DE10145374C1 - Video coding method identifies monotonously decreasing level sequence for using different entropy coding to provide monotomy information and entropy coding value - Google Patents

Abstract

The video coding method estimates actual image information from transmitted image regions, with provision of a prediction error matrix (PFM) linearised via coefficient sampling (SCAN) and converted into a series of levels and zero sequence lengths. When the levels exhibits a monotonously decreasing sequence a different entropy coding is used for delivery of a monotony information and an entropy coding value at the output (OUT). An Independent claim for a video coding device is also included.

Description

The invention relates to a method and a device, in the or from the already transferred image areas current image information is estimated and only one Prediction error is transmitted. This is usually done by dividing an input image into blocks and a corresponding block in for each of these blocks previous images is searched, which is an error or Distance dimension minimized. The prediction error is caused by DCT or integer transformation from the spatial to the frequency domain transformed, quantized and the resulting Coefficients are compressed by means of entropy coding and sent to the recipient.

From international patent application WO 00/27128 is a Method and an apparatus known in which the quantized data in the frequency domain with quantized data Data in the frequency range of a second block compared and a prediction error matrix is formed.

In these methods known per se, the property exploited the DCT or integer transformation that the Energy is focused on coefficients, the low one Represent frequencies. After quantization lies in As a rule, a sparse matrix with only a few Coefficients near the top left corner are non-zero. In the next step, this matrix is replaced by a Linearized coefficient scan that measures the coefficients in the Order of their probability of being non-zero, scans; a ZigZag scan is often used for this. This List is usually viewed as a sequence of run / level Tuples are shown, where the level is the value of the specifies coefficients other than zero and the run the Number of zeros between this and the  previous coefficients are not equal to zero. This Sequence is the last step of entropy coding fed. Common methods for entropy coding are e.g. B. 3D Huffman codes with a representation as a 3-tuple Level / Run / Last, which is used in the ITU-T standard H.263 finds 2-D UVLCs with a representation as a tuple level / run and an indication of the end of the block by coding a Levels 0, that of the proposed ITU-T standard H.26L is provided, or an independent coding of the levels and runs through an arithmetic encoder.

The object underlying the invention now exists therein the efficiency of the entropy coding of the To enable prediction errors and thereby the transmission of video signals with better quality at the same Data rate or with the same quality but lower To achieve data rate.

The invention is in terms of the method by Features of claim 1 and in terms of arrangement solved the features of claim 5.

The subclaims relate to advantageous refinements of the inventive method.

The invention essentially consists in the fact that at a such video coding often fulfills the property of monotonically falling amounts of the level for a more efficient coding is used, the value range the values to be coded and thus the number of code words can be restricted and, for example, only one block Information about the existence of such a monotony must also be transferred.

An embodiment of the invention is described below the drawing explained in more detail.  

There is only one part relevant to the invention a device for video coding by means of a frequency range based motion estimation shown only based on a prediction error matrix PFM. This prediction error matrix is represented by a Coefficient scanning SCAN linearized as well as in a sequence of levels l and lengths of zero sequences (runs) converted. These sequences of level 1 are checked by a test unit MC checks whether the amounts | l | the level in a monotonous descending order. By a monotonic signal m is effected with the help of a selection unit SEL that in the case a monotonically descending sequence, an entropy coding EC2 and otherwise entropy coding EC1 takes place. The Monotonic information m and the entropy coding The values formed are then sent to an output bit stream OUT added.

If the amounts of the levels or levels in monotonous descending order, so with this Know the levels of coding much more efficiently, since with Coding a level with Betag | l | = k the range of values for the following levels to the interval [-k; k] restricted and thus the number of possible code words is reduced. In addition, this approach allows a better statistical Modeling the occurrence probabilities of the individual Symbols and thus a further increase in coding efficiency. Another advantage of this procedure is that it basically regardless of the type of subsequent Entropy coding is - one use is with both VLC codes as well as possible with an arithmetic encoder.

The decoder must be communicated in some way whether a given block meets this requirement. This can be due to the monotonicity signal m, for example Macroblock or block level, analogous to a Coded / NonCoded Decision made, taking a block in the case of "coded" is coded because it has at least one coefficient unequal  Contains zero). In addition, encoders and decoders in the case VLC entropy coding via appropriately adapted Code tables or in the case of an arithmetic Entropy coding via corresponding Probabilities of occurrence.

An exemplary embodiment of a programmatic implementation of the invention is shown below:
The prerequisite for this is the existence of entropy coding that supports the following functionality:
Entropy_code_sortedflag (s)

The probability p (s = 1) or p (s = 0) = 1 - p (s = 1) is required for this.
Entropy_code_level-sorted (l, r, k)

The conditional probabilities of the occurrence of a level 1 are required provided that the last coded level had the amount k p (l | k) for all l ∈ [-k, k], for all 1 ≦ k ≦ n.
Entropy_code_level (l, r)

The probabilities p (l) are required for all l ∈ [-n, n].

These probabilities can, but do not have to imperative, adaptively adapted to the actual circumstances become. A coding of the lengths of zero sequences (runs) is not taken into account here for the sake of simplicity.

The following flowchart is that of this invention report underlying innovations highlighted in italics.  

Step 1 (checking the criterion)

Test whether the sequence of the coefficients to be encoded is monotonous is falling. If so, then sorted = 1, otherwise sorted = 0

2nd step (coding of the block)

Entropy_code_sortedflag (sorted)

If falling monotonously, ie sorted = 1:
Prev_level = n

For all coefficients <level, run <in their given order:
Entropy_code_level_sorted (level, run, prev_level)
Prev_level = abs (level)
Entropy_code_level_sorted (0, 0, prev_level)

Otherwise:
For all coefficients <level, run <in their given order:
Entropy_code_level (level, run)
Entropy_code_level (0, 0)

3rd step

As long as blocks are still to be coded, continue with step 1.

The invention also includes all combinations of hardware and Software for carrying out the method according to the invention, and computer program products with a corresponding Storage medium as well as the corresponding storage medium itself.

Claims (5)

1. video coding method,
in which current image information is estimated from already transmitted image areas and a prediction error matrix (PFM) is transmitted,
in which the prediction error matrix is linearized by a coefficient scan (SCAN) and converted into a sequence of levels (l) and lengths of zero sequences,
in which the amounts (| l |) of the level are checked as to whether they are in a monotonically descending sequence,
in which, in the case of a monotonically descending sequence, a different entropy coding takes place than otherwise and
in which monotonic information (m) and the values formed by the entropy coding are added to an output bit stream (OUT). 2. The method according to claim 1, in the case of a monotonically descending sequence a Entropy coding takes place in such a way that the code words in Dependence of the respective current level (l) and the respective previous level (k) are formed, wherein the code words the conditional probabilities (p (l | k)) the occurrence of a current level depending on the previous coded levels are assigned. 3. The method according to claim 1 or 2, in which the entropy coding with a variable length code he follows. 4. The method according to any one of claims 1 or 2, in which entropy coding with an arithmetic encoder he follows.   5. Video coding device,
which have means for receiving input video data and for forming a prediction error matrix (PFM),
in which a scanning unit is provided in such a way that the prediction error matrix is linearized by a coefficient scan (SCAN) and converted into a sequence of levels (l) and lengths of zero sequences,
in which a test unit (MC) is provided in such a way that the amounts (| l |) of the level are checked as to whether they are in a monotonically descending sequence,
in which a selection unit (SEL) is present in such a way that in the case of a monotonically descending sequence, a different entropy coding is carried out than otherwise and
in which the output means are present in such a way that monotony information (m) and the values formed by the entropy coding are added to an output bit stream (OUT).