KR100224686B1 - Method for storing mpeg decoded image data - Google Patents

Abstract

Disclosed is a method of storing MPEG restored image data. A method of storing MPEG-reconstructed image data consisting of a plurality of horizontal pixels comprises: an initial step of dividing horizontal pixels into a pixel group consisting of N (where N is a natural number of two or more) pixels, each adjacent pixel of the pixel group A calculation step of sequentially calculating the differences between the two to obtain N-1 differences, an operation step of obtaining N-1 pixels computed corresponding to the levels of the N-1 differences, and the calculated N-1 pixels If it is within the range, the result from the combining step and the combining step of sequentially combining the first pixels among the N pixels in the pixel group, the calculated N-1 pixels, and the characteristics of the N-1 pixels are emmpted. And storing the data as compressed data of the reconstructed image data, and compressing and storing the pixels by using similarities between adjacent pixels, thereby reducing the memory size. It has the effect of reducing the dragon.

Description

MPEG storage method

The present invention relates to a method for storing data, and more particularly, to a method for storing restored MPEG (Motion Picture coding Experts Group) data.

In general, in the case of data transmission, in order to overcome the limited storage medium capacity and the transmission channel capacity and to transmit a large amount of data per hour, the data to be transmitted is compressed at the transmitting side and the compressed data is recovered at the receiving side.

Recently, video signals are compressed and reconstructed in a digital manner based on the MPEG standard. MPEG decoder uses an external memory because it processes a large amount of data during the restoration process. However, when storing the image data reconstructed through the MPEG decoder in an external memory, the intra (I: Intra) picture, the P (Predictive) picture, and the bidirectionally predictive (B: bidirectionally predictive) picture are all stored due to the nature of the decoder. Since both the luminance signal and the chroma signal are stored per picture, a considerable amount of storage capacity is required. In other words, assuming that all I pictures, P pictures, and B pictures are stored, horizontal pixels 720 * vertical pixels 480 * pictures 3 * luminance signals and chroma signals 2 * bits per pixel 8 = 15.82 Mbits. Therefore, since a large amount of memory must be used to store the restored data, there is a costly problem.

An object of the present invention is to provide a method for storing MPEG reconstructed image data in which pixels are compressed and stored by using correlation between adjacent pixels.

1A and 1B are diagrams for describing a compression process of image data when adjacent pixels have similarities.

2 (a) and 2 (b) are diagrams for explaining a restoration process of image data stored through the compression process of FIG.

3A and 3B are diagrams for describing a process of compressing image data when at least one of adjacent pixels does not have similarities.

FIG. 4 is a diagram for describing a restoration process of image data stored through the compression process of FIG. 3.

5 is a flowchart illustrating a method of storing MPEG restored image data according to the present invention.

In order to achieve the above object, the method for storing MPEG restored image data according to the present invention includes a method for storing MPEG restored image data including a plurality of horizontal pixels, wherein horizontal pixels are N (where N is a natural number of 2 or more). An initial step of dividing into pixel groups consisting of two pixels, a step of sequentially calculating the differences between adjacent pixels of the pixel groups to obtain N-1 differences, and respectively calculating corresponding to the levels of the N-1 differences Calculating the N-1 pixels, if the calculated N-1 pixels are within a predetermined range, the first pixel of the N pixels of the pixel group, the N-1 pixels, and the N-1 pixels It is preferable that the combination step of sequentially combining the bits indicating the characteristics of the characteristics and the storage step of storing the results from the combination step as compressed data of the MPEG-reconstructed video data.

Hereinafter, a method of storing MPEG restored image data of the present invention will be described with reference to the accompanying drawings.

MPEG reconstructed image data (hereinafter, simply referred to as image data) has similarities between adjacent pixels. Having similarity reduces the difference between adjacent pixels within a predetermined range. The storage method of the image data of the present invention utilizes this feature.

1A and 1B are diagrams for explaining a compression process of image data according to the present invention when adjacent pixels have similarities, and FIG. 1A shows a difference between adjacent pixels. FIG. 1B is a diagram illustrating a process of performing a predetermined operation according to the level of FIG. 1, and FIG. 1B illustrates the format of a compressed pixel group.

Before storing image data in an external memory, the image data is divided into predetermined pixel units in a horizontal direction and temporarily stored in a buffer. The format shown in (a) of FIG. 1 represents a pixel group obtained by dividing into eight pixel units. Here, since one pixel is 8 bits, one pixel group becomes 64 bits, and eight pixels are temporarily stored in a 64-bit buffer having an address of 0 to 63.

For the format shown in Fig. 1A, the difference between adjacent pixels is calculated from the eight temporarily stored pixels, and a predetermined operation is performed according to the level of the difference. The pixel value of each pixel shown in FIG. 1A is, for example, that when the adjacent pixels have similarity, the difference between the pixels is reduced to a range of -48 or more and 47 or less, and is -48 or more and 47 or less. These are the values.

First, subtracting the second pixel value from the first pixel value results in 128-125 = 3. Thus, if the difference between the two values is within the range of -16 to 15, the difference is left as it is. Next, subtracting the third pixel value from the second pixel value results in 125-157 = -32. In this way, if the difference between the two values is within the range of -32 to -17, 16 is added to be modified to 125-157 + 16 = -16. Next, subtracting the fourth pixel value from the third pixel value gives 157-126 = 31. Thus, if the difference between the two values is in the range of 16 to 31, 16 is subtracted and transformed into 157-126-16 = 15. Next, subtracting the fifth pixel value from the fourth pixel value yields 126-174 = -48. Thus, if the difference between the two values is in the range of -48 to -33, 32 is added to 126-174 + 32 = -16. Subsequently, subtracting the sixth pixel value from the fifth pixel value, 174-180 = -6. At this time, since the difference between the two values is in the range of -16 to 15 as described above, the difference is left as it is. Next, subtracting the seventh pixel value from the sixth pixel value results in 180-133 = 47. Thus, if the difference between the two values is in the range of 32 to 47, 32 is subtracted and transformed into 180-133-32 = 15. Finally, subtracting the eighth pixel value from the seventh pixel value, 133-153 = -20. At this time, since the difference between the two values is in the range of -32 to -17 as described above, 16 is added to transform 133-153 + 16 = -4.

That is, similarity between adjacent pixels, but once the difference between adjacent pixels is greater than or equal to -48 and less than or equal to 47, in order to reduce the pixels having 8 bits to 5 bits, the difference between adjacent pixels is modified to a value of -16 or more and 15 or less. do. As a result, if the difference between adjacent pixels is -48 to -33, 32 is added, -32 to -17 is added 16, -16 to 15 is not modified, 16 to 31 is reduced to 16, and 32 to 47 is added. Then subtract 32.

The result of performing the above-mentioned predetermined operation is temporarily stored in the 56-bit buffer again. The first 8 bits store the original first pixel in the format shown in FIG. 1A, the next 35 bits sequentially store seven pixels compressed to 5 bits, and the next 13 bits to 5 bits. The characteristics of the seven pixels are sequentially stored. Here, 7 bits of 13 bits sequentially store the reference bits of 7 pixels compressed to 5 bits, the next 5 bits sequentially store the additional bits of the reference bits, and the last 1 bits store the identification bits. .

Specifically, 128, the first pixel value, is stored in addresses 0 to 7. Next, '0' is stored at address 43 while the modified second pixel value 3 is stored at addresses 8-12. Here, '0' is a reference bit indicating that 3 is a value in which the difference between adjacent pixels is maintained. Next, '1' is stored at address 44 while the modified third pixel value -16 is stored at addresses 13 to 17. Here, '1' is a reference bit indicating that -16 is a value modified from a difference between adjacent pixels. In the same manner, '0' or 'at addresses 45 to 49' are sequentially stored in the remaining bits of the fourth to eighth pixel values 15, -16, -6,16 and -4 in the order of 5 bits at addresses 8 to 42. 1 'is stored sequentially.

Further, additional bits are stored in addresses 50 to 54 according to the operational characteristics of the reference bits of '1' stored in addresses 43 to 49. That is, '0' is stored as an additional bit when 16 is added or subtracted from the difference between adjacent pixels, and '1' is stored as an additional bit when 32 is added or subtracted from the difference. Finally, an identification bit of '1' is stored at address 55 to indicate that compression is performed using similarity of adjacent pixels.

Therefore, the image data is divided into a 64-bit pixel group consisting of eight pixels, and the divided pixel group is compressed into a 56-bit pixel group and stored in an external memory.

2A and 2B are diagrams for describing a restoration process of image data stored through the compression process of FIG. 1, and FIG. 2A illustrates a process of restoring compressed pixels. 2B is a diagram illustrating a format of a reconstructed pixel group.

When reading image data stored in the external memory through the compression process, the data is read in 56 bit units and temporarily stored in the buffer. When restoring the compressed 56-bit video data, first check the bit value stored at the last address 55 of the 56 bits. If the bit stored at address 55 is '1', it indicates that the image data is compressed using the similarity of adjacent pixels. Therefore, if the compression process described with reference to FIGS. 1A and 1B is applied in reverse, do.

In the format shown in (a) of FIG. 2, 128, the first pixel value stored at addresses 0 to 7, is restored as it is. The second to eighth pixel values stored in the remaining addresses 8 to 42 in 5-bit units are restored by performing a predetermined operation according to the respective bits read from the addresses 43 to 55. Specifically, if any of the reference bits read from addresses 43 to 49 is '0', only the next pixel value is subtracted from the immediately restored pixel value. However, if any reference bit is '1', an additional bit corresponding thereto is further read out from addresses 50 to 54. If the read-out bit is '0', 16 is added (if the next pixel value is negative) or subtracted (if the next pixel value is positive) after subtracting the next pixel value from the immediately restored pixel value. On the other hand, when the read-out bit is '1', after subtracting the next pixel value from the immediately restored pixel value, 32 is added (if the next pixel value is negative) or subtracted (if the next pixel value is positive). do.

When the above-described restoration process is described by substituting the numerical value shown in FIG. 2A, first, the first pixel value 128 is restored as it is. Next, 125 is restored by subtracting 3, which is the second pixel value, from 128 in accordance with the reference bit '0' read from the address 43. Next, the third pixel value -16 is subtracted from 125 in accordance with the reference bit '1' read from the address 44 and the additional bit '0' read from the address 50, and then 16 is added to restore 157. Next, the fourth pixel value 15 is subtracted from 157 in accordance with the reference bit '1' read from the address 45 and the additional bit '0' read from the address 51, and then 16 is subtracted to restore 126. Next, after subtracting -16, the fifth pixel value -16 from 126, 32 is subtracted to restore 174 according to the reference bit '1' read from the address 46 and the additional bit '1' read from the address 52. Next, 180 is restored by subtracting the sixth pixel value -6 from 174 in accordance with the reference bit '0' read from address 47. Next, after subtracting 15 from 180 to 15 in accordance with the reference bit '1' read from the address 48 and the additional bit '1' read from the address 53, 32 is subtracted to restore 133. Next, after subtracting -4 from 133 in accordance with the reference bit '1' read from the address 49 and the additional bit '0' read from the address 54, 16 is added to restore 153.

The reconstructed pixel values are sequentially restored in units of 8 bits, as shown in FIG. The reconstructed pixel values are the same as the pixel values of the pixel group shown in FIG.

3A and 3B are diagrams for explaining a process of compressing image data when at least one of adjacent pixels does not have similarities, and FIG. 3A illustrates a difference between adjacent pixels. Is a diagram illustrating a process of obtaining a C, and FIG. 3B illustrates a format of a compressed pixel group.

As described with reference to FIG. 1A, the image data is divided into eight pixel units in the horizontal direction and temporarily stored, and the difference between adjacent pixels is calculated. FIG. 3A shows a fifth pixel value different from that of FIG. 1A. Subtracting the fifth pixel value from the fourth pixel value results in 126-175 = -49, and this value is not applied to a predetermined operation when the difference between adjacent pixels is -48 or more and 47 or less. Therefore, it is not possible to reduce an 8-bit pixel to 5 bits.

As a result, if at least one of the differences between adjacent pixels is less than -48 or greater than 47, i.e., if there is no similarity between adjacent pixels, it cannot be compressed using the similarity of adjacent pixels. If there is no similarity between adjacent pixels, as shown in FIG. 3 (b), simply divide each pixel value by 2 to sequentially store the divided result, and separate reference bits and additional bits. Do not save.

Here, all the divided results can be stored in 7 bits. However, in order to indicate that the division is performed by dividing by 2, the last bit stored at address 55 is arbitrarily set to '0'. In FIG. 3A, the final pixel value 154 divided by 2 is actually 77, but as shown in detail in FIG. 3B, 76 is stored by arbitrarily replacing the least significant bit with '0'.

Therefore, even if compression is performed without using the similarity between adjacent pixels, a 64-bit pixel group consisting of eight pixels is compressed and stored into a 56-bit pixel group.

In addition, in the compression process of FIGS. 1A and 1B, even if the difference between all adjacent pixels in the pixel group consisting of eight pixel units is -48 or more and 47 or less, the 64-bit pixel group is 56. If it is not possible to compress the pixel group of bits, it divides by 2 and compresses it. That is, except for the identification bit, the sum of 8 bits occupied by the original first pixel, 35 bits occupied by the compressed pixels, 7 bits occupied by the reference bits, and 5 bits occupied by the additional bits exceeds 55 bits. Then, the compression is stopped using the similarity of adjacent pixels, and the division is again divided by 2 to compress.

FIG. 4 is a diagram for describing a restoration process of image data stored through the compression process of FIG. 3. When restoring the compressed data, check the bit value stored at the last address 55 of the 56 bits. If the bit stored at the address 55 is '0', it indicates that the video data is divided by 2 and compressed.

Multiplying each pixel of the pixel group compressed by 56 bits shown in FIG. 3B yields a reconstructed pixel group as shown in FIG. In this case, the data reconstructed after dividing the image data by 2 and compressing again differs by about 1 or 2 bits from the image data before compression, but such an error has little effect on using the reconstructed data. In practice, the image data has many similarities between adjacent pixels, so the probability of dividing each pixel by two is very low.

5 is a flowchart illustrating a method for storing MPEG restored image data according to an embodiment of the present invention, wherein an initial step (step 10) of dividing image data into N pixels in a horizontal direction is performed. An operation step (steps 12 to 28) for performing a predetermined operation according to the level; a first compression step (steps 30 and 32) for compressing the result of the operation step if there is similarity between adjacent pixels; If there is no similarity between them, or if the result of the first compression step does not satisfy a predetermined condition, the second compression step (steps 30, 34 and 36) and the first step of compressing the N pixels in the initial step are performed. Or storage steps (steps 42, 44 and 46) for storing the result of the second compression step.

Specifically, first, the MPEG-reconstructed image data is divided into N pixels in the horizontal direction to form an initial pixel group (step 10). Taking (a) of FIG. 1 described above as an example, N is 8.

After the tenth step, the difference between adjacent pixels in the initial pixel group is sequentially calculated (step 12). After the twelfth step, a predetermined operation is performed according to the level of the difference. First, it is determined whether the difference is within the first range (step 14). After step 14, if the car is within the first range, the car is kept as it is (step 20), otherwise it is determined whether the car is within the second range (step 16). For example, if a pixel of 8 bits is to be reduced to a pixel of 5 bits, the first range is -16 to 15. Accordingly, if the difference between pixels having 8 bits is within the first range, the difference may be represented as 5 bits. After step 16, if the car is within the second range, the first predetermined number is added / subtracted to the car (step 22); otherwise, it is determined whether the car is within the third range (step 18). For example, the second range is -48 to -33 or 32 to 47, the first predetermined number is 32, and the difference can be represented by 5 bits by adding / subtracting the first predetermined number from the difference in the second range. After step 18, if the car is within the third range, the second predetermined number is added / subtracted to the car (step 24); otherwise, the car is kept as it is (step 20). For example, the third range is -32 to -17 or 16 to 31, the second predetermined number is 16, and the difference can be represented by 5 bits by adding / subtracting the first predetermined number from the difference in the third range.

After any one of steps 20 to 24, it is determined whether the difference is the N-1th difference (step 26). That is, it is checked whether the difference of N-1 is obtained by sequentially calculating the difference between adjacent pixels from the initial N pixels. If it is not the N-th difference, the process returns to step 12 and steps 12 to 24 are performed again. Otherwise, N-1 pixels respectively calculated according to the level of the N-1 differences are obtained (step 28).

After step 28, it is determined whether at least one of the N-1 pixels respectively calculated according to the level of the N-1 differences is out of the first range (step 30). If the initial N pixels have similarities between adjacent pixels, the N-1 pixels respectively calculated according to the level of the N-1 differences may be represented by 5 bits. In operation 30, when at least one of the calculated N-1 pixels is out of the first range, the N pixels constituting the initial pixel group are divided by 2 (operation 36). The image data is stored as compressed data (step 38).

In operation 30, if all of the calculated N-1 pixels can be represented by 5 bits, a reference representing characteristics of the first pixel, the calculated N-1 pixels, and the N-1 pixels among the initial N pixels is shown. The bit parts are combined (step 32). After step 32, it is determined whether the total number of combined bits exceeds a predetermined bit (step 34). If the combined total bits are predetermined bits, the combined result is stored as compressed data of the image data (step 38). Otherwise, the N pixels that constituted the initial pixel group are each divided by two (step 36), and the divided result is stored as compressed data of the image data (step 38). That is, even though all of the calculated N-1 pixels may be represented by 5 bits, if all the bits including the reference bit portion exceed a predetermined bit, they are not compressed and stored using the calculated N-1 pixels.

The above-described method for storing MPEG reconstructed image data of the present invention reduces image data by approximately 2 megabits. Conventionally, if I, P, and B pictures are all stored, horizontal pixels 720 * vertical pixels 480 * pictures 3 * luminance and chroma signals 2 * bits per pixel 8 = 15.82 megabits of storage Do. However, using the method of storing the MPEG-reconstructed image data of the present invention, since the storage capacity of one pixel per eight pixel units can be saved in the horizontal direction, the storage capacity of the entire horizontal pixels 720 to 90 pixels Can save. That is, 90 * 480 * 3 * 2 * 8 = 2,073,600 bits = 1.977 megabits ≒ 2 megabits of memory capacity can be saved.

As described above, the method for storing MPEG reconstructed image data according to the present invention has the effect of reducing the memory size and reducing the cost since the pixels are compressed and stored using the similarity between adjacent pixels.

Claims (7)

In the method for storing MPEG restored image data consisting of a plurality of horizontal pixels, An initial step of dividing the horizontal pixels into a pixel group consisting of N pixels, where N is a natural number of two or more; A calculation step of sequentially calculating differences between respective adjacent pixels of the pixel group to obtain N-1 differences; Obtaining N-1 pixels respectively calculated according to the levels of the N-1 differences; If the calculated N-1 pixels are within a predetermined range, the first pixel of the N pixels of the pixel group, the calculated N-1 pixels, and bits representing characteristics of the N-1 pixels are sequentially Combination step of combining; And And storing the result from the combining step as compressed data of the MPEG-reconstructed image data. The method of claim 1, wherein the calculating step, A first calculation step of adding a first predetermined number if the difference is within a first range, and subtracting the first predetermined number if the difference is within a second range; A second operation step of adding a second predetermined integer greater than the first predetermined number if the difference is within a third range including a lower limit value, and subtracting the second predetermined number if the difference is within a fourth range including an upper limit value; A third operation step of maintaining the difference as it is if the difference is between the first range and the second range; And A fourth operation step of maintaining the difference as it is, if the difference is smaller than the third range and larger than the fourth range, The range in which the difference may belong is subdivided into the first to fourth ranges, and the subdivided ranges have a relationship between the third range, the first range, the second range, and the fourth range, and the difference is the lower limit and the fourth range. If the value between the upper limit value is reduced to a value between the first range and the second range of the MPEG-restored image data storage method. The method of claim 2, wherein the combining step, If all of the calculated N-1 pixels are the results of any one of the first to third operation steps, the first pixel and the calculated N-1 pixels of the N pixels of the pixel group are selected. Sequentially temporarily storing; For each of the calculated N-I pixels, a first reference bit is sequentially added if it is a result from the first or second operation step, and if it is a result from the third operation step, the first reference bit. Sequentially adding a second reference bit in an inverted state of; For each of the first reference bits, a first additional bit is sequentially added if the result is from the first operation step, and if the result is from the second operation step, a second part is inverted of the first additional bit. Adding bits sequentially; And And adding the first identification bit. The method of claim 3, wherein after the combining step, If at least one of the calculated N-1 pixels is not within a predetermined range, a first division step of dividing the N pixels of the pixel group by a predetermined number to obtain N divided pixels; And And storing the result of the first division step as compressed data of the MPEG-reconstructed image data. The method of claim 4, wherein the first division step, If at least one of the calculated N-1 pixels is a result of the fourth operation step, dividing the N pixels of the pixel group by a predetermined number, and temporarily storing the divided results sequentially; And And replacing the least significant bit of the last result among the divided results with a second identifier that is an inverted state of the first identification bit. The method of claim 3, wherein after the combining step, A second division step of dividing the N pixels of the pixel group by a predetermined number to obtain N divided pixels when the total bits combined in the combining step exceed a predetermined bit; And And storing the result from the second division step as compressed data of the MPEG restored image data. The method of claim 6, wherein the second division step, If the total bits combined in the combining step exceeds a predetermined bit, dividing the N pixels of the pixel group by a predetermined number, and temporarily storing the divided results sequentially; And And replacing the least significant bit of the last result among the divided results with a second identifier that is an inverted state of the first identification bit.

Images