KR19990085045A - Lossless data compression and decompression - Google Patents

Abstract

The present invention relates to a lossless data compression and decompression method by iterative circulation, comprising: receiving N bits of information and dividing the information into K bits; Determining a predetermined number of representative values representing two or more values among 2 ^K values of divided K bits of information and dividing the information by information bits; 2 but classify ^K kinds of values into two or more groups based on the frequency of in the K-bit information, the group that have a low frequency value that group that the value of the number of frequencies to include the representative value is the representative value Including the value deleted due to being determined; Generating a first bit field for distinguishing each group, and generating a new value by combining one or more values included in two or more groups selected from each group according to the generated first bit; And if bit savings are made, additional information is received from the bitstream to be compressed, combined with the bitstreams with bit savings, and the compression process is repeated again. If no bit savings are made, the information of N bits is kept as it is. And receiving a new N-bit information from the compression target bitstream and repeating the compression process.

According to the present invention, data can be losslessly compressed at a high compression rate, and the losslessly compressed data can be easily extended.

Description

Lossless data compression and decompression

The present invention relates to a data compression and decompression method, and more particularly, to a lossless data compression and decompression method by iterative circulation.

As the amount of information of data processed by computers and networks increases rapidly, recording media and high-speed information transmission for storing a large amount of information are required. However, the advancement speed of hardware technology to increase the recording capacity of the recording medium and the information transmission speed through the network are gradually increased compared to the expansion speed of the information amount, so that a large amount of information is efficiently stored in the limited capacity recording medium. In addition, research is being actively conducted on how to transmit data through a network in a short time.

In order to store a large amount of data in a limited-capacity recording medium, there is a method of compressing and storing data, and expanding and using the compressed and stored data. In addition, in order to transmit more data to the network within a limited time, there is a method of compressing and transmitting data at a transmitting end and extending and using the compressed and transmitted data at a receiving end. Such data compression methods include a lossless compression method in which all information of compressed data can be restored as it is, and a loss data compression method in which some data is lost during compression and restoration.

Currently, the lossless data compression method can achieve a compression ratio of about 3: 1, and the lossy data compression method can achieve a compression ratio of 100: 1 or more. The lossy data compression method has a much better data compression ratio than the lossless data compression method. However, since some data are lost, it is impossible to restore all the information. . Therefore, development of a data compression method having a high data compression rate without ultimately losing data is required.

SUMMARY OF THE INVENTION The present invention has been made to solve the above problems, and a first object of the present invention is to provide a lossless data compression method of repeatedly compressing data and receiving new information.

It is a second object of the present invention to provide a lossless compressed data decompression method for decompressing lossless compressed information by iterative circulation.

1 is a flowchart illustrating a lossless data compression process according to the present invention.

2 is a flowchart illustrating an embodiment of a lossless data compression method according to the present invention.

3 is a flowchart illustrating step 240 of FIG. 2 in more detail.

4 is a diagram for explaining the number of cases where values belonging to group 1 and values belonging to group 2 are combined.

5 is a flowchart illustrating another embodiment of a lossless data compression method according to the present invention.

6 is a flowchart illustrating a process of expanding lossless compressed data according to the present invention.

In order to achieve the first object described above, the lossless data compression method according to the present invention includes the steps of: (a) receiving N bits of information and dividing the information into K bits (N>K>1); (b) the representative value of the decided number of predetermined representing at least two values from the values of 2 ^K of the K-bit information which are divided in the step (a) divided by the information bits; (c) to group within, but divided into two or more groups according to frequency, to groups within the low frequency value, including the representative value and number of the frequency value at the of the K bits of the 2 ^K kinds of values of the information Including a value deleted due to the representative value being determined; (d) generating a head bit field for distinguishing each group, and generating a new value by combining one or more values included in two or more groups selected from each group according to the generated head bits; And (e) in the case where the bit reduction is made through the steps (a) to (d), additional information is received from the bitstream to be compressed and combined with the bitstream where the bit reduction is made. If step (d) is repeated and bit saving is not achieved, the N-bit information is kept as it is and new N-bit information is received from the bit stream to be compressed, and the steps (a) to (d) are performed. It characterized in that it comprises the step of repeatedly performing.

In order to achieve the first object described above, another lossless data compression method according to the present invention comprises the steps of: (a) receiving N bits of information and dividing it by K bits (N>K>1); (b) 2 in 2 ^K of values which are information of K bits, divided from the (a) step, but divided into two or more groups based on the frequency in the said second ^K kinds of values of the K bits of the information Assuming that a predetermined number of representative values representing more than one value is specified, the groups to which the values belong to a small frequency include values corresponding to the hypothesized representative values, and the assumed representative values to a group to which many values belong. Including the deleted value due to the determination to make a deviation in the number of branches belonging to each group; (c) generating a first bit field for distinguishing each group, and generating a new value by combining one or more values included in two or more groups selected from each group according to the generated first bit; And (d) in the case where the bit reduction is made through the steps (a) to (c), additional information is received from the bitstream to be compressed and combined with the bitstream where the bit reduction is made. If step (c) is repeated and bit saving is not achieved, the N-bit information is kept as it is and the new N-bit information is received from the bit stream to be compressed, and the steps (a) to (c) are performed. It characterized in that it comprises the step of repeatedly performing.

In order to achieve the above second object, the method for decompressing a lossless compressed bitstream into the original bitstream comprises: (a) taking N bits of information from the compressed bitstream; (b) extracting the recovered X bits from the N bits of information; (c) decompressing the remaining (N-X) bits from the N bits of information into M bits (M> N-X); And (d) continuing the step (b) by changing N to M if the start bit is '0' after the decompression, and continuing the step (a) if the start bit is '1'. It is done.

Hereinafter, the present invention will be described in detail with reference to the accompanying drawings.

Referring to FIG. 1, a process of losslessly compressing a bitstream according to the present invention is as follows.

First, N bits of information are taken from the compression target bitstream (step 100).

Next, N bits of information are compressed into less than N bits of information by a predetermined compression method (step 110). In this case, let the size of the compressed information be X bits (N> X).

Next, it is checked whether there is more information to be compressed (step 120), and if there is no more information to be compressed, the compression process is terminated.

If there is more information to be compressed, new Y bits of information are taken from the bit stream to be compressed (step 130), and new M (M = X + Y) bits are converted into the information of the compressed X bits and the new Y bits. Build and continue with step 110 (step 140). In this case, Y may be the same value as (N-X), but may be different from (N-X) by setting a predetermined rule between the compression process and the decompression process.

Referring to FIG. 2, a process of performing an embodiment of a lossless data compression method according to the present invention is as follows.

The process of taking N bits from a bitstream requiring compression and shortening them by more than one bit is defined as one compression cycle. First, N bits of information are taken from the compression target bitstream (step 200).

When the N bit bitstream to be compressed is divided by K (N>K> 1) bits, the divided K bit values each have one of 2 ^K values of 0 to (2 ^K -1) (210). step). Therefore, to calculate the frequency of 2 ^K of the K bits of the bit stream that correspond to the values of 0 ~ (2 ^K -1) in the compressed bit stream of N-bit destination. In this case, two or more values having the lowest frequency correspond to one representative value and are divided into flags (step 220). Then, to include one or more of the representative value, if needed, by each of the ^K 2 according to the frequency is divided into two or more groups, each group of values (step 230). Next, by scanning the bit stream to be compressed, it creates a head bit field that can distinguish the group to which the value belongs, and scans the head bit field and marks it by a new combined value combining one or more values belonging to each group. The method to reduce the total number of bits is found (step 240).

Hereinafter will be described a method for converting into a 2 ^K kinds of values into three groups.

First, four values having the lowest frequency among the values of 0 to ( ^2K- 1) are taken, and these four values are recorded in the F1 / F2 flag of the information field. The two least frequent values correspond to F1, and the next two less frequent values correspond to F2. Then, (2 ^K-2 -3) values having the least frequency next to the four values among the values of 0 to (2 ^K -1) are classified into Group 1 together with the values corresponding to F1 and F2.

(2 ^K-2 -1) values having the highest frequency among values of 0 to (2 ^K -1) are classified into Group 2, and the remaining 2 ^{K -1} values are classified into Group 3.

Group classification Group3 Group2 Group 1 2 ^K-1 pieces X X (2 ^K-2 -1) F1 F2 (2 ^K-2 -3)

In the grouping of Table 1, the 'X' marks in Group 1 and Group 2 assume that the four values with the lowest frequency correspond to F1 and F2, respectively, so that two values are deleted from Group 1 and Group 2, respectively. I mean.

Step 240 of FIG. 2 is described in detail by FIG.

First, the sum of the frequency of the values selected as the group 1 and the sum of the frequency of the values selected as the group 2 are compared (step 300). That is, in order to achieve bit savings according to the present invention, the following conditions are required.

<Requirement>: The sum of the frequencies of the values selected as Group 1 shall be less than or equal to the sum of the frequencies of the values selected as Group 2.

If the above <condition> is satisfied, it is processed as follows.

2 ^K-1 values of group 3 are separated by the first bit '1', (2 ^K-2 -1) values of group 2 are separated by the first bit '00', and (2 ^K-2) of group 1 The -1) values are separated by the first bit '01' to generate a first bit field corresponding to the order of the N bit bitstream to be compressed.

While generating the first bit field and sequentially scanning the N-bit compression target bitstream, K-bit values are converted as follows (step 310).

In Group 1, the values corresponding to F1 are '0', the values corresponding to F2 are '1', and the remaining values are in order from '2' (2 ^K-2 -2) To. Similarly, the values belonging to group 2 are converted from '0' to (2 ^K-2 -2) in order from the smallest values, and the values belonging to group 3 are ordered from the smallest values Convert from '0' to (2 ^K-1 -1). The converted value of group 1 and the converted value of group 2 can be represented by (K-2) bits, and the converted value of group 3 can be represented by (K-1) bits.

Therefore, the values belonging to each group are all expressed as K bits by adding up the number of each leading bit and the number of bits for the conversion value. In other words,

Group 3: '1' + 2 ^K-1 values,

Group 2: '00' + 2 ^K-2 -1 values (values with high frequency),

Group 1: '01' + 2 ^K-2 −1 values (values with less frequency).

The first bit field is sequentially scanned, and the converted values corresponding to the first bits of the group 3 are left as they are, and the converted values corresponding to the first bits of the group 1 and the group 2 are expressed as a combined value combined (step 320). That is, when the first bit of the first group 1 is encountered, the converted value corresponding to the position of the first group 2 is found, the converted value corresponding to the position of the first bit of the first group 2 is found, and the combined value corresponding to the found two converted values is displayed. do. The mapping of the two transformed and combined values can be used in the compression and decompression processes by setting up a table. At this time, the order in which the transform values belonging to the group 1 and the transform values belonging to the group 2 should be kept constant.

As described above, the reason for combining the transform values of group 1 and the transform values of group 2 will be described.

The conversion value of group 1 is from 0 to (2 ^K-2 -2) to (2 ^K-2 -1), and the conversion value of group 2 is also from 0 to (2 ^K-2 -2) to (2 ^K-2 -1) Thus, the number of cases of combining these converted values is a total of ^{(2 K-2} -1) 2 kinds, that is, ^{(2 2K-4 -2 K-} 1 +1). Therefore, some of the number of cases where the two conversion values are combined are represented by (2K-5) bits,

(2 ^2K-4 -2 ^K-1 +1) -2 ^2K-5 = (2 ^2K-5 -2 ^K-1 +1)

Branches are represented by (2K-4) bits. On the other hand, the number of cases combined with the converted values of F1 and F2 in Group 1 is (2 ^K-2 -1), respectively,

2 × (2 ^K-2 -1) = (2 ^K-1 -2)

It is eggplant.

If (2K-5) bits are assigned when combined with the converted values of F1 and F2 in group 1, the remaining number of cases in which (2K-5) bits can be allocated is

2 ^2K-5- (2 ^K-1 -2) = (2 ^2K-5 -2 ^K-1 +2)

It is eggplant. Therefore, except when combined with the converted values of F1 and F2 in Group 1, the number of cases in which (2K-5) bits can be allocated is exactly one more than the number of cases in which (2K-4) bits are allocated. many. This relationship is shown in FIG.

In this case, when combined with the transformed values of F1 and F2 of Group 1, that is, when the number is indicated by Equation 2, each of the two values corresponding to F1 and F2 for the two bitstreams indicated by the first K bits It takes 1 bit to correctly restore either, 4 bits for the leading bit, and (2K-5) bits to indicate the number of combined cases so that the total number of bits remains unchanged.

In addition, since the number represented by Equation 3 is one more than the number represented by Equation 1, the remaining cases except for this one case represent 4 bits for the first bit and the number of combined cases. It takes one bit to distinguish from the case shown in (2K-5) bits and the equation (1), so that the total number of bits remains unchanged.

In addition, even in the case shown in Equation 1, 4 bits are used for the first bit and (2K-4) bits are required to indicate the number of combined cases, so that the total number of bits remains unchanged.

However, only one of the cases shown in Equation 3 takes 4 bits for the first bit and (2K-5) bits to indicate the number of combined cases, thereby reducing the total number of bits. Therefore, it is advantageous to include the value in the highest frequency flag of the information field and allocate (2K-5) bits for the case with the highest frequency among the cases where F1 and F2 are not included.

If the converted value of group 2 remains after combining the converted value of group 1 and the converted value of group 2 by the above-described method, the remaining converted value of group 2 remains as it is, or the combined values of group 2 are combined to further add bits. A method of saving may be used (step 330). In case of combining only the conversion values of group 2, there is no case of combining with F1 and F2. Therefore, the case of (2 ^K-2 -1) can be selected, and each bit can be reduced by 1 bit.

In this way, a given compression target bitstream can be compressed. If the <condition> is not satisfied, several compression methods are considered simultaneously to find a compression method.

In the first method, when the <condition> is not satisfied, the bitstream is reconstructed using an exclusive logic sum (XOR).

According to this method, an N-bit bitstream is divided by K bits, and the first K-bit value and the second K-bit value following this are exclusively logically summed and the value generated by the exclusive logical sum is added to the second K-bits. Replace with the value of. Subsequently, the value of the second K-bit before the substitution and the value of the third K-bit following it are exclusively replaced with the value of the third K-bit. In this manner, the exclusive logical sum of all the N-bit bitstreams in order is called a first-order exclusive logical sum conversion.

After the first exclusive logic sum, it is checked whether the <condition> is satisfied. If the <condition> is satisfied, the compression is performed by converting the values belonging to the group 1 and the values belonging to the group 2 and combining them, and < If the condition> is not satisfied, the second exclusive logical OR is performed in the same manner. If the conversion is repeated repeatedly, the bitstream that exactly matches the bitstream of the N bits before the first exclusive logical sum conversion will always appear again. The conversion at this time is called the M-order exclusive logical sum conversion, and the order conversion is the limit. (340 steps).

Before the M-order exclusive logical conversion is performed, the exclusive logical sum of a specific bit value promised for all of the N bits must be performed at least once in a promised sequence so that a predetermined flow according to the exclusive logical-conversion of all N-bit values is arbitrarily selected. The conversion increases the probability of obtaining a bit state that meets the condition.

Various methods may also be used for the exclusive logical sum conversion method. For example, a method of exclusive logical sum of only the value of the K bits in the multiples of four and the value of the next K bits may be used.

In this way, information on how many times the conversion is made or whether the random conversion is performed in the middle is displayed on the XOR count flag, the XOR random conversion flag, and the XOR method flag of the information field so that it can be used when restoring the original data. do.

The second method is used when the above <Condition> is not satisfied by the first method, and goes through the following process.

First, the excess number is obtained by subtracting the total frequency of the values belonging to Group 1 by subtracting the total frequency of the values belonging to Group 2, and generating a leading bit field and a transform value for the N-bit compression bitstream (steps 350 and 360). .

Then, after combining the transformed values belonging to Group 1 and Group 2 in a one-to-one, and restoring the original four values corresponding to F1 and F2 to make the new converted value , Group 1 has (2 ^K-2 +1) transform values. Among the transformed values, the two transformed values having the lowest frequency are allocated by expressing a predetermined value of (K-2) bits, and then one more bit is added to distinguish the two transformed values, and the remaining (2 ^{K- 2} -1) conversion values are expressed by allocating different values of (K-2) bits. Therefore, the number of values subtracted by the total frequency of the values belonging to Group 2 from the total frequency of the values belonging to Group 1 is preselected from Group 1, with some selection methods defined and 2 The method with the lowest total frequency of eggplant values is selected (steps 370 and 380).

Here, the information on the excess number and the selection method should be recorded in the excess number flag and the excess sorting flag of the information field, respectively. The excess number flag can be omitted because it can be calculated from the number of each group.

Referring to FIG. 2 again, a process of shortening a given N-bit bitstream by more than one bit in the above-described manner is called one compression cycle. At the end of the compression cycle, it is checked whether there is information to be further compressed, and if there is no further compression target information, the compression process is terminated (step 250).

The compression cycle is successful only if the number of bits saved by compression is one or more than the number of bits used for the information field. If the compression cycle progresses and the above-described methods are not applied, and the bit stream to be compressed cannot be shortened by more than one bit, the first compression cycle is performed with the start bit set to '1' while leaving the N bit stream as it is. Start (step 260, 280). If the compression cycle is successful, input additional information to start the compression cycle again (steps 260 and 270). The start bit is set to '1' at the beginning of the first compression cycle and '0' from the next compression cycle so that the limit in which the compression cycle is repeated upon recovery can be distinguished. It is included in the information field as a start bit.

Hereinafter, a specific number will be described as an example. It is assumed here that ^2K values are classified into two groups.

If the size of the bit stream to be compressed is N = 3,840,000 bits and the bit stream to be divided is divided by the division unit K = 6 bits, 640,000 values are generated. These 640,000 values take one of the 64 values of 0 to 63, and the 64 values each have an average frequency of 10,000, but the corresponding frequencies may be different.

Here, four of the 64 values are selected with the lowest frequency, and the values are assumed to be (20, 14, 55, 8) in order of decreasing frequency. Then, the values 20 and 14 correspond to F1, and the values 55 and 8 correspond to F2. These four values are recorded with 6 × 4 = 24 bits assigned to the information field.

The group 1 together with F1 and F2 is selected by selecting (2 ^(6-1) -3), that is, 29 values having the lowest frequency among the 60 values except the 4 values among the 64 values. The remaining 31 values that are not selected are classified into Group 2 as values having a high frequency.

Group classification Group2 Group 1 X X 31 high frequency values F1 = 20 F2 = 55 29 values of low frequency

In Table 2, 'X' is intended to assume that the values of 14 and 8 corresponding to F1 and F2 are deleted from the 32 values of each group (actually, 31 values are not included in Group 1 Values belong to).

The 3,840,000-bit bitstream to be compressed is scanned by 6 bits, and if the value is 33 values belonging to group 1, the first bit is marked as '1', and if the 31 values belonging to group 2, the first bit is set to '0' 'To create a 640,000-bit leading bit field.

Meanwhile, among the values belonging to Group 1 while generating the first bit field, the values 20 and 14 corresponding to F1 are '0', and the values 55 and 8 corresponding to F2 are '1', and the remaining 29 kinds are selected. The values of are converted into '2' to '30' in order of their magnitude. The values belonging to the group 2 are also converted into '0' to '30' in order of magnitude.

Now, compare the total frequency of the values belonging to group 1 with the total frequency of the values belonging to group 2. First, it is assumed that the total frequency of 33 values having a low frequency belonging to group 1 is less than or equal to the total frequency of 31 values having a high frequency belonging to Group 2. The following procedures are defined as "Method 1".

By putting group 1 at the head, the values belonging to group 1 and the values belonging to group 2 are processed one by one. In other words, while scanning 640,000 values of the first bit field, the bit values of '1' and '0' are sequentially processed one by one. Since there are 31 values in each group, the values generated by combining two values The number of cases is 31 x 31 = 961 types.

By the way, it was originally expressed as two 6-bit values, that is, 12 bits. Since the first bit corresponding to each converted value used 2 bits, the number of 961 cases was less than 10 bits to represent each combined value. You can express it. The number of these 961 cases includes the number of cases combined with F1 and F2.

The 961 cases are expressed as shown in FIG. 4 as follows.

Among the 961 cases, 31 cases combined with F1 and 31 cases combined with F2 are distinguished first.The remaining 899 cases are again 450 and 449 cases. Separate by number of cases.

In this way, the number of 62 cases combined with F1 and F2 is represented by 9 bits. At this time, for the number of 62 cases combined with F1 and F2, the original value of the two values 20 and 14 coupled to F1 is restored and the original value of the two values 55 and 8 coupled to F2. Since 1 bit must be added to recover the data, a total of 10 bits are used as it is to represent the combined value. Of the remaining 899 cases, 898 except for one case are divided into two classes of 449, and the remaining 9 bits are used except for the 9-bit value used for the 62 cases combined with F1 and F2. A total of 10 bits are used to represent the combined value since each value is assigned and one bit is used to distinguish the class.

However, only one remaining case is represented by 9 bits, thereby saving 1 bit. Therefore, the value of the most frequent number among the 899 cases is expressed as 9 bits in this case, and the value is recorded by assigning 9 bits to the highest frequency flag of the information field.

Thus, the combined value of the two groups combined after the leading bit field 640,000 bits is assigned by 10 bits, respectively, except that the combined value is assigned with 9 bits. Therefore, in the decompression process of the compressed data, the first 9 bits compare the combined value with the 9 bits of the information field maximum frequency flag, and restore only 9 bits in the same case as it is.

If the total frequencies of the values belonging to the two groups are the same, the process is completed by the method as shown in FIG. However, if the total frequency of the values belonging to group 2 is greater than the total frequency of the values belonging to group 1, 31 converted values are expressed using 5 bits for the excess values of group 2, except in group 2 Additional bit savings can be achieved by representing the most frequent value as 4 bits of '1111'.

The structure of the bitstream converted by the above method is shown in Table 3.

Structure of Converted Bitstream Group 1 Group 2 Group 3 4th group Bitstream Info Field Lead bit field (640,000 bits) Combined value of group 1 and group 2 Variant of Unjoined Group2 Restoration method Read 9 bits first, then read 1 additional bit based on the result Read 4 bits first, then read 1 additional bit based on the result

In Table 3, the number of bits is reduced by the frequency of one case represented by 9 bits in group 3. If the value is larger than the number of bits required for the group 1 information field, the number of bits in the converted bitstream is the first N bits. Less than a number. One such cycle is completed, and the new cycle is started again by accepting the new bits as many as the saved bits.

On the other hand, when the total frequency of the values belonging to the group 1 is greater than the total frequency of the values belonging to the group 2, that is, when the <condition> is not satisfied, the following various methods are compressed.

The following describes the procedures for "Method 2".

According to "Method 2", the bit stream of 3,840,000 bits is divided into 6-bit units and converted by exclusive logic sum.

In this manner, in the embodiment of the present invention, in the first method, the value of the odd 6-bit unit is exclusively summed with the value of the next even 6-bit, and the value of the even 6-bit is replaced. Adopts the method of performing the exclusive logic on the first two 6-bit values to replace the latter value, and then performing the exclusive logic on the next 6-bit value with the replaced 6-bit value to replace the next 6-bit value. do.

In this case, in case of using the two methods, it checks whether the <Condition> is suitable after the Exclusive Logic Conversion, records this state in the information field if it meets the <Condition>, and uses one compression cycle using "Method 1". To complete. However, if it does not meet <condition>, the exclusive logical sum conversion continues until it meets <condition> based on the converted value, and the number of repetitions of the transformation is limited to 255 or less so that the XOR flag of the information field is Decide to have 8 bits.

The XOR random conversion flag is a flag for arbitrarily converting the value of N bits by causing the promised bit value to be exclusive logically summed with N bits after the promised exclusive logic sum count without continuing exclusively 255 times. In this case, it is conducted once after 10 exclusive logic sums. Therefore, if the XOR count flag value is 10 or less, the exclusive logical sum conversion may not be performed.If the XOR arbitrary conversion flag value is 0, the exclusive logical sum conversion is not performed.If the 1 is 10, the exclusive logical sum conversion is performed immediately after 10 times. Handle

If <condition> is met before 255 iterations are made, the status is recorded in the information field and compressed through "Method 1".

Therefore, in the embodiment of the present invention, 8 bits for the XOR number flag in the information field, 1 bit for the XOR arbitrary conversion flag, and 1 bit for the XOR method flag are allocated (at this time, if the value of the XOR number flag is 0, Since no XOR conversion has occurred, the bit allocation for the XOR method flag is omitted).

The following describes the procedures for "Method 3".

If the <Condition> is not satisfied even by "Method 2", the number of values exceeding the total frequency of the values belonging to the group 2 from the total frequency of the values belonging to the group 2 is selected and extracted from the values belonging to the group 1 .

First, the number is indicated on the excess number flag of the information field so as to know how many are exceeded (in the embodiment of the present invention, 15 bits are allocated for the excess number flag because it does not exceed 20,000). Then, the first bit field is scanned to find the conversion values corresponding to the positions of the values indicated by '1' and to select values determined by one of the following screening methods.

As the sorting method used in the embodiment of the present invention

First, choose a multiple of 2 from the beginning, but only overtime.

Second, choose a multiple of 3 from the beginning, but only the excess number.

Third, select the multiple of 5 from the beginning, but only the excess number.

Fourth, select a multiple of 7 from the beginning, but select only the excess number.

Fifth, the number of consecutive selections from the beginning.

Sixth, successive selections after the value determined in the fifth method above.

Seventh, the number of consecutive selections after the value determined in the sixth method above.

Eighth, successive selections after the value determined in the seventh method above.

The criteria for selecting one of the eight screening methods described above is to recover both F1 and F2 and change it to 33 values, and then select the method with the smallest total frequency of the 2 of the 33 values. To choose. Accordingly, the information field is allocated 15 bits for the excess number flag and 3 bits for the excess sorting flag.

After the selection, if the sum of the total frequency and the number of information bits of the two least frequent values among the values selected by the above method is less than the number of bits saved by "Method 1," Group 1 is not selected. Since the total number of transform values belonging to and group 2 is the same, the transform value belonging to group 1 and the transform value belonging to group 2 are combined and processed by "Method 1", respectively, and the selected transform values are represented by 5 bits. However, two values with the least frequency are indicated by 5 bits of the same value and 1 bit is added to distinguish.

Looking at the configuration of the information field used in the embodiment of the present invention as described above are as follows.

▼ Start bit (1 bit): Set to '1' to start the first compression cycle and to '0' from the next compression cycle.

F1 / F2 flag (24 bits): Records four values represented by F1 and F2 (6 bits x 4).

▼ Group 1 selection flag (60 bits): It displays the status of 60 values classified by group except 4 values represented by F1 and F2.

▼ Maximum Frequency Flag (9 bit): When the combined value of Group 1 and Group 2 except F1 and F2 is combined, record the combined value with the highest frequency.

▼ XOR number flag (8 bits): Records the number of exclusive logical sum conversion. If the value is 0, it means that the exclusive logical sum is not performed.

▼ XOR method flag (1 bit): Record which XOR method was used. Not assigned if the value of XOR flag is 0.

▼ Overnumber flag (15 bits): If the total frequency of the values belonging to group 1 is large, the excess frequency is recorded. Not assigned if the value of XOR flag is 0. This flag can be omitted because it can be seen from the number of each group.

▼ Over-screen flag (3 bits): Record how the group 1 selected the excess number recorded in the excess number flag. Not assigned if the value of the excess number flag is zero.

▼ XOR Arbitrary Conversion Flag (1 bit): Indicates whether or not the number of exclusive logic sums is performed with the promised value in the number of promises. For example, if the value of the flag is '1', two bits may be allocated to indicate the number of times.

The total number of bits saved by completing one compression cycle through the above process is displayed separately only the values belonging to Group 2 when the frequency of occurrence represented by 9 bits in "Method 1" and the total frequency of Group 2 are large. The total number of bits consumed is the sum of the number of information bits and the total frequency of the two values with the lowest frequency among the values selected in "Method 3".

Therefore, if the total number of bits saved by one or more is larger than the total number of bits consumed, the total number of bits displayed in one compression cycle is reduced by the number of bits, so that a new compression target bit corresponding to the number of bits is accepted and the compression cycle is performed again. Repeat. If the total number of bits saved is less than or equal to the total number of bits consumed, the N-bit bitstream is kept as it was before the last compression cycle started, and the new N-bit bitstream is accepted and the first compression cycle is started again. .

Meanwhile, the method of using two representative values of F1 and F2 to represent a specific value has been described so far.How to set one representative value of F1 and three representative values of F1, F2 and F3 or more In the method of setting the representative value, the compression process can be performed similarly to the above.

Referring to FIG. 5, another embodiment of the lossless data compression method according to the present invention is based on the following process.

Steps 500 to 510 and steps 530 to 570 except for step 520 of FIG. 5 are processed similarly to steps 200 to 210 and 240 to 280 of FIG. 2.

However, in step 520, one group includes less than 2 ^K-2 branches by including one or more values among the 2 ^K-2 branches, the group reference number, without using the representative value. Some groups, including more than 2 ^K-2 branches, contain variations in the baseline number of groups. That is, but divided into two or more groups based on the frequency in the 2 ^K of the the K-bit values, the predetermined number of representatives of which represents at least two values from the 2 ^K of values which have a K-bit information Assuming that the value is set, each group of low frequency values contains values corresponding to the hypothesized representative value, and a group of high frequency values includes the deleted value because the hypothesized representative value is determined. The number of branches in the group is varied. Accordingly, in step 540, a given bitstream may be compressed by combining two or more changed groups in pairs.

6, the decompression process of the lossless compressed data according to the present invention is as follows.

Basically, the data decompression process may restore the original bitstream using various flags of the information field generated in the compression process in the reverse order of the compression process.

First, N bits of data are taken from the compressed bitstream (step 600).

Next, by using the information bit field included in the N-bit data, the X-bit that has already been restored from the N-bit information is extracted and stored (step 610), and the remaining (NX) bits in the N-bit data are M bits (M > NX) (step 620).

After the data decompression, if the start bit is '0', M is regarded as N, and step 610 is continued. If the start bit is '1', it is checked whether there is more compressed data to decompress (steps 630 to 650). If there is more compressed data to be decompressed, the decompression process is continued by receiving the new decompression bitstream in step 600. Otherwise, the decompression process of the lossless compressed data is terminated.

According to the present invention, data can be losslessly compressed at a high compression rate, and the losslessly compressed data can be easily extended.

Claims (10)

In the method of lossless compression of a bitstream, (a) taking N bits of information from the bitstream; (b) compressing the N-bit information into less than N bits of X bits (N> X); (c) taking new Y bits of information from the bitstream; (d) constructing new M (M = X + Y) bits with compressed X bits of information and new Y bits of information; And (e) repeating steps (b) to (d) until there is no new information to take in step (c). (a) receiving N bits of information and dividing the information into K bits (N> K> 1); (b) the representative value of the decided number of predetermined representing at least two values from the values of 2 ^K of the K-bit information which are divided in the step (a) divided by the information bits; (c) to group within, but divided into two or more groups according to frequency, to groups within the low frequency value, including the representative value and number of the frequency value at the of the K bits of the 2 ^K kinds of values of the information Including a value deleted due to the representative value being determined; (d) generating a head bit field for distinguishing each group, and generating a new value by combining one or more values included in two or more groups selected from each group according to the generated head bits; And (e) In the case where the bit reduction is made through the steps (a) to (d), additional information is received from the bitstream to be compressed and combined with the bitstream in which the bit reduction is made. If step d) is repeated and bit saving is not achieved, the N-bit information is kept as it is and the N-bit information is newly received from the bit stream to be compressed to perform steps (a) to (d). And repeatedly performing the data compression method. (a) receiving N bits of information and dividing the information into K bits (N> K> 1); (b) 2 in 2 ^K of values which are information of K bits, divided from the (a) step, but divided into two or more groups based on the frequency in the said second ^K kinds of values of the K bits of the information Assuming that a predetermined number of representative values representing more than one value is specified, the groups to which the values belong to a small frequency include values corresponding to the hypothesized representative values, and the assumed representative values to a group to which many values belong. Including the deleted value due to the determination to make a deviation in the number of branches belonging to each group; (c) generating a first bit field for distinguishing each group, and generating a new value by combining one or more values included in two or more groups selected from each group according to the generated first bit; And (d) In the case where the bit reduction is made through the steps (a) to (c), additional information is received from the bitstream to be compressed and combined with the bitstream where the bit reduction is made. Repeat step c), and if no bit reduction is made, the N-bit information is kept as it is and new N-bit information is received from the bit stream to be compressed to perform steps (a) to (c). And repeatedly performing the data compression method. (a) accepting the information of N bits divided into K bits, (N>K> 1), obtaining the frequency of the K-bit information corresponding to 2 ^K of values; (b) wherein (a) represented by the two values the low frequency to F1 in 2 ^K of values of the phase and, the frequency is to include in the group 1 with the F1 values of a number of small predetermined, a number of frequency by including the predetermined number of values of the second group, comprising: divided into two or more groups based on the frequency of the 2 ^K kinds of values; (c) combining the values belonging to the group 1 and the values belonging to the group 2 in order to convert them into combined values; And (d) In the case where the bit reduction is made through the steps (a) to (c), additional information is received from the bitstream to be compressed and combined with the bitstream where the bit reduction is made. Repeat step c), and if no bit reduction is made, the N-bit information is kept as it is and new N-bit information is received from the bit stream to be compressed to perform steps (a) to (c). And repeatedly performing the data compression method. The method of claim 4, wherein step (c) (c.1) comparing the magnitude of the total frequency of the values belonging to group 1 with the total frequency of the values belonging to group 2; (c.2) If the total frequency of the values belonging to group 2 is greater than or equal to the total frequency of the values belonging to group 1 in step (c.1), the values belonging to group 1 and the values belonging to group 2 are selected. Combining them in turn to convert them into combined values; And (c.3) If the total frequency of the values belonging to the group 2 is less than the total frequency of the values belonging to the group 1 in step (c.1), the step (c.1) by converting the bit configuration to recover And retrying a comparison of the data compression method. The method of claim 5, wherein step (c.2) (c.2.1) generating a head bit field indicating a group to which the value belongs in order of N bits of information divided by K bits, and converting each value into a new conversion value given for each group to which the value belongs; (c.2.2) combining one transform value belonging to group 1 and one transform value belonging to group 2 into pairs and converting the result into a new combined value; And (c.2.3) data compression, comprising assigning and expressing a bit number of bits less than other transform values for the transform value having the lowest frequency among the redundant transform values belonging to Group 2 Way. The method of claim 6, wherein the conversion value 0 for the values represented by F1 in Group 1, one or more values mapped to the order of magnitude of the remaining values, and 0 mapped to the order of magnitude for the values belonging to Group 2. The data compression method characterized by the above value. The method of claim 5, wherein if the conversion of the bit configuration of step (c.3) is repeated more than a predetermined number of times, the comparison of step (c.1) is stopped and performed. (c.4.1) obtaining an excess number by subtracting the total frequency of the values belonging to Group 2 from the total frequency of the values belonging to Group 1; (c.4.2) generating a first bit field indicating a group to which the value belongs by sequentially converting N bits of information divided by K bits, and converting each value into a new conversion value given for each group to which the value belongs; (c.4.3) The excess number of transformed values among the transformed values belonging to Group 1 are selected, but the total frequency of the two transformed values having the lowest frequency after restoring F1 to the corresponding original value is lowest. Selecting the transform values to include; (c.4.4) converting one transform value from one of the transform values belonging to group 1 and not selected in step (c.4.3) and one transform value belonging to group 2 into pairs to convert to a new combined value; ; And (c.4.5) converting the transformed values selected in the step (c.4.3) into a new transformed value, and then adding an additional 1 bit for the two transformed values having the least frequency. The method of claim 1, further comprising a step. The method of claim 4, wherein the data information compressed through the steps (a) to (c) Data having a value of '1' when the first N bits of information are received, and a start bit having a value of '0' when the compression target information is additionally received when bit reduction is made. Compression method. A method of extending a lossless compressed bitstream to an original bitstream, (a) taking N bits of information from the compressed bitstream; (b) extracting X bits already restored from the N bits of information; (c) decompressing the remaining (N-X) bits from the N bits of information into M bits (M> N-X); And (d) after decompression, if step M is a predetermined value, consider M as N and continue with step (b); if step A is different from the predetermined value, continuing step (a); Lossless compressed data decompression method characterized in that.