourth embodiment.
FIG. 12 is a flowchart illustrating processing contents according to a fifth embodiment.
FIG. 13 is a flowchart showing processing contents according to the sixth embodiment.
FIG. 14 is a flowchart showing processing contents according to a seventh embodiment;
FIG. 15 is a flowchart showing processing contents according to an eighth embodiment.
FIG. 16 is a flowchart showing processing contents according to the ninth embodiment;
[Explanation of symbols]
101 Image input unit
102 Data storage unit
103 Discrete wavelet transform unit
104 Coefficient quantization unit
105 Coding parameter selection unit
106 Golomb Rice Encoding Unit
107 bit plane scanning unit
108 buffers
109 Code output section

Claims (13)

A symbol group of a predetermined unit to be encoded is determined based on a certain parameter , a variable length code portion whose code amount is determined depending on the contents of the symbol group and the parameter, and a code amount is determined only based on the parameter. in the coding method for coding the coded data represented by a fixed length code part, the resulting code amount estimation processing for symbol groups of a predetermined unit, the difference value of the code amount of the variable length code portion against the respective parameters And determining an optimum value of the parameter that is a minimum code amount for the symbol group of the predetermined unit .   The encoding method according to claim 1, wherein the difference value of the code amount of the variable-length code portion is a difference value when the parameter value is different by 1 or when the parameter value is different by 2 or more.   2. The encoding method according to claim 1, wherein when the difference value of the code amount of the variable length code portion is calculated, information indicating the number of 1 in each bit plane of the symbol group to be encoded is used.   The encoding method according to claim 1, wherein a difference value sequence in which difference values of code amounts of the variable-length code portions are arranged in ascending or descending order of parameter values is evaluated from the end.   The subsequent evaluation order is determined based on a result of evaluating a value in the middle of a difference value sequence in which the difference value of the code amount of the variable length code portion is arranged in ascending or descending order of parameter values. The encoding method according to claim 1. A symbol group of a predetermined unit to be encoded is determined based on a certain parameter , a variable length code portion whose code amount is determined depending on the contents of the symbol group and the parameter, and a code amount is determined only based on the parameter. In an encoding method for encoding encoded data represented by a fixed-length code portion, the code amount of a variable-length code portion obtained by code amount estimation processing for the symbol unit of the predetermined unit depends on the size of the predetermined unit. The optimum value of the parameter that is the minimum code amount for the symbol group of the predetermined unit is determined by comparing the value obtained by adding the predetermined value determined in this way with the code amount of the variable length code portion according to the adjacent parameter. An encoding method characterized by the above. Based on a certain parameter , a symbol group of a predetermined unit to be encoded, a variable-length code portion determined depending on the content of the symbol and the parameter, and a fixed-length code whose code amount is determined only based on the parameter in the encoding apparatus for encoding in the encoded data represented by a portion obtained by the code amount estimation processing for symbol group of the predetermined unit, calculates a difference value of the code amount of the variable length code portion against the respective parameters Means for comparing the difference value with a predetermined value determined depending on the size of the predetermined unit, and determining an optimum value of the parameter that is a minimum code amount for the symbol group of the predetermined unit based on the comparison result And a coding device.   8. The encoding apparatus according to claim 7, wherein the difference value of the code amount of the variable length code portion is a difference value when the parameter value is different by 1 or by two or more.   8. The encoding according to claim 7, wherein the means for calculating the difference value of the code amount of the variable length code portion further includes means for counting the number of 1 in each bit plane of the symbol group to be encoded. apparatus.   8. The encoding according to claim 7, further comprising means for sequentially comparing a difference value sequence in which the difference value of the code amount of the variable length code portion is arranged in ascending or descending order of the parameter value with a predetermined value from the end. apparatus.   Means for comparing the difference values of the code amount of the variable length code portion in the ascending order or descending order of the parameter values with a predetermined value, and controlling the subsequent comparison order based on the comparison result The encoding apparatus according to claim 7, further comprising means for performing comparison and means for performing comparison under the control means. A symbol group of a predetermined unit to be encoded is determined based on a certain parameter , a variable length code portion whose code amount is determined depending on the contents of the symbol group and the parameter, and a code amount is determined only based on the parameter. In the encoding device that encodes the encoded data represented by the fixed-length code portion, the code amount of the variable-length code portion obtained by the code amount estimation process for the symbol group of the predetermined unit is calculated based on each parameter. Means for adding a predetermined value determined depending on the size of the predetermined unit to the code amount obtained by calculation for each parameter , and the value obtained by the addition according to the adjacent parameter comparing means for comparing the code amount of the variable length code portion, based on the comparison result, to determine the optimum value of the parameter having the minimum code amount for symbol group of said predetermined unit Encoding apparatus characterized by a means. A computer-readable storage medium storing a computer program for causing a computer to function as the encoding device according to claim 6 or 12 .

Images