Principle of MPEG
The Principle of MPEG
Following, the aspects of the video coding are outlined in this area. MPEG audio coding is subject of an independent chapter. General aspects and fundamentals are shown here. Further references are attached at suitable places.
Both MPEG 1 and MPEG 2 utilise a number of circumstances which result from the human vision or from the specific characteristics of video data:
Colour sensivity of human vision
Due to the less sensitiveness of the eye for colour perception, the colour representation (chrominance) is reduced by the factor 4 in relation to the brightness representation (luminance). The colour information of four pixels will be combined together to a single value. In total the data volume is halved. A 4:1:1 YCC colour system is applied.
Perception of details
Information must be rejected in every case to achieve the target data rates. It is of importance for the visible quality, that only details are impaired having a subjectively low influence. To achieve this, the original data has to be converted by the discrete cosine transform (DCT), followed by weighting the data (quantization). With the help of these two steps, it can be distinguished approximately between important (relevant) and unimportant (irrelevant) picture information. The procedures are nearly identical with those used for JPEG coding.
Differences between pictures
At many scenes only a part of the picture moves or changes, while the background remains unchanged. The contents of the previous or following picture sections can be addressed, only the differences have to be encoded.
Movement of details.
Modifications from picture to picture frequently occur, if central motives change their position or if the background moves itself. By determining the movement of certain picture elements, a clear saving can be achieved. It is only necessary to indicate the changes in the position by movement vectors.