Method for producing video coding and programme-product
According to the invention, there are provided sets of contexts specifically adapted to encode special coefficients of a prediction error matrix, on the basis of previously encoded values of level k. Furthermore, the number of values of levels other than 0 is explicitly encoded and numbers of appropriate contexts are selected on the basis of the number of spectral coefficients other than 0.
1. A device for video coding for a current image represented by macroblocks and a data stream, the device comprising a processor and a machine-readable medium, the machine readable medium storing instructions that, when loaded on the processor, configure the processor to:
receive motion vectors for a macroblock of the current image based on one or more previously received reference images stored in an image memory;
calculate a prediction image based on the received motion vectors, the prediction image representing predicted movement of the macroblocks of the reference image based on the received motion vectors;
generate a prediction error matrix, representing a prediction error after transformation by a transformer and quantization by a quantizer, the prediction error representing a subtraction result generated by subtracting the prediction image from the current image by a substractor;
convert the prediction error matrix into a series of symbols using coefficient sampling; and
perform context-adaptive arithmetic encoding of the symbols on the basis of symbol frequencies, a distribution of the symbol frequencies depending at least in part on an already encoded symbol;
wherein the context-adaptive arithmetic encoding of the symbols includes, for a symbol being encoded, selecting from different predetermined distributions of symbol frequencies a particular predetermined distribution of symbol frequencies based on the symbol encoded immediately beforehand; and
the predetermined distribution of symbol frequencies indicates the likelihood of different types of symbols occurring immediately following the type of the symbol encoded immediately beforehand based on known statistical interdependencies between different types of symbols occurring in succession; and
form a bit stream using context-adaptive arithmetic encoding of the symbols.
2. A device in accordance with claim 1 , wherein converting the prediction error matrix includes coefficient sampling into a series of levels and lengths of zero sequences.
3. A device in accordance with claim 2 , wherein levels and lengths are represented as unary values.
4. A device in accordance with claim 2 , further comprising arithmetically encoding the levels context-adaptively in each case, wherein a distribution of underlying level value frequencies depends at least in part on a level value encoded beforehand in each case.
5. A device in accordance with claim 2 , further comprising arithmetically encoding the lengths context-adaptively in each case, wherein a distribution of the underlying length value frequencies depends at least in part on the length value encoded beforehand in each case.
6. A device in accordance with claim 2 , further comprising sorting level values before context-adaptive arithmetic encoding in accordance with level values.
7. A device in accordance with claim 1 , further comprising encoding and transmitting a number of symbols read out for coefficient sampling.
8. A device in accordance with claim 7 , wherein a selection of the distribution of symbol frequencies depends at least in part on the number of symbols read out.