Moving Picture Experts Group MPEG
The MPEG compression standards were developed by the Moving Picture Experts Group (MPEG).
This technology defines compression standards for audio and video information and makes it suitable for broadcast transmission. Let’s describe the types of standards in sequence.
Compression standard MPEG -1
AND
The MPEG -1 standard assumes a transition from spatial to spectral representation of images. This allows the multidimensional structure of the image to be selectively processed to achieve its highest quality characteristics. In the MPEG -1 standard, for this the discrete cosine formation method (DCT) is used, which applies to all the same 8×8 blocks, into which the image is divided. The DCT application gives a matrix of 64 spectral components – coefficients. The sequential sum of coefficients during image reconstruction, starting from zero, determined by the average brightness of the original block, leads to restoration of the block to a satisfactory image. Experience shows that for the average image of images in more than 50% of all blocks, 8x 8 can be restored to a good coefficient state by adding approximately 20 spectral components of 64. This reduces the number of bits needed for encoding. The compression ratio reaches eight.
Searching for higher compression degrees up to hundreds of times when using DKM (8x 8) technology leads to strong distortion of the original image. Using 16×16 drives requires significant computational cost.
In accordance with the MPEG -1 standard, audio and video information (streams) is transmitted at a speed of 150 Kb / s. The flow control selects key video frames and fills only the areas that change between frames.
Simply put, MPEG -1 compression technology includes the following steps.
Before removing the temporal redundancy encoding, as we have described above, it occurs in temporal compression technology in a group of 12-frame images – GOP, where the selected I -frames (intra-frame – frame) which is compressed without modification, the P-frames (pre-directed frame), during encoding what part of the information is erased, and during playback, information from the previous I or P-frames is used. The most “compressed” frame type is B (bidirectional frame) – frames in which, during encoding, information losses are significant and, during playback, information from previous I or P frames is used. If we take the most typical frame chain IBBPBBPBBIBB formed during encoding, then the information compression process is obvious, since in the GOP chain the relationship between frames (I: P: B) is 2: 2: 8, that is prevalence of the most compressed frames by 4 times.
After framing is complete, the I-frames are divided into 8×8-pixel blocks and the P and B-frames are “compressed” using “motion prediction” technology. The motion prediction algorithm obtains a block from the current frame (8×8 pixels) and similar blocks from previous I or P frames as processed information.
After going through processing, eventually at this “compression” stage there is the following information:
the motion vector of the current block relative to the previous ones,
the difference between the present and past blocks, subject to additional coding.
The encoding process goes through three stages
DCT (discrete cosine preobazovanie – DCT (discrete cosine transformation);
quantization (quantization): the transfer of information from an analog to a discrete form;
conversion of the resulting blocks from matrix to linear form.
The MPEG -1 standard in the color television transmission system turned out to be of little use, since in the YUY format (see expressions (4) and (5) in Part I of this article) it is optimized for use with the parameters:
240 lines per frame (line per frame – lpf);
352 points on a line (point per line – ppl);
30 minute staff scan (frames per second – fps);
therefore, it provides a video quality lower than that transmitted according to the color television standard.
However, with the development of the MPEG -1 format, audio coding techniques have been developed. The result was presented as a series of Layer I, Layer II (Musicam), Layer III (MP 3) audio encoders.
