KR101612281B1 - Binary data compression and restoration method and apparatus - Google Patents

Abstract

The present invention relates to a method and an apparatus for compressing and restoring binary data to rapidly and efficiently compress and restore the binary data through simple operation and hardware configuration. A binary data compression method includes the steps of: scanning, by a compression unit, raw binary data in a first direction by n-bit unit and obtaining cyclic blocks having n-bit symbols, wherein each cyclic block is a binary block having the symbols encountered until all different 2^n types of symbols appear once or more when the raw binary data is sequentially scanned in the first direction; obtaining, by the compression unit, the cyclic groups respectively including one or more cyclic blocks based on a distribution chart of frequencies of respective symbols appearing in one or more cyclic blocks, or neighboring cyclic symbols; compressing, by the compression unit, the respective cyclic groups with a Huffman coding process for the symbols in the respective cyclic groups and then generating compressed cyclic groups; generating, by the compression unit, compressed data by combining the compressed cyclic groups.

Description

TECHNICAL FIELD [0001] The present invention relates to a binary data compression and restoration method and apparatus,

The present invention relates to a method and apparatus for compressing and restoring binary data, and more particularly, to an apparatus and method for efficiently and efficiently compressing and restoring binary data through a simple operation and a hardware configuration, And more particularly to a method and apparatus for compressing and restoring binary data.

In general, since the frequency bandwidth available in a normal transmission channel is limited, various transmission systems such as a modem have used an effective data compression technique to compress or reduce the amount of transmission data in order to transmit a large amount of data.

One of the various compression schemes is the CCITT V.42 bis employed in a data transmission system such as a modem with a coding algorithm standardized by the International Telecommunication Union (ITU). The basis applied to this coding standard is a Ziv-Lempel code (ZLC). In this method, an address value of a dictionary storing the same phrase as the previous input data is formed as a codeword while adaptively forming a dictionary from the input data. Lt; / RTI > The dictionary operation performs a continuous string matching with the input data to update the dictionary by adding the unmatched characters to the maximum matching string and adding them to the dictionary.

However, such a conventional compression method requires complicated processing of data compression and decompression, requires a relatively high-performance hardware device, limits the improvement of the processing speed, and increases the reliability of the compression result value there was.

The background art of the present invention is disclosed in Korean Patent Laid-Open Publication No. 1999-0022960 (published on Mar. 25, 1999).

SUMMARY OF THE INVENTION The present invention has been made in view of the above problems, and it is an object of the present invention to provide a data compression method and a data compression method that can compress and restore binary data quickly and efficiently through simple computation and hardware configuration, And a method and apparatus for compressing and restoring binary data that can improve transmission efficiency and speed. And more particularly to enhance the compression ratio of the Huffman coding scheme more effectively.

According to one aspect of the present invention, there is provided a method of compressing binary data in a compression apparatus, the compression unit comprising: a first step of scanning the original binary data in units of n bits in a first direction to generate a plurality of cyclic blocks each having a plurality of symbols each having n bits Wherein each of said circular blocks is a binary block having symbols that meet until at least one or more of the 2 ^{< n >} different symbols of the original binary data are sequentially scanned in the first direction, Block acquisition step; Obtaining a plurality of cyclic groups each including at least one cyclic block based on the scattering of the frequencies at which each of the symbols appears within one cyclic block or a plurality of neighboring cyclic blocks; The compression unit compresses each of the cyclic groups by performing Houghman coding on the symbols in the respective cyclic groups to generate a plurality of compression cyclic groups; And compressing the compressed data by combining the plurality of compression cyclic groups to generate compressed data.

In the present invention, the step of acquiring the cyclic group may include calculating a first scatterplot of the frequencies at which the symbols appear in the current cyclic block; Comparing the first scatter diagram with a reference scatter diagram; And obtaining the current circulating block as the circulating group if the first scattering degree is equal to or greater than the reference scattering degree.

In the present invention, if the first scattering degree is less than the reference scattering degree, a second scattering degree of the frequencies in which the symbols appear in the current circular block and at least one subsequent circular block is calculated, and if the second scattering degree is smaller than the reference scattering degree The method further comprising: acquiring the current circular block and the at least one next circular block as the circular group, wherein the number of the at least one next circular block is such that the second scatter is greater than or equal to the reference scatter And the minimum number is.

In the present invention, an index of whether the distribution of variance, standard deviation, skewness or kurtosis of the frequency numbers is employed, or whether at least some kinds of symbols occupy more than a reference rate of the total frequency can be employed.

In the present invention, for the binary number of the remaining group remaining after the plurality of cyclic group acquisition steps, the compression unit performs Huffman coding based on the frequencies at which the symbols appear in the remaining group to obtain the compressed residual group And further comprising:

In the present invention, the step of generating the plurality of compression cyclic groups may include generating Huffman codes for each symbol according to the frequency of each symbol in each cyclic group; And replacing each symbol in each of the circulation groups with the corresponding Huffman code.

In the present invention, the step of generating the plurality of compression cyclic groups may include a Huffman code sequence generated by concatenating the Huffman codes corresponding to the respective symbols in a row, and a bit length of each of the Huffman codes in the Huffman code sequence And generating a coded dictionary corresponding to the coded word.

In the present invention, the step of generating the plurality of compressed cyclic groups may further include generating an encoded dictionary including information on the frequency of occurrence of each symbol in each cyclic group.

According to another aspect of the present invention, there is provided a method for restoring binary data compressed by a binary data compression method, the restoration unit comprising: a decompression unit that refers to the encoding dictionary for each cyclic group, And restoring the group into a plurality of cyclic groups to restore the binary data.

According to still another aspect of the present invention, the present invention provides a method for scanning original binary data in units of n bits in a first direction to obtain a plurality of circular blocks having a plurality of symbols of n bits length, Obtaining a plurality of cyclic groups each including at least one cyclic block based on a scattering degree of frequencies in which each of the symbols appears in a plurality of neighboring cyclic blocks, and performing a Huffman coding for the symbols in each cyclic group And a compression unit for compressing each of the cyclic groups to generate a plurality of compression cyclic groups and generating compressed data by combining the plurality of compression cyclic groups, wherein each of the cyclic blocks includes: until all of the receive more than 2 ⁿ different types of the symbol is at least one when sequentially scanned in a first direction It provides a binary data compression apparatus characterized in that the binary block having symbols meet.

In the present invention, at the time of acquiring the circulation group, the compression unit calculates a first scattering degree of the frequencies in which the symbols appear in the current circulation block, compares the first scattering degree with a reference scattering degree, The current circulating block is acquired as the circulating group.

In the present invention, when the first scattering degree is less than the reference scattering degree, the compression unit calculates a second scattering degree of the frequencies in which the symbols appear in the current circular block and at least one next circular block, Wherein if the reference scatter is greater than or equal to the reference scatter, the current circular block and the at least one next circular block are obtained as the circular group, and the number of the at least one next circular block is equal to or greater than the minimum number .

In the present invention, an index of whether the distribution of variance, standard deviation, skewness or kurtosis of the frequency numbers is employed, or whether at least some kinds of symbols occupy more than a reference rate of the total frequency can be employed.

In the present invention, with respect to the binary numbers of the remaining groups remaining after acquiring the plurality of cyclic groups, the compression unit may perform Huffman coding based on the frequencies at which the symbols appear in the remaining group to obtain further compressed residual groups .

In the present invention, when generating the plurality of compression cyclic groups, the compression unit generates Huffman codes for each symbol according to the frequency of each symbol in each cyclic group, and outputs each symbol in each cyclic group to the corresponding Huffman Code. ≪ / RTI >

In the present invention, the compression unit may further generate a coding dictionary including a Huffman code string generated by line-by-line the Huffman codes corresponding to the respective symbols, and a bit length of each of the Huffman codes in the Huffman code string .

In the present invention, the compression unit may further generate a coding dictionary including information on the frequency of occurrence of each symbol in each cyclic group.

According to still another aspect of the present invention, there is provided an apparatus for restoring binary data compressed by a binary data compression apparatus, the apparatus comprising: means for referring to the encoding dictionary for each cyclic group, And a restoring unit for restoring the binary data by restoring the binary data to a cyclic group.

The method and apparatus for compressing and restoring binary data according to the present invention are capable of quickly and efficiently compressing and restoring binary data through a simple operation and a hardware configuration, and also have excellent compression rate and reliability of compressed data and restored data Not only the transmission efficiency and the speed of data transmission can be improved. In particular, according to the present invention, the compression ratio of the Huffman coding scheme can be more effectively increased.

1 is a block diagram of a binary data compression apparatus and a decompression apparatus according to an embodiment of the present invention.
2 is a flowchart illustrating a method of compressing binary data according to an embodiment of the present invention.
3 shows a binary tree structure used in Huffman coding.
FIGS. 4 and 5 are reference diagrams for explaining allocation of Huffman codes for each symbol.
Figure 6 schematically illustrates a first circulating block and a second circulating block.
FIG. 7 schematically shows a first circulating block and a second circulating block according to a maximum circulating block method.
FIG. 8 schematically shows a first circulating block and a second circulating block according to a minimum circulating block method.
FIG. 9 is a conceptual diagram illustrating a process of constructing a Huffman tree on the restoration side by using the appearance frequency information of sequentially stored symbols to find Huffman codes on a symbol-by-symbol basis.
FIG. 10 conceptually shows a method of separating Huffman codes from a Huffman code string during data restoration and restoring symbol-by-symbol Huffman coding dictionaries for each cyclic group.
11 shows an outline of the operation principle of the circulation group acquisition unit.
FIG. 12 illustrates a process of restoring data from actual compressed data to a source symbol using the Huffman coding dictionary for each symbol of each cyclic group restored in FIG.

Hereinafter, embodiments of the present invention will be described in detail with reference to the accompanying drawings so that those skilled in the art can easily carry out the present invention. The present invention may, however, be embodied in many different forms and should not be construed as limited to the embodiments set forth herein. In order to clearly illustrate the present invention, parts not related to the description are omitted, and like parts are denoted by similar reference numerals throughout the specification.

Throughout the specification, when an element is referred to as "comprising ", it means that it can include other elements as well, without excluding other elements unless specifically stated otherwise.

FIG. 1 is a block diagram of a binary data compression apparatus and a decompression apparatus according to an embodiment of the present invention. Referring to FIG. 1, an embodiment according to the present invention will be described below.

As shown in FIG. 1, the binary data compression apparatus 100 according to the present embodiment includes a compression unit 110 and an output unit 120. The compression unit 110 includes a cyclic group acquisition unit 111 and a compressed data generation unit 112.

First, the cyclic group obtaining unit 111 scans original binary data in units of n bits in a first direction to obtain a plurality of cyclic blocks having a plurality of symbols of n bits length. The cyclic group obtaining unit 111 obtains a plurality of cyclic groups each including at least one cyclic block based on the scattering degree of the frequencies in which the symbols appear in one cyclic block or a plurality of neighboring cyclic blocks do. Herein, each of the circulating blocks is a binary block having symbols that meet until at least one of the 2 ⁿ kinds of symbols appears at least once when the original binary data is sequentially scanned in the first direction.

When acquiring the circulation group, the circulation group acquisition unit 111 calculates a first scattering degree of the frequencies in which the symbols appear in the current circulation block, compares the first scattering degree with a reference scattering degree, And acquires the current circulating block as the circulating group if the scattering degree is more than the scattering degree. If the first scattering degree is less than the reference scattering degree, the cyclic group obtaining unit 111 calculates a second scattering degree of the frequencies in which the symbols appear in the current cyclic block and at least one next cyclic block, And if the two scatterplots are equal to or greater than the reference scattering degree, the current circulating block and the at least one next circulating block are obtained as the circulating group. In this case, the number of the at least one next circulating block is a minimum number that makes the second scattering degree equal to or greater than the reference scattering degree.

The dispersion may be such that variance, standard deviation, skewness, or kurtosis of the frequency numbers may be employed, or an indicator of whether the minimum number of symbols occupies more than the reference frequency of the entire frequency may be employed, May be applied to other various ways of calculating the scattering degree. A detailed description thereof will be described later.

11 shows an outline of the operation principle of the circulation group acquisition unit.

The compression data generation unit 112 generates Huffman coding for the symbols in the respective cyclic groups to generate a plurality of compression cyclic groups by compressing each of the cyclic groups, And generates data.

When generating the plurality of compression cyclic groups, the compressed data generation unit 112 generates Huffman codes for each symbol according to the frequency of each symbol in each cyclic group, and outputs the corresponding symbols Huffman code to generate the compressed cyclic group. The compressed data generating unit 112 further generates a coding dictionary including a Huffman code string generated by concatenating the Huffman codes corresponding to the respective symbols in a line and a bit length of each Huffman code in the Huffman code string can do.

On the other hand, with regard to the binary numbers of the remaining groups remaining after acquiring the plurality of cyclic groups, the compression unit 110 performs Huffman coding based on the frequencies at which the symbols appear in the residual group to obtain a compressed residual group, It can be applied when generating compressed data.

The output unit 120 outputs the compressed data generated by the compression unit 110 to a target device such as the binary data restoration device 200 or the like.

1, the apparatus for recovering binary data 200 according to the present embodiment includes an input unit 210 and a decompression unit 220. The input unit 210 receives compressed data transmitted through the output unit 120 and transmits the compressed data to the decompression unit 220.

The restoring unit 220 includes a circulating group restoring unit 221 and a combining unit 222.

The cyclic group restoring unit 221 restores the plurality of compression cyclic groups into a plurality of cyclic groups by referring to the coding dictionary for each cyclic group, and the combining unit 222 combines the plurality of cyclic groups, .

The operation and operation of this embodiment thus configured will be described in detail with reference to Figs. 1 to 5. Fig.

FIG. 2 is a flow chart for explaining a method of compressing binary data according to an embodiment of the present invention. FIG. 3 shows a binary tree structure used in Huffman coding, and FIGS. Code, and a method of compressing binary data according to this embodiment will be described with reference to FIG.

Before explaining the compression method of binary data according to the present embodiment, Huffman coding will be described first.

For example, if a data file of 109,505 bytes is taken as an example, it is cut by 4 bits and its distribution is checked as shown in Table 1 below. The 4-bit binary number obtained at this time can have 16 (= 2 ⁴ ) values ranging from "0000" to "1111" ranging from 0 to 15, which is called a symbol. Table 1 shows the appearance frequency of each symbol in the 109,505 byte data file.

4-bit symbol Appearance frequency 0 24434 One 13255 2 13528 3 12391 4 13765 5 12907 6 13236 7 13298 8 12635 9 12377 10 13259 11 12420 12 11828 13 12956 14 12673 15 14048

Huffman coding is a type of entropy coding used for lossless compression and is an algorithm for generating a prefix code (a code in which a symbol of one symbol does not become a prefix of another symbol code) from the frequency of a symbol, The less frequently occurring symbol (symbol with lower frequency) allocates a longer code, and the frequent symbol (the symbol with higher frequency) allocates a shorter code. And generates a variable length code for a fixed length symbol.

There are many ways to design Huffman codes. One example is a method of generating Huffman codes using a binary tree. FIG. 3 shows a binary tree structure used in Huffman coding. In an binary tree structure, an outer node or an outer node or a leaf node represents a corresponding symbol. The Huffman code for an arbitrary symbol is generated by combining all codewords assigned to each branch, starting from the root node and going down to the leaf node corresponding to the symbol.

In FIG. 3, in the case of the A symbol, "0" is sequentially applied to the arm, which is a path descending from the root node through the inner node, in order of Huffman code, "10" Quot; 110 "in the case of the symbol C, and" 111 "

Referring to FIG. 3, a method of constructing the Huffman tree will be described in more detail with reference to an example of a minimum-likelihood Huffman code. The probability of occurrence between two newly generated symbols after combining two symbols having the lowest probability or frequency Is summed again with the probability of the remaining symbols, and then the code generation process is repeated. If we design a minimal variance Huffman code for a set of five symbols S = {A, B, C, D, E} Let's say once, twice, once, and once. Of course, it is also possible to calculate the probability of occurrence in addition to the number of occurrences, and to follow the same calculation procedure.

The start of the minimal transition Huffman code begins by combining the two symbols D and E with the lowest probability of occurrence or frequency of occurrence. Different bits are assigned to the last bit of the code word corresponding to these two symbols. In this case, the rule for assigning different bits is such that "1" is assigned to the side where the probability of occurrence or frequency of occurrence is "0", and vice versa. Once determined, however, the rules must be applied consistently throughout the entire tree generation process to simplify the decoding process.

If the probabilities or appearance frequencies are the same, a specific criterion can be selected and assigned according to the sort order of the symbols. In this embodiment, "0" is assigned to a symbol positioned above when the symbols are arranged in ascending order. However, if the symbols are rearranged in the descending order, the positions of the symbol located above and the symbol positioned below are changed, and the allocation order of "0" and "1 " may be changed. However, since the structure of the Huffman tree is basically maintained, It does not affect

That is, in FIG. 4, "0" is assigned to D and "1" is assigned to E, and it is also possible to assign "1" to D and "0" to E. However, once the rules are specified in descending order or ascending order, it is necessary to apply them consistently to simplify the decoding process.

Next, when the two symbols D and E are combined to form a new symbol n1, the occurrence frequency of the new symbol n1 is equal to the sum of the occurrence frequencies of the two symbols D and E. Therefore, the frequency of n1 is 2. If there is an existing symbol having the same probability or appearance frequency as the probability of a newly generated symbol at the time of reordering the symbols according to the frequency of occurrence of the four symbols A, B, n1 and C, Place the symbol above the existing symbol with the same probability or frequency of occurrence. This is the difference between a regular Huffman code and a minimal-variant Huffman code.

The above process is summarized in FIG. The sum of the signs of the paths down to each leaf node is Huffman code per symbol. Table 2 shows the result of applying this example to each symbol of an actual binary cluster (binary data) to generate a Huffman code.

symbol Appearance frequency Huffman code A 2 10 B 4 00 C 2 11 D One 010 E One 011

In the present embodiment, a method using a minimum mutation tree is mainly used as a Huffman coding method. However, it is needless to say that it is possible to perform coding and decoding using a very wide variety of known Huffman tree generation methods such as general Huffman tree and adaptive Huffman coding.

Table 2 shows the size of the minimum displacement after encoding according to the Huffman tree implementation.

Table 3 shows the coding of each symbol of the 109,505 byte data file described in Table 1 by performing Huffman coding.

4-bit symbol Appearance frequency Huffman code Compression size Original size 0 24434 000 73302 97736 One 13255 1001 53020 53020 2 13528 1100 54112 54112 3 12391 0010 49564 49564 4 13765 1101 55060 55060 5 12907 0110 51628 51628 6 13236 1000 52944 52944 7 13298 1011 53192 53192 8 12635 0100 50540 50540 9 12377 11111 61885 49508 10 13259 1010 53036 53036 11 12420 0011 49680 49680 12 11828 11110 59140 47312 13 12956 0111 51824 51824 14 12673 0101 50692 50692 15 14048 1110 56192 56192 total 875811 876040

As can be seen from Table 3, in some cases, the extrusion rate may be relatively low even if the actual compressed data excluding the Huffman coding dictionary is only 229 bits (= 876040-875811). Thus, the embodiment according to the present invention described below can improve the extrusion efficiency even more if the Huffman coding result according to the frequency of occurrence of symbols is not good, thereby effectively increasing the extrusion rate.

2, the cyclic group obtaining unit 111 of the compression unit 100 scans the original binary data in units of n bits in the first direction to obtain a plurality of cyclic groups having a plurality of symbols of n bits length Block (S201). That is, the cyclic group obtaining unit 111 scans the original binary data in units of n bits from the most significant bit to the lower bit direction or vice versa. Here, n is an arbitrary natural number. For example, if n = 4, a plurality of symbols having a length of 4 bits, that is, 16 (= 2 ⁴ ) kinds of symbols from "0000" to "1111" are scanned. Of course, if n = 6, 64 (= ²⁶ ) kinds of symbols of 6 bits length, i.e., "000000" to "111111" are obtained by scanning.

The circular block is a binary block having symbols that meet until at least one of the 2 ⁿ kinds of symbols appears at least once when the original binary data is sequentially scanned in the first direction. That is, in the case of scanning from the most significant bit to the lower bit direction, if n = 4, the circular block is a binary number of all symbols scanned until 16 (= 2 ⁴ ) kinds of symbols from "0000" Block. In this case, the frequency at which each symbol appears in one cyclic block may vary. For example, 0000 is once, 0001 is 5 times, 1111 is 10 times, and so forth. The above process of acquiring the circular block is repeatedly performed until the end of the original binary data.

Next, the cyclic group obtaining unit 111 obtains a plurality of cyclic groups each including at least one cyclic block based on the scattering degree of the frequencies at which the respective symbols appear in one cyclic block or a plurality of neighboring cyclic blocks (S202). The dispersion may be such that variance, standard deviation, skewness, or kurtosis of the frequency numbers may be employed, or an indicator of whether the minimum number of symbols occupies more than the reference frequency of the entire frequency may be employed, May be applied to other various ways of calculating the scattering degree. These indicators are common in that they indicate the degree of uneven distribution over the value of the frequency of appearance.

The concept of "index of at least some kinds of symbols occupying more than the reference rate of the total number of frequencies" means that when the frequencies of the smallest number of symbols are summed in order from the largest frequency to the largest, Of the total population. For example, if the reference rate is 50% of the total frequency, it is the minimum number of types of symbols occupying 50% or more of the total frequency. The smaller the number of symbols, the greater the proportion of a certain number of specific symbols.

Two settings are possible in this regard. A part of the total frequency and a minimum B type are specified, and when the symbols of the symbol group satisfying A and B are exceeded, Huffman coding is performed.

For example, as shown in Table 4, in the case of a circulation group in which 330 symbols (a part of which is abbreviated due to ground) appear in 256 symbols from 0 to 255, the sum of the frequencies corresponding to the frequencies of 50% Dog.

symbol Appearance frequency 0 272 One 131 2 85 3 164 4 137 5 100 6 101 7 138 8 112 9 95 10 107 240 82 241 81 242 134 243 97 244 90 245 109 246 140 247 84 248 68 249 112 250 93 251 95 252 101 253 85 254 95 255 99

As shown in Table 5 below, the minimum number of 80 species is 15,131, which is the first 50% or more of the total, when Table 4 is sorted in descending order according to the frequency of appearance, As the value of 80 is smaller, it can be seen that more symbols are distributed in a specific kind.

symbol Appearance frequency 138 979 40 960 162 959 69 431 81 420 20 397 0 272 36 211 146 196 92 195 73 177 183 170 56 167 70 167 113 166 3 164 72 164 91 163 200 162 219 162 140 160 149 160 28 159 227 157 105 156 218 155 130 154 142 153 145 153 25 152 32 151 199 151 50 150 198 150 90 149 181 146 228 146 35 145 57 144 107 143 115 143 24 142 237 142 114 141 214 141 99 140 158 140 231 140 246 140 100 139 7 138 84 138 4 137 103 137 110 137 121 137 201 137 204 137 207 137 48 135 79 135 112 135 182 135 21 134 206 134 242 134 109 133 134 133 150 132 192 132 One 131 59 131 147 131 222 131 12 130 14 130 30 130 44 129 210 129 75 128

More specifically, the cyclic group obtaining unit 111 calculates a first scattering degree of the frequencies in which the symbols appear in the current cyclic block, and compares the calculated first scattering degree with a reference scattering degree . The scattering degree such as dispersion or standard deviation is a value representing the degree of distribution of the frequency numbers (degree of scattering of the variables), and if the value is large, the deviation between the frequencies is large, meaning that the frequency is relatively large and small. . For example, if the frequency of a specific symbol A is considerably large and the frequency of a specific symbol B is relatively small, etc., a large number of frequency distributions are scattered around the average value, so that a specific symbol (s) do. Accordingly, compression of symbols having a relatively high frequency of appearance compared to other symbols can increase the data extrusion rate.

If the first scattering degree is equal to or greater than the reference scattering degree as a result of the comparison, the circulating group acquiring unit 111 acquires the current circulating block as the circulating group. Here, the cyclic group is a symbol group serving as a criterion for performing the Huffman encoding operation described below. If the first scattering degree, which is the scattering degree of the current cyclic block, is equal to or greater than the reference scattering degree, the frequency values of the symbols are scattered and distributed in the cyclic block, meaning that specific symbols are concentrated. Therefore, since the compression efficiency by Huffman encoding becomes relatively high with respect to such a cyclic block, the corresponding cyclic block is acquired as a cyclic group.

On the other hand, if the first scattering degree is less than the reference scattering degree, the circulation group obtaining unit 111 does not create the circulation group by only the current circulation block but calculates the scattering degree considering the following next circulation block. That is, the cyclic group obtaining unit 111 calculates a second scattering degree of the frequencies in which the symbols appear in the current cyclic block and the next cyclic block, and compares the second scattering degree with the reference scatterplot. If the second scattering degree is equal to or greater than the reference scattering degree, the circulating group obtaining unit 111 obtains the current circulating block and the next circulating block as the circulating group. If the second scattering degree which is the scattering degree of the current cyclic block and the next cyclic block is equal to or greater than the reference scattering degree, it can be seen that the specific symbols are concentrated in the two cyclic blocks, so that the compression efficiency by Huffman coding becomes relatively high, Lt; / RTI > as a cyclic group.

On the other hand, if the second scattering degree is also less than the reference scattering degree, the circulation group obtaining unit 111 repeatedly performs the above-described process such that the scattering degree is calculated considering the next circulating block and compared with the reference scattering degree, . Therefore, the obtained cyclic group may be composed of only one cyclic block, but it may be composed of various numbers of cyclic blocks according to the characteristics of data such as 2, 3, 5, and so on. However, since the number of the circulating blocks forming the circulating group is the minimum number that makes the scattering degree of the corresponding circulating blocks be greater than the reference scattering degree, once a certain number of the circulating block (s) becomes greater than the reference scattering degree, If satisfied, it acquires the cyclic group without considering the next cyclic block.

Next, the compressed data generation unit 112 performs Huffman encoding on the symbols in the respective cyclic groups, and compresses the cyclic groups to generate a plurality of compressed cyclic groups (S203). Specifically, the compressed data generation unit 112 generates a Huffman code for each symbol according to the appearance frequency of each symbol in each of the cyclic groups, and replaces each symbol in each cyclic group with the corresponding Huffman code. As described above, in the Huffman coding scheme, a relatively simple Huffman code is assigned to symbols having a high frequency and a relatively complex Huffman code is assigned to symbols having a low frequency. Thus, each cyclic group can be effectively compressed by assigning and replacing the corresponding Huffman codes for each symbol in each cyclic group.

When the compressed cyclic group is generated, the compressed data generation unit 112 generates a Huffman code string generated by concatenating the Huffman codes corresponding to the respective symbols in a row, and a bit length (DIC_SIZE) of each Huffman code in the Huffman code string ). ≪ / RTI > Of course, other known encoding dictionary generation methods may be used. If the Huffman coding is performed on each cyclic group of the original binary data, a compression cyclic group is created and the data is compressed. However, in order to restore the original binary data, it is necessary to know information about which Huffman codes are allocated to which symbols do. Therefore, Huffman codes corresponding to the respective symbols are connected in a row to generate a Huffman code string, and a bit length of each Huffman code included in the Huffman code string and an encoding dictionary for decoding using the Huffman code string will be.

For example, when compressing using symbols with a bit length of n = 4, Huffman codes 000, 11111, and 1100 for symbols 0000, 0001, 0010, 0011, ..., 1111 in a certain cyclic group , 0010, ..., 11110 correspond to each other, a Huffman code string such as 000/11111/1100/0010 /.../11110 made by concatenating the Huffman codes and a bit of each Huffman code in this code string 4, ..., 5) for the length of the data. This encoding dictionary creation is performed for each cyclic group.

On the other hand, if there are binary numbers of remaining groups remaining after the plurality of cyclic group acquiring steps, that is, if there are residual binary numbers (remaining groups) remaining after not being a circulating group after the cyclic group acquiring step (S202) The encoding unit 112 may perform Huffman encoding based on the frequencies at which the symbols appear in the residual group to obtain a compressed residual group. Of course, depending on the embodiment, the binary number of the remaining group may be transmitted without compression.

Next, the compressed data generation unit 112 combines the plurality of compression cyclic groups generated in the above-described manner to generate compressed data, and outputs the compressed data to the destination apparatus such as the restoration apparatus 200 through the output unit 120 (S204) . If there is a compressed residual group as described above, compressed data is also generated including the compressed residual group. At this time, the compression apparatus 100 can output the encoding dictionary for the generated cyclic groups together with the compressed data, and outputs the encoded dictionaries through a channel separate from the compressed data, As shown in FIG.

When the binary data is compressed and output through the above process, the binary data decompression apparatus 200 receives the compressed data through the input unit 210 and transmits the compressed data to the decompression unit 220. As shown in FIG. 1, the restoration unit 220 includes a circulation group restoration unit 221 and a combining unit 222. The cyclic group restoring unit 221 restores the plurality of compression cyclic groups into a plurality of cyclic groups by referring to the coding dictionary for each cyclic group, and the combining unit 222 combines the plurality of cyclic groups, .

In the present embodiment, it is described that the original binary data is scanned to acquire a plurality of cyclic blocks, and a cyclic group is acquired and Huffman coding is performed for each cyclic group. However, the present invention is not limited to this, To obtain a cyclic group while acquiring a cyclic block, and performing compression by performing Huffman coding immediately on the obtained cyclic group. That is, according to the present invention, even if scanning is completed for the entire original binary data or a circular block is not obtained for all of the original binary data, when at least one circular block obtained sequentially is satisfied as a circular group forming condition, The obtained cyclic group includes the case of directly performing Huffman coding.

The binary data compression method and the decompression method according to the present embodiment will be described in more detail with reference to the following embodiments.

Any original binary data was selected to perform compression according to the present embodiment. First, when a symbol consisting of 4 bits is sequentially read from the most significant bit and all the possible symbols are first encountered one or more times, it is referred to as a first circular block. For example, in the present embodiment, when the 101st data 10 (binary number "1010") is encountered, all the symbols from 0 to 15 are encountered at least once. Table 6 shows the result of counting after reading the 100th data, Table 7 shows the result of counting after reading up to the 101st data, which is the next data.

symbol Appearance frequency 0 33 One 5 2 4 3 6 4 7 5 8 6 9 7 7 8 5 9 One 11 2 12 3 13 4 14 4 15 2 total 100

symbol Appearance frequency 0 33 One 5 2 4 3 6 4 7 5 8 6 9 7 7 8 5 9 One 10 One 11 2 12 3 13 4 14 4 15 2 total 101

As shown in Tables 6 and 7, by encountering the 101st data 10 ("1010" in binary number) starting from the 1st, all the possible symbols 0 to 15 can be met for the first time more than once. By doing so, a first circular block, the first circular block, is obtained.

In the second cycle, the distribution of the frequency up to the 101st first cycle is divided and the continuous reading starts from the 102nd time, and conceptually again all occurrences of symbols 0 to 15 are encountered for the first time more than once . In this embodiment, the 102nd data through 1082st data constitute the second circular block.

Table 8 shows the distribution of the results of reading the data from the 102nd to the 1181st data, and Table 9 shows the distribution of the results of reading the data from the 102nd to the 1182rd data.

symbol Appearance frequency 0 1035 One 3 2 5 3 4 4 3 5 One 6 5 7 4 8 2 9 2 10 4 11 3 12 4 13 3 15 2 total 1080

symbol Appearance frequency 0 1035 One 3 2 5 3 4 4 3 5 One 6 5 7 4 8 2 9 2 10 4 11 3 12 4 13 3 14 One 15 2 total 1081

As shown in Tables 8 and 9, it can be seen that all the symbols from 0 to 15 meet again for the first time by meeting 14 of the 1182th.

Figure 6 schematically illustrates a first circulating block and a second circulating block.

As an additional variant of the method of distinguishing the cyclic blocks, when all the possible symbols are encountered for the first time, the symbol of that position (for example, the 101st, 1182st symbols, ... in the above figure) However, it is also possible to configure a single circular block including the case where the same symbols are consecutively displayed in succession. In the present embodiment, the former method is referred to as a minimum circulating block configuration method, and the latter method is referred to as a maximum circulating block configuration method. In this embodiment, both methods can be used, and a specific one can be selected by setting.

That is, FIG. 7 shows the maximum circulating block method. In the case where the same symbols as the 102nd to 199th symbols are consecutively consecutively, the first up to the 199th symbol is included in the first symbol block, The frequency of appearance of the symbol is initialized, and the second cyclic block is newly constructed from the 200th.

In the case of the minimum cyclic block scheme, as shown in FIG. 8, all the symbols that can be completely represented in the fixed-length bit extraction scheme in the 101-th symbol are met at least once, where the first cyclic block is generated, The appearance frequency of all the symbols is initialized, and the second circular block is newly constructed from the 102nd symbol.

In this embodiment, an example of the minimum block configuration method is described above, and the same algorithm can be applied by configuring the circulation block with the maximum block configuration method.

If the first circular block is identified as described above, the compression effect may be large when Huffman coding is performed on the first circular block using a distribution table of symbols and appearance frequency of the corresponding circular block, for example, Can be calculated in advance. For example, there may be a variety of methods for predicting the compression effect using an indicator. In this embodiment, as an example, the variance value of the appearance frequency per symbol is used as one index, Any form of statistical or numerical indicators that indicate trends in the frequency distribution may be used, but there may be some differences in compression efficiency. It is important to be able to confirm whether the compression effect of Huffman coding can be maximized. From Table 7, the variance value of the frequency of appearance of each symbol of the first cyclic block is approximately 56.495. Since the formula for obtaining the dispersion value can be applied by a known method, a specific calculation method is omitted.

Here, when this variance value is compared with the preset reference value T and the variance value is equal to or larger than T, it can be known that the data is sufficiently distributed and distributed. Therefore, it can be expected that the compression effect using Huffman coding is large. T = 10000 in Fig.

Therefore, the compression effect of sufficient Huffman coding can not be obtained with only the data distribution up to the first circular block in Table 7. [ In this case, the variance value of the appearance frequency of the data as shown in Table 10, which is the sum of the distribution values of the first circulation block and the second circulation block as the next circulation block, should be obtained. Table 10 shows the distribution of the first circular block of Table 7 and the second circular block of Table 9 together.

symbol Appearance frequency 0 1068 One 8 2 9 3 10 4 10 5 9 6 14 7 11 8 7 9 3 10 5 11 5 12 7 13 7 14 5 15 4 total 1182

Calculating the variance of the frequency of appearance in Table 10 is 70826.25. Therefore, since T = 10000, the first circulating block and the second circulating block are grouped and acquired as a circulating group to be referred to as a first circulating group (GRP = 1), and the first circulating group internally has two circulating blocks (RL = 2), the range is defined as from the 1st data to the 1182st data. The Huffman coding for the first cyclic group is performed to perform compression, as shown in Table 11 below. CV is the kind of symbol (0 ~ 15, expressed as binary numbers, 0000 ~ 1111), and HUFF is the Huffman coding result according to the minimum variation Huffman tree. VAR_S means the variance value of the frequency of occurrence (FRQ) in the corresponding cyclic group (GRP = 1). DIC_SIZE is bit length information of Huffman code for each symbol. On the other hand, in Table 11 below, the expected compression size expressed as COMPRESSED is 1620 bits and the original size is 4728 bits. In this case, the original is the size of the first to 1182-th data including the first circulating block and the second circulating block.

CV FRQ GRP RL VAR_S COMPRESSED DIC_SIZE ORIGINAL CV HUFF 0 1068 One 2 70286.25 1068 One 4272 0 0 One 8 One 2 70286.25 40 5 32 One 10100 2 9 One 2 70286.25 45 5 36 2 10010 3 10 One 2 70286.25 50 5 40 3 10000 4 10 One 2 70286.25 50 5 40 4 10001 5 9 One 2 70286.25 45 5 36 5 10011 6 14 One 2 70286.25 56 4 56 6 1100 7 11 One 2 70286.25 44 4 44 7 1110 8 7 One 2 70286.25 35 5 28 8 11010 9 3 One 2 70286.25 18 6 12 9 101011 10 5 One 2 70286.25 25 5 20 10 11110 11 5 One 2 70286.25 25 5 20 11 11111 12 7 One 2 70286.25 35 5 28 12 10110 13 7 One 2 70286.25 35 5 28 13 10111 14 5 One 2 70286.25 25 5 20 14 11011 15 4 One 2 70286.25 24 6 16 15 101010

That is, for each of the first to 1182 th data of the first and second circular blocks, bits are allocated to Huffman codes on the basis of the encoding table of Table 11 to generate compressed data.

After the compression for the first cyclic group is completed, the second cyclic group continues to be compressed on a group-by-group basis by applying the same method as the first cyclic group, starting from the 1183rd data, which is the next data. In the second cyclic group, the process of generating the first cyclic group may be similarly applied by considering that the 1183rd data is the same as the first data.

Table 12 shows the compression results for the second cyclic group. It can be seen that the second cyclic group consists of 16 cyclic blocks. That is, it can be seen that the variance with respect to the frequency of symbols from the third to the eighteenth cyclic block is 70582.36, which exceeds T = 10000 for the first time, and includes 2069 data which is the sum of the frequency (FRQ) The data is up to the 3251th data. As a result, it can be seen that the original data of 8276 bits is compressed to 5760 bits.

CV FRQ GRP RL VAR_S COMPRESSED DIC_SIZE ORIGINAL CV HUFF 0 1125 2 16 70582.36 1125 One 4500 0 0 One 61 2 16 70582.36 305 5 244 One 10100 2 85 2 16 70582.36 340 4 340 2 1111 3 57 2 16 70582.36 285 5 228 3 11000 4 53 2 16 70582.36 265 5 212 4 11101 5 57 2 16 70582.36 285 5 228 5 11001 6 71 2 16 70582.36 355 5 284 6 10010 7 74 2 16 70582.36 370 5 296 7 10001 8 59 2 16 70582.36 295 5 236 8 10111 9 54 2 16 70582.36 270 5 216 9 11100 10 77 2 16 70582.36 385 5 308 10 10000 11 63 2 16 70582.36 315 5 252 11 10011 12 56 2 16 70582.36 280 5 224 12 11010 13 56 2 16 70582.36 280 5 224 13 11011 14 60 2 16 70582.36 300 5 240 14 10110 15 61 2 16 70582.36 305 5 244 15 10101

The method of calculating the variance by summing the number of distributions of the circulating blocks in relation to the reference value T of the variance values is also economical in terms of speed. The Huffman coding is actually performed by summing the distributions of the respective circulating blocks, It is also possible to calculate the summation combination of the circulating blocks and perform compression while grouping them into one group.

That is, in the above description, although the first cyclic block and the second cyclic block are grouped into the first group using the variance of the frequency of occurrence of each symbol exceeding T = 10000, actual Huffman coding is performed on the first cyclic block, A compression ratio, a compression ratio obtained by performing actual Huffman coding on the first cyclic block + the second cyclic block, and a compression ratio obtained by performing actual Huffman coding on the first cyclic block + the second cyclic block + the third cyclic block, To the b-th cyclic block, it is possible to compute the most effective combination among the compression ratios of the result of the actual Huffman coding and to perform compression while grouping the groups.

Alternatively, each time one circular block is identified, one circular block may be recognized as a cyclic group and compression may be performed through Huffman coding without considering the variance value of the frequency of symbol occurrence in the circular block.

If the compression is performed for each group, the data is compressed until the end of the data. If the compression is performed after grouping the data of the last group, It reaches the end. The last unfinished group is referred to as the last group. Even if the last group does not satisfy the condition of the reference value T, the Huffman encoding is performed as it is. In the present embodiment, the 28th group is the last group, and the compression result is shown in Table 13. Since the final group does not satisfy the dispersion condition, the variance value (VAR_S) is meaningless, and it is not significant whether several circulating blocks are bundled.

CV FRQ GRP RL VAR_S COMPRESSED DIC_SIZE ORIGINAL CV HUFF 0 219 28 219 One 876 0 One One 25 28 100 4 100 One 0101 2 29 28 116 4 116 2 0100 3 11 28 55 5 44 3 01110 4 17 28 85 5 68 4 00011 5 13 28 65 5 52 5 01100 6 41 28 164 4 164 6 0000 7 35 28 140 4 140 7 0010 8 11 28 55 5 44 8 01111 9 10 28 60 6 40 9 000100 10 5 28 30 6 20 10 011011 11 6 28 36 6 24 11 011010 12 9 28 54 6 36 12 001110 13 9 28 54 6 36 13 000101 14 7 28 42 6 28 14 001111 15 16 28 80 5 64 15 00110

Table 14 is a table showing compression effects of all groups in this embodiment. From the ground level, the 3 to 26 groups are omitted.

On the other hand, in the case of the final cyclic group, Huffman coding can be forcibly performed in a state in which all the symbols do not appear at least one time. In this case, since there is no Huffman code for the symbols not yet displayed until the end of the binary data, In the case of the cyclic group, 0 is recorded in the bit length (DIC_SIZE) for symbols that do not exist, so that Huffman codes for the corresponding symbols can not be allocated when the Huffman codes are divided for each symbol in the Huffman code sequence.

CV FRQ GRP RL VAR_S COMPRESSED DIC_SIZE ORIGINAL CV HUFF 0 1068 One 2 70286.3 1068 One 4272 0 0 One 8 One 2 70286.3 40 5 32 One 10100 2 9 One 2 70286.3 45 5 36 2 10010 3 10 One 2 70286.3 50 5 40 3 10000 4 10 One 2 70286.3 50 5 40 4 10001 5 9 One 2 70286.3 45 5 36 5 10011 6 14 One 2 70286.3 56 4 56 6 1100 7 11 One 2 70286.3 44 4 44 7 1110 8 7 One 2 70286.3 35 5 28 8 11010 9 3 One 2 70286.3 18 6 12 9 101011 10 5 One 2 70286.3 25 5 20 10 11110 11 5 One 2 70286.3 25 5 20 11 11111 12 7 One 2 70286.3 35 5 28 12 10110 13 7 One 2 70286.3 35 5 28 13 10111 14 5 One 2 70286.3 25 5 20 14 11011 15 4 One 2 70286.3 24 6 16 15 101010 0 1125 2 16 70582.4 1125 One 4500 0 0 One 61 2 16 70582.4 305 5 244 One 10100 2 85 2 16 70582.4 340 4 340 2 1111 3 57 2 16 70582.4 285 5 228 3 11000 4 53 2 16 70582.4 265 5 212 4 11101 5 57 2 16 70582.4 285 5 228 5 11001 6 71 2 16 70582.4 355 5 284 6 10010 7 74 2 16 70582.4 370 5 296 7 10001 8 59 2 16 70582.4 295 5 236 8 10111 9 54 2 16 70582.4 270 5 216 9 11100 10 77 2 16 70582.4 385 5 308 10 10000 11 63 2 16 70582.4 315 5 252 11 10011 12 56 2 16 70582.4 280 5 224 12 11010 13 56 2 16 70582.4 280 5 224 13 11011 14 60 2 16 70582.4 300 5 240 14 10110 15 61 2 16 70582.4 305 5 244 15 10101 0 477 27 8 12572.7 477 One 1908 0 One One 55 27 8 12572.7 220 4 220 One 0110 2 65 27 8 12572.7 260 4 260 2 0100 3 26 27 8 12572.7 156 6 104 3 000010 4 54 27 8 12572.7 216 4 216 4 0111 5 51 27 8 12572.7 255 5 204 5 00000 6 100 27 8 12572.7 400 4 400 6 0001 7 80 27 8 12572.7 320 4 320 7 0011 8 23 27 8 12572.7 138 6 92 8 001010 9 23 27 8 12572.7 138 6 92 9 001001 10 12 27 8 12572.7 72 6 48 10 010111 11 16 27 8 12572.7 96 6 64 11 010110 12 24 27 8 12572.7 144 6 96 12 001000 13 25 27 8 12572.7 150 6 100 13 000011 14 17 27 8 12572.7 102 6 68 14 001011 15 30 27 8 12572.7 150 5 120 15 01010 0 219 28 219 One 876 0 One One 25 28 100 4 100 One 0101 2 29 28 116 4 116 2 0100 3 11 28 55 5 44 3 01110 4 17 28 85 5 68 4 00011 5 13 28 65 5 52 5 01100 6 41 28 164 4 164 6 0000 7 35 28 140 4 140 7 0010 8 11 28 55 5 44 8 01111 9 10 28 60 6 40 9 000100 10 5 28 30 6 20 10 011011 11 6 28 36 6 24 11 011010 12 9 28 54 6 36 12 001110 13 9 28 54 6 36 13 000101 14 7 28 42 6 28 14 001111 15 16 28 80 5 64 15 00110

A method of compressing the original data size of 876,040 bits to 857,929 bits as a result of the compression and constructing an encoding dictionary (dictionary information) for reconstruction is described here.

In the case of Huffman compression, there are two methods for constructing a coding dictionary. The first method is to store the symbol-to-appearance frequency information in one cyclic group in the coding dictionary when the cyclic group generation condition in compression is satisfied , A dynamically implementing a Huffman tree for generating symbol-by-symbol Huffman codes at the decompression side (information decompression side) upon decompression, and using the dynamically implemented symbols of the Huffman tree (Decompression). In the second method, it is disadvantageous in terms of speed to generate the Huffman tree from the restoration side (decompressed side) dynamically. Therefore, in compression, the symbol-by-Huffman code Data is stored in an encoding dictionary in a Huffman code sequence, and each Huffman code is separated into symbols in each of the Huffman code strings, Bit length information (DIC_SIZE) of only the codes is also stored, and the Huffman codes are separated in the compressed data at the time of restoration and converted into symbols.

In the case of the former, only the frequency-of-occurrence information sequentially listed for each symbol, which is denoted by FRQ for each cyclic group, may be stored as shown in Tables 15 and 16 below. This is because Huffman tree can be created for use in decompression in one cyclic group by using the 16-symbol-by-occurrence frequency information of all possible symbols 0 to 15. Of course, it is also possible to recompress only the appearance frequency information values in the following table using a separate compression method.

CV FRQ 0 1068 One 8 2 9 3 10 4 10 5 9 6 14 7 11 8 7 9 3 10 5 11 5 12 7 13 7 14 5 15 4 0 1125 One 61 2 85 3 57 4 53 5 57 6 71 7 74 8 59 9 54 10 77 11 63 12 56 13 56 14 60 15 61 ... ... 0 477 One 55 2 65 3 26 4 54 5 51 6 100 7 80 8 23 9 23 10 12 11 16 12 24 13 25 14 17 15 30 0 219 One 25 2 29 3 11 4 17 5 13 6 41 7 35 8 11 9 10 10 5 11 6 12 9 13 9 14 7 15 16

That is, as shown in FIG. 9, symbol-appearance frequency information can be obtained from only the appearance frequency information of sequentially stored symbols in units of 16, and a Huffman tree is constructed using the same Huffman coding method as that of compression Then, Huffman codes can be found by symbol as shown in Table 16.

symbol Huffman code 0 0 One 10100 2 10010 3 10000 4 10001 5 10011 6 1100 7 1110 8 11010 9 101011 10 11110 11 11111 12 10110 13 10111 14 11011 15 101010

By using the Huffman code thus obtained, the Huffman code can be separated from the compressed data and restored (decompressed) with respect to the symbol.

Next, the second method for constructing the encoding dictionary will be described as follows.

The Huffman codes (HUFF codes) in Table 14 are sequentially stored in a row in order for the encoding dictionary (dictionary information) for decompression (decompression). In other words, the Huffman codes of Group 1 in Table 14 are shown in Table 17 as an example. If they are stored in a row, they will be equal to "010100100101000010001 ... 101010".

HUFF 0 10100 10010 10000 10001 10011 1100 1110 11010 101011 11110 11111 10110 10111 11011 101010

In order to precisely and precisely divide the Huffman code strings constituted by such a series in symbol units of 0 to 15 in each circulation group, each Huffman code of Table 14, which is the length of Huffman codes corresponding to each symbol of each Huffman code string The bit length (DIC_SIZE) information of the DIC_SIZE information should also be sequentially stored. It is effective to store the DIC_SIZE information separately in the Huffman coding, compress and store the values in Table 14 rather than storing the values represented by the values 1 through 7 . Table 18 shows the distribution values for the DIC_SIZE values in Table 14, and Table 19 shows the Huffman coding and compression size results according to the bit length (DIC_SIZE) of each Huffman code and the frequency of application, see.

DIC_SIZE Appearance frequency One 7 2 4 3 15 4 284 5 112 6 22 7 4 total 448

DIC_SIZE Appearance frequency Huffman coding Compression size One 7 00110 35 2 4 001110 24 3 15 0010 60 4 284 One 284 5 112 01 224 6 22 000 66 7 4 001111 24 total 448 717

If DIC_SIZE is not compressed, the 3-bit binary number to represent 1 to 7 is multiplied by 448 times, so that 1344 bits can be obtained. It will be necessary.

Using the DIC_SIZE information, Huffman codes, which are sequentially mapped to symbols in each cyclic group represented in Table 14, are divided into 16 Huffman codes corresponding to all the symbols in each cyclic group in the Huffman code sequence in which the Huffman codes are connected in series And in this way eventually, Huffman codes can be mapped for restoration (decompression) for all symbols in all cyclic groups. As described above, Huffman coding can be forcibly performed in a final cyclic group even when all symbols are not present at least one time. Due to the characteristics, in the final cyclic group, DIC_SIZE is set to 0 for a specific symbol , The Huffman code for the symbol can be prevented from being allocated when the Huffman code sequence is divided.

On the other hand, Huffman compressed data 010100100101000010001 ... 101010 of the bit length (DIC_SIZE) information of the Huffman code for each symbol of the group 1 is added to the Huffman code string "010100100101000010001 ... 101010" Can be decrypted using the encoding table in Table 19. < tb > < TABLE > DIC_SIZE is 1, 5, 5, 5, 5, 4, 4, 5, 6, 5, .... Using this information, the Huffman code sequence is divided into the bit length of the corresponding DIC_SIZE value When divided, the Huffman code shown in Table 17 is completely restored. It is natural that each Huffman code is a symbol-by-Huffman code used by one cyclic group every 16 for a 4-bit fixed length extracted symbol. The symbols at this time are the mapping of the Huffman codes according to the sequential order from 0 to 15. That is, if it is known how many bits fixed length extraction has been performed, it can be known whether Huffman codes are used by one cyclic group for each of several Huffman codes.

Therefore, in summary as an encoding dictionary (dictionary information) for decompression (decompression), a coding table of Huffman code sequence / DIC_SIZE in which Huffman codes of each cell in Table 16 are connected is needed.

For reference, in the decompression (decompression) as described above, information (RL) about how many cyclic blocks are included in each group is not needed because the frequency of occurrence of symbols in the cyclic block If the variance is less than a specific reference scale (T = 10000 in the above example), the frequency of appearance of symbols is calculated naturally by adding up to the next cycle block. On the other hand, It is because.

The restoration method (decompression method) will be described as follows.

FIG. 10 conceptually shows a method of separating Huffman codes from a Huffman code string during data restoration and restoring symbol-by-symbol Huffman coding dictionaries for each cyclic group. FIG. 12 illustrates a process of restoring data from actual compressed data to original symbols using the Huffman-encoded dictionaries for each symbol of each cyclic group restored in FIG. The Huffman codes for the original binary data and the DIC_SIZE information for the compressed data are used to decode the Huffman codes for each symbol from 0 to 15 to be applied to each group. Is a sequential Huffman code for each symbol 0-15 for decompression to be applied to one group per 16 Huffman codes.

Referring to FIG. 10, 16 Huffman codes are separated from the DIC-SIZE decompressed data for the Huffman code string. Table 20 shows symbol-by-symbol Huffman code encoding dictionary information necessary for decoding for the first cyclic group. Now, moving from the most significant bit of the compressed data to the lower bit direction, decoding is performed on the symbols assigned to each of the lower Huffman codes.

However, in the continuous Huffman code in the compressed data, a theoretical explanation is given as to how the last Huffman code for decoding the last symbol of the first cyclic group and the first Huffman code for decoding the first symbol of the second cyclic group are separated naturally If the last Huffman code of the first cyclic group is H1, then the most significant first bit of the second cyclic group Huffman code necessarily starts from either 0 or 1. However, if it is assumed that H1 + '0' and H1 + '1' are Huffman codes used to decode certain other symbols of the first cyclic group, the prefix codes of Huffman codes for certain symbol- code, that is, a specific Huffman code does not satisfy the basic Huffman code characteristic that it is not included as a prefix of another Huffman code, such a Huffman code can not exist. That is, in the first cyclic group, there is no code including the H1 code as a prefix code of another Huffman code such as H1, H1 + '0 ....', H1 + '1 ....' Therefore, the last Huffman code in the first cyclic group can be uniquely identified as H1, thereby deciphering the last symbol to thereby generate all kinds of emergent symbols that constitute the last cyclic block in the first cyclic group The symbol separation between the cyclic groups from the compressed data occurs very smoothly.

GRP symbol HUFF One 0 0 One One 10100 One 2 10010 One 3 10000 One 4 10001 One 5 10011 One 6 1100 One 7 1110 One 8 11010 One 9 101011 One 10 11110 One 11 11111 One 12 10110 One 13 10111 One 14 11011 One 15 101010

In this embodiment, since the minimum cyclic block method is used at the time of compression, the frequency distribution of decoded symbols is calculated while decoding is performed, In the case of one or more encounters, the scattering degree is calculated from the frequency of appearance of the collected symbols until that time. If the value is equal to or greater than the reference scattering degree, the corresponding cyclic block is divided into the cyclic groups. However, The corresponding cyclic block can not be separated into the cyclic groups and continuously collects the symbols. In this case, although the corresponding cyclic block is not independent of the cyclic group, in order to check the next cyclic block, the frequency distribution of the previously calculated symbols is reset to 0, and the frequency of appearance of the symbol for the next cyclic block is Should be calculated.

In this way, after confirming a new cyclic block in which all the symbols are first to be encountered more than once for the first time, the scattering degree is calculated from the appearance frequency information of the symbols about the set of symbols in the two cyclic blocks, It is naturally understood that the two circulating blocks can be separated into the circulating group only if the reference scattering degree is greater than the reference scattering degree. That is, each time a decoding block is identified, the scattering degree of all the symbols in the collected block is calculated, and whether the value is equal to or greater than the reference dispersion is itself a separation criterion of the cyclic group.

However, if the circulating block configuration method is applied to the maximum cyclic block configuration method, if the frequency distribution of decoded symbols is calculated while decoding, if all the symbols are encountered more than once for the first time, It is confirmed whether or not the same symbol as the reference symbol continues to be displayed next to the position of the reference symbol at the time of the encounter, and if it occurs, it is recognized as a circular block in one circular group when all the symbols are encountered. The procedure for obtaining the scattering degree for the circular blocks and obtaining the circular group is the same as that for the minimum circular block method.

In the case of the cyclic group 1, the decoding using the table of Table 20 for the symbols corresponding to the first cyclic block and the second cyclic block is completed.

Next, Huffman code for each symbol from 0 to 15 to be applied to the 17th to 32nd second cyclic groups is decoded using Huffman code string and DIC_SIZE information for Huffman code compressed data for original binary data, Table 21 shows a Huffman decoding table for decoding the second cyclic group. Now, from the next data of the data up to the first group decoding from the compressed data, the decoding is performed on the symbols allocated to the lower Huffman codes while moving in the lower bit direction. The decryption is performed in the same manner as the decryption of the first cyclic group but continues as the data of the second cyclic group on the basis of the table of Table 21. [

GRP symbol HUFF 2 0 0 2 One 10100 2 2 1111 2 3 11000 2 4 11101 2 5 11001 2 6 10010 2 7 10001 2 8 10111 2 9 11100 2 10 10000 2 11 10011 2 12 11010 2 13 11011 2 14 10110 2 15 10101

As described above, the above process is continuously repeated. In the case of the last group, the decompression process is continued by referring to the Huffman table of the group until the end of the compressed data is reached, .

If the effect of the present invention can be seen more quantitatively, data of 857, 892 bits + DIC_SIZE compressed data 717 bits and DIC_SIZE decompression table (Huffman code data per DIC_SIZE in Table 19) The final compressed data can be obtained. However, as shown in Table 2, if Huffman coding is performed using the minimum variation Huffman tree for the same data, only 875,811 bits are compressed, and the value is also a value excluding the Huffman coding dictionary (Huffman dictionary table).

Thus, according to the present embodiment, an additional compression effect of about 17,000 bits can be obtained. In particular, it is possible to use the data distribution in which the occurrence frequency is almost even and the efficiency of Huffman coding is inferior.

In this embodiment, 2 ^k symbols of but on the basis of 16 symbols, to the basis of the symbols separated by the n-bit units from 0 to 2 ^k-1 from 0 to 15 on the basis of symbols that are separated by 4 bits, The user can designate the reference value T appropriately according to the overall size and dispersion of the data so as to show the optimum compression ratio.

For example, when data is divided into 8-bit units, 256 symbols from 0 to 255 are possible.

When the original binary data of 5,682,432 bits is read in units of 8 bits, 710,304 bytes are obtained.

symbol Appearance frequency 0 7465 One 3878 2 3455 3 4163 4 2237 5 2696 6 3644 7 2883 8 2029 9 2193 10 2383 11 2475 12 3067 13 2506 14 12051 15 2262 16 1638 17 2112 18 2429 ... ... 240 3181 241 3418 242 1180 243 1797 244 1722 245 1394 246 1966 247 1421 248 3151 249 1392 250 1756 251 1721 252 2103 253 1650 254 1740 255 1779

A method of dividing 256 symbols into one of the minimum circulation block or the maximum circulation block unit every time the first intersection of the 710,304 symbols is shifted from the top to the bottom (or vice versa) The frequency of occurrence of 256 symbols including the corresponding cyclic group and the corresponding new block is calculated and then the variance value of the appearance frequency is calculated and when the specific value (reference value) is not exceeded (10,000 in this embodiment) And the new cyclic blocks are integrated to continue the cyclic grouping. If the distribution exceeds a certain threshold value, Huffman coding is repeated for the cyclic group. In the above description, only the 4 bit fixed bit stream is extracted or the 8 bit fixed bit stream is extracted.

However, in the case of extracting an 8-bit fixed bit stream, in this specification, an example in which compression is performed in a manner of dividing into a maximum cyclic block unit in a cyclic block configuration is shown in Table 23.

As a result of applying the method according to the present embodiment to all 5,682,432 bits, it was compressed to 5,320,225 bits (including a coding dictionary), and it was found that the compression rate was reduced to 5,544,278 bits have.

As shown in the following Table 23, 25 recurring groups (GRPs) are generated. Since the last 25 recurring groups have reached the end of data while the variance value in the recurring group has not exceeded the reference value, The result of summing 12 cyclic blocks in the case of the first cyclic group is that the variance value of the occurrence frequency per symbol exceeds the reference value (10000 or more in the present invention), and 12 cyclic blocks are generated And is a result of performing Huffman coding on the first cyclic group.

CV FRQ GRP RL VAR_S COMPRESSED DIC_SIZE ORIGINAL HUFF 0 1479 One 12 16946.68 5916 4 11832 1110 One 96 One 12 16946.68 768 8 768 11000000 2 79 One 12 16946.68 711 9 632 001001000 3 112 One 12 16946.68 896 8 896 10001000 4 88 One 12 16946.68 792 9 704 000001000 5 85 One 12 16946.68 765 9 680 000101000 6 118 One 12 16946.68 944 8 944 01111110 7 93 One 12 16946.68 744 8 744 11010110 8 82 One 12 16946.68 738 9 656 000111000 9 79 One 12 16946.68 711 9 632 001001001 10 80 One 12 16946.68 720 9 640 001000100 11 87 One 12 16946.68 783 9 696 000011010 ... ... ... ... ... ... ... ... ... 250 69 One 12 16946.68 621 9 552 011000101 251 75 One 12 16946.68 675 9 600 001011111 252 115 One 12 16946.68 920 8 920 10000101 253 88 One 12 16946.68 792 9 704 000000011 254 101 One 12 16946.68 808 8 808 10101111 255 150 One 12 16946.68 1200 8 1200 00110100 ... ... ... ... ... ... ... ... ... 0 1780 25 8900 5 14240 00001 One 176 25 1408 8 1408 01001010 2 295 25 2065 7 2360 1000001 3 151 25 1208 8 1208 01111010 4 95 25 855 9 760 001110000 5 283 25 1981 7 2264 1001101 6 169 25 1352 8 1352 01010010 7 105 25 945 9 840 001000110 8 201 25 1608 8 1608 00101100 9 186 25 1488 8 1488 00111111 10 100 25 900 9 800 001011110 ... ... ... ... ... ... ... 250 144 25 1152 8 1152 10001111 251 114 25 1026 9 912 000000101 252 92 25 828 9 736 010000011 253 132 25 1056 8 1056 10111100 254 101 25 909 9 808 001010111 255 147 25 1176 8 1176 10000101

As described above, the method and apparatus for compressing and restoring binary data according to the present embodiment can quickly and efficiently compress and restore binary data through a simple operation and a hardware configuration, Not only the reliability of data can be increased, but also transmission efficiency and speed can be improved in data transmission.

While the invention has been shown and described in detail in the foregoing description, it is evident that many alternatives, modifications and variations will be apparent to those skilled in the art, Of the right.

100: binary data compression device
110: compression unit 111: circulation group acquisition unit
112: compressed data generation unit
120: Output section
200: Binary data restoration device
210: input unit 220:
221: circulation group restoring unit 222:

Claims (18)

A method of compressing binary data of a compression device,
Wherein the compressing unit scans the original binary data in units of n bits in a first direction to obtain a plurality of recursive blocks having a plurality of symbols of n bits in length, a binary block the acquisition phase rotation block having symbols meet until all 2 ⁿ different types of the symbol is at least once when scanned sequentially;
Obtaining a plurality of cyclic groups each including at least one cyclic block based on the scattering of the frequencies at which each of the symbols appears within one cyclic block or a plurality of neighboring cyclic blocks;
The compression unit compresses each of the cyclic groups by performing Houghman coding on the symbols in the respective cyclic groups to generate a plurality of compression cyclic groups; And
And the compressing unit combines the plurality of compression cyclic groups to generate compressed data.
The method according to claim 1,
Wherein acquiring the cyclic group comprises:
Computing a first scatterplot of the frequencies at which the symbols appear in the current cyclic block;
Comparing the first scatter diagram with a reference scatter diagram; And
And obtaining the current cyclic block as the cyclic group if the first scatterplot is equal to or greater than the reference scatterplot.
3. The method of claim 2,
Calculating a second scattering degree of the frequencies at which the symbols appear in the current circular block and at least one next circular block if the first scatterplitude is less than the reference scatterplitude, And acquiring the at least one next circular block as the circular group,
Wherein the number of the at least one next circular block is a minimum number that allows the second scatterplot to be greater than the reference scatterplot.
The method according to claim 1,
Wherein an index is employed as to whether the variance, standard deviation, skewness, or kurtosis of the frequency numbers are employed, or whether at least some kinds of symbols occupy more than a reference rate of the total frequency. Way.

The method according to claim 1,
Further comprising performing Huffman coding on the binary numbers of the remaining groups remaining after the plurality of cyclic group acquiring steps, based on the frequencies at which the compression unit appears the symbols in the remaining group to obtain a compression residual group A binary data compression method.
The method according to claim 1,
Wherein the generating the plurality of compressed cyclic groups comprises:
Generating a Huffman code for each symbol according to the frequency of each symbol in each cyclic group; And
And replacing each symbol in each cyclic group with the corresponding Huffman code.
The method according to claim 6,
Wherein the generating the plurality of compressed cyclic groups comprises:
Generating a coding dictionary including a Huffman code sequence generated by concatenating the Huffman codes corresponding to the respective symbols in each of the circulation groups in a row and a bit length of each of the Huffman codes in the Huffman code sequence A binary data compression method.
The method according to claim 6,
Wherein the generating the plurality of compressed cyclic groups comprises:
And generating an encoding dictionary including information on the frequency of appearance of each symbol in each cyclic group.
A method for restoring binary data compressed by the binary data compression method according to claim 7 or 8,
Wherein the restoration unit reconstructs the plurality of compression cyclic groups into a plurality of cyclic groups by referring to the encoding dictionary for each cyclic group to restore the binary data.
A binary data compression apparatus comprising:
Scanning the original binary data in units of n bits in a first direction to obtain a plurality of circulating blocks having a plurality of symbols of n bits length,
Obtaining a plurality of cyclic groups each including at least one cyclic block based on a scattering degree of frequencies in which each of the symbols appears within one cyclic block or a plurality of neighboring cyclic blocks,
Performing a Huffman encoding on the symbols in each of the circulation groups to generate a plurality of compression cyclic groups by compressing each of the cyclic groups,
And a compression unit for combining the plurality of compression cyclic groups to generate compressed data,
Wherein each of the circulating blocks is a binary block having symbols that meet until at least one of the different 2 ^{< n >} kinds of symbols appear at least once when the original binary data is sequentially scanned in the first direction. Device.
11. The method of claim 10,
Wherein the compressing unit calculates a first scattering degree of the frequencies in which the symbols appear in the current cyclic block, compares the first scattering degree with a reference scattering degree, and if the first scattering degree is equal to or greater than the reference scattering degree, As the cyclic group, the cyclic block of the binary data.
12. The method of claim 11,
If the first scatterplit is less than the reference scatterplitude, the compressing unit computes a second scatterplot of the frequencies at which the symbols appear in the current circular block and at least one subsequent circular block, and if the second scatterplot is greater than or equal to the reference scatterplot The current circular block and the at least one next circular block are obtained as the circular group,
Wherein the number of the at least one next circular block is a minimum number that allows the second scatterplot to be equal to or greater than the reference scatterplot.
11. The method of claim 10,
Wherein an index is employed as to whether the variance, standard deviation, skewness, or kurtosis of the frequency numbers are employed, or whether at least some kinds of symbols occupy more than a reference rate of the total frequency. Device.
11. The method of claim 10,
Wherein the compression unit further obtains a compressed residual group by performing Huffman encoding based on the frequencies at which the symbols appear in the residual group, with respect to binary numbers of remaining groups remaining after acquiring the plurality of cyclic groups. Compression device.
11. The method of claim 10,
When generating the plurality of compression cyclic groups, the compression unit generates a Huffman code for each symbol according to the frequency of each symbol in each cyclic group, and replaces each symbol in each cyclic group with the corresponding Huffman code Wherein the binary data compression device comprises:
16. The method of claim 15,
Wherein the compression unit further generates a coding dictionary including a Huffman code string generated by concatenating the Huffman codes corresponding to the respective symbols in a row and a bit length of each Huffman code in the Huffman code string. Compression device.
16. The method of claim 15,
Wherein the compression unit further generates an encoding dictionary including information on an occurrence frequency of each symbol in each cyclic group.
An apparatus for restoring binary data compressed by the binary data compression apparatus according to claim 16 or 17,
And restoring the binary data by restoring the plurality of compression cyclic groups into a plurality of cyclic groups by referring to the encoding dictionary for each cyclic group.

Images