DE102012212068A1 - Method for generating and transcoding encoded video data stream (VDS) of sequence of images, involves carrying out encoding mode for image frames of VDS, based on comparison result, for encoding image blocks with selected coding mode - Google Patents

Abstract

The method involves comparing transformation coefficients of image blocks (BB1,BB2) of sequence of images (BS) in encoded video data stream (VDS), based on an error metric. The sequence of images in encoded VDS is provided with interdependent images (P11,P12,P21,P22). The images (P11,P12) serve as respective image frames (SP1,SP2). An encoding mode is carried out for image frames, based on comparison result, for encoding image blocks (BB11,BB12) with selected coding mode. An independent claim is included for a device for generating and transcoding encoded video data stream (VDS) of sequence of images.

Description

The invention relates to a method and apparatus for generating and transcoding a coded video data stream.

Conversion of encoded data of a first format into encoded data of a second format is referred to as transcoding. The transcoding is applied, for example, to encoded video data streams encoded in a first format to convert it into a different format encoded video data stream. In a transcoding of video data, which are present in individual images (for example in JPEG format or in H.264 / AVC format with INTRA coding) into a video data stream using prediction in terms of time, usually a decoding of the coded image data and a re-encoding into the desired video format. For example, the image data stream is completely decoded and then re-encoded by an H.264 / AVC encoder. This approach is very complex and time consuming.

The object of the invention is to provide a method and devices in which the transcoding of a coded image data stream into a coded video data stream is enabled faster and in comparison with the above-described procedure with lower processing power.

These objects are achieved by the independent claims. Further developments of the invention are contained in the dependent claims.

The invention proposes a method for generating a coded video data stream from a sequence of first pictures, wherein the first pictures are coded individual pictures (SP1, SP2) which have no (direct) data dependency and are provided as picture data stream. The encoded video data stream comprises an image sequence with interdependent second images (P21, P22). The respective images each have a plurality of image blocks having a plurality of respective transformation coefficients.

In the method, the following steps are performed according to the invention: The transformation coefficients of at least one image block of a first frame serving as a reference image are compared with the transformation coefficients of a picture block of the same location (co-located) of a second frame to be coded based on an error metric. Subsequently, a selection of a coding mode for the second individual image to be encoded is performed in a second image of the video data stream based on the comparison result. Subsequently, the coding of the image block of the second individual image to be coded with the selected coding mode ensues.

In contrast to the prior art, for transcoding an image data stream in a video data stream, it is no longer necessary to first decode and then re-encode the image data stream. Rather, transcoding is done by directly processing the encoded image data. A decision as to how a coding should take place in the video data stream takes place here on the basis of the transform coefficients contained in the coded individual pictures (SP1, SP2). This makes it possible to make the decision as to which coding mode is ideally used faster. As a result, a comparatively lower computing power is needed for the transcoding. In addition, the approach provides for higher image quality due to the direct processing of the transform coefficients of the coded first images. There is no inverse transformation and re-transformation of the coefficients from the pixel domain required. Finally, the proposed approach has a low complexity because the coding mode selection is based on information contained in the transformation domain.

As a first and a second single image, two temporally adjacent first images can be selected for the comparison of the transformation coefficients.

In a development of the method, the coding mode used in dependence on the comparison result is the skip coding mode or the INTER coding mode or the INTRA coding mode. In this case, depending on the result of the comparison, the optimum coding mode for a particular picture block is always selected.

In one development of the invention, the transformation coefficients may also be represented by their level and a quantization step size. If the transformation coefficients are represented by their levels, the computation steps required to carry out the method can be carried out in a particularly simple and efficient manner.

As an error metric, either the mean squared error (MSE) or the sum of squared differences (SSD) is processed. In principle, other error metrics can also be used.

According to a variant, the coding mode Skip is selected if the result of the error metric is smaller than a first threshold value. In an alternative variant, if the result of the error metric is smaller than a first threshold and at the same time the implicitly determined motion vector of the picture block being processed is not equal to zero, then the coding mode INTER is selected, in which case a parameter representing the motion vector is explicitly selected to be zero.

That is, the coefficients of the coded difference image are explicitly zeroed and transmitted to the decoder in the coded video stream.

If the result of the error metric is greater than or equal to the first threshold value, then the coding mode INTER or INTRA is possible. The decision can be made, for example, by the rate distortion optimization method known from the prior art, as described for example in [1].

According to a further embodiment, if the result of the error metric is smaller than a first threshold value and at the same time the implicitly determined motion vector of the image block being processed is not equal to zero, then the coding mode INTER is selected, wherein a parameter representing the motion vector is explicitly selected to be zero. This means that the corresponding parameters of the coded difference image are set to zero. By doing so, the coding mode Skip is selected when the motion vector of the picture block under consideration is zero, whereby the block to be coded can be easily copied. If, in the case in which the skip coding mode is selected on the basis of the result of the error metric, it results that the implicitly determined motion vector is not equal to zero (this can be derived from the neighboring picture blocks of the second frame), then the skip is copied Coding mode the INTER coding mode is selected. In order to be able to mimic the skip coding mode accordingly, the motion vector is explicitly selected to be zero in this case in order to correspond to the condition of the image block of the same location.

In a further embodiment, a second threshold value is predetermined which is greater than the first threshold value. In a variant, a) INTER is chosen as the coding mode if the result of the error metric is greater than the first threshold value and smaller than the second threshold value. In a variant b), the coding mode INTRA or INTER is selected with a motion vector not equal to zero if the result of the error metric is greater than the second threshold value. In the case mentioned in variant a), one or more parameters of the INTER coding mode representing the motion vector are explicitly selected such that the motion vector of the INTER coding is zero.

The decision in case b) between the coding modes INTRA or INTER (where the motion vector is not equal to zero) can be made, for example, based on the already mentioned rate distortion criterion of the algorithm described in [1].

The first and second thresholds will vary depending on the quality of the output data stream, i. of the encoded video data stream. In particular, if the transformation coefficients are represented by their levels, this can be done as a function of the selected quantization step size.

In a further refinement, with the INTER-coding mode selected in the transformation plane, a residual to be coded for the single image to be coded in the video data stream is determined from the difference of the transformation coefficients of the image blocks of the same location of the first and the second individual image. As a result, the transcoding can be realized in a particularly simple manner, since it is possible to work in the transformation domain.

It is also expedient if a flag is set in the encoded video data stream which suppresses filtering of the video data stream image data during decoding. Additional operations, such as deblocking filtering or adaptive loop filtering may degrade image quality. By setting the flag, these filter mechanisms can be deactivated, which keeps the quality of the transcoding high.

In a further expedient refinement, for processing the error metric for selecting the coding mode and / or for calculating the residual, a transformation coefficient which comprises or represents a difference value of a DC value is calculated back to an absolute value of the DC value. The DC value corresponds to a DC component, such as the average brightness, in one of the picture blocks of the first pictures. In part, with INTRA coding applied to JPEG images, DC value coefficients are estimated to reduce the data rate. Typically, a differential pulse code modulation (DPCM) is used for this, ie the DC transformation coefficient of the considered image block is coded as the difference with respect to the corresponding value of the previous block in the same image. This results in an estimation chain which starts from the first image block in the viewed frame. In order to be able to determine the value of the error metric for the coding mode decision and / or to correctly determine the transformation coefficients of the residual, this procedure is reversed according to the invention. This means that the DC values are determined from the difference values of the DC value.

In a further expedient refinement, for the processing of the error metric for the selection of the coding mode and / or for the calculation of the residual, the quantization step size is normalized when the quantization step size varies between the first and the second individual image.

• – die ersten Bilder codierte Einzelbilder (SP1, SP2) sind, die keine Datenabhängigkeit zueinander aufweisen und als Bilddatenstrom bereitgestellt sind;
• – der codierte Videodatenstrom eine Bildsequenz mit voneinander abhängigen zweiten Bildern (P21, P22) umfasst;
• – die jeweiligen Bilder jeweils eine Mehrzahl an Bildblöcken mit einer Mehrzahl an jeweiligen Transformationskoeffizienten aufweisen;

umfassend die folgenden Einheiten:

• – ein erste Einheit zum Vergleich der Transformationskoeffizienten zumindest eines Bildblocks eines als Referenz dienenden ersten Einzelbilds mit den Transformationskoeffizienten eines Bildblocks gleichen Orts eines zweiten, zu codierenden Einzelbilds basierend auf einer Fehlermetrik;
• – eine zweite Einheit zur Auswahl eines Codiermodus für das zu codierende zweite Einzelbild in ein zweites Bild des Videodatenstroms basierend auf dem Vergleichsergebnis;
• – eine dritte Einheit Codieren des Bildblocks des zu codierenden zweiten Einzelbilds mit dem ausgewählten Codiermodus.

The invention further provides an apparatus for generating a coded video data stream from a sequence of first images, wherein

• The first images are coded individual images (SP1, SP2) which have no data dependency on one another and are provided as an image data stream;
• - the coded video data stream comprises an image sequence with interdependent second images (P21, P22);
• The respective images each have a plurality of image blocks having a plurality of respective transformation coefficients;

comprising the following units:

• A first unit for comparing the transformation coefficients of at least one image block of a reference frame of the first frame with the transformation coefficients of an image block same location of a second frame to be encoded based on an error metric;
• A second unit for selecting a coding mode for the second frame to be encoded into a second frame of the video stream based on the comparison result;
• A third unit coding the picture block of the second frame to be coded with the selected coding mode.

• – ein erste Einheit zum Vergleich der Transformationskoeffizienten zumindest eines Bildblocks eines als Referenz dienenden ersten Einzelbilds mit den Transformationskoeffizienten eines Bildblocks gleichen Orts eines zweiten, zu codierenden Einzelbilds basierend auf einer Fehlermetrik;
• – eine zweite Einheit zur Auswahl eines Codiermodus für das zu codierende zweite Einzelbild in ein zweites Bild des Videodatenstroms basierend auf dem Vergleichsergebnis;
• – eine vierte Einheit zum Erzeugen des transcodierten Videodatenstroms durch Einfügen des mit dem ausgewählten Codiermodus codierten Bildblocks des zu codierenden zweiten Einzelbilds in den transcodierten Videodatenstrom.

Finally, the invention comprises a transcoding device for generating a transcoded video data stream from a coded image data stream which can be generated according to the method described above, comprising the following units:

• A first unit for comparing the transformation coefficients of at least one image block of a reference frame of the first frame with the transformation coefficients of an image block same location of a second frame to be encoded based on an error metric;
• A second unit for selecting a coding mode for the second frame to be encoded into a second frame of the video stream based on the comparison result;
• - A fourth unit for generating the transcoded video data stream by inserting the encoded with the selected encoding mode image block of the second frame to be encoded in the transcoded video data stream.

The transcoding device enables an implementation of the transcoding method, wherein the individual method steps can be carried out by means of the units mentioned. The advantages are analogous to the transcoding method.

The invention and its developments are explained in more detail below with reference to an embodiment in the drawing. Show it:

1 a sequence of coded frames (SP1, SP2) which have no data dependence on each other and from which a coded video data stream can be generated, and

2 a flow chart and an apparatus for generating a coded video data stream.

Elements with the same function and mode of operation are provided with the same reference numerals in the figures.

In a video coding method such as ITU H.264 / AVC (Advanced Telecommunication Union AVC), a picture sequence BS having plural pictures P21, P22 is encoded into a video data stream VDS. In the encoded video data stream VDS this image sequence BS is encoded.

The images P21, P22 are generated by transcoding from images P11, P12. The images P11, P12 are coded frames (SP1, SP2) SP1, SP2, which have no data dependency on each other and are provided as image data stream BDS.

The images P11, P12, P21, P22 are divided into image blocks BB having a size of 4 × 4 or 8 × 8 pixels. In general, the image blocks may take on arbitrary forms, the sizes mentioned in the US Pat Standard H.264 be used. The coding of the images is block by block, whereby a reduction of the amount of data is achieved by the coding.

When encoding image blocks, the following coding modes are generally known:
INTRA: An image block is coded without reference to at least one other image block.

INTER Prediction: The coding of an image block of an image takes place by prediction on an image area, wherein the image area lies in an image past or behind the image. This image area is called a reference image area or reference.

INTRA Prediction: The coding of an image block of an image is done by prediction on an image area, the image area, i. the reference is in the same picture as the picture block.

The transcoding of the individual images (SP1, SP2) SP1, SP2 proposed in accordance with the invention into an image sequence BS of the coded video data stream VDS with the images P21, P22 is effected by a comparison of transformation coefficients of an image block BB11 of the reference frame SP1 with the transformation coefficients of an image block BB2 same location of the second, to be coded frame SP2. The comparison is based on an error metric. This first step S1 is performed by a first unit E1, cf. 2 , Based on the comparison result, in step S2, a second unit E2 selects a coding mode for the image block BB2 to be coded into a corresponding image block B21 of the image P22 of the video data stream VDS. In a step S3, a unit E3 codes the image block of the individual image SP2 to be coded with the selected coding mode. The coded image block is then inserted into the video data stream VDS by a unit E4 in a step S4.

Depending on the result of the comparison, optionally the skip coding mode, the INTER coding mode or the INTRA coding mode are used as coding modes.

berechnet. Darin repräsentieren c_i und c_i,Ref die Koeffizienten i eines Blocks im gegenwärtigen Bild und in dem Referenzbild. N repräsentiert die Anzahl der Koeffizienten in dem Bildblock BB. The individual images (SP1, SP2) SP1, SP2 are preferably two temporally adjacent first images. In principle, however, temporally adjacent images can not be selected for the comparison of the transformation coefficients. The comparison of the transformation coefficients of the image blocks of the same location (so-called co-located image blocks) takes place by means of an error metric. For example, the mean squared error (MSE) can be used for this purpose. The MSE between image blocks of the same place in two images is according to

calculated. Therein, c _i and c _{i, Ref represent} the coefficients i of a block in the current picture and in the reference picture. N represents the number of coefficients in the image block BB.

The coding mode for the considered image block, which also output data (image data stream BDS) is selected according to the value of MSE. In the case where MSE is below a first threshold th _{1, MSE} , the skip coding mode is selected without offset. The term "without offset" means that the motion vectors of the image blocks are implicitly determined to be zero. This means if the output data stream, that is, the encoded video data stream is decoded, the relevant image block of the reference image is encoded in the image block of the same location of the current image, without further refinement of the pixel values.

In the other case, ie when MSE is greater than or equal to the first threshold th _{1, MSE} , the INTER or INTRA encoding mode is selected:

The decision as to whether INTER or INTRA is chosen as the coding mode is made according to the known methods of rate distortion optimization as described in [1].

In the event that it is not possible to select the skip coding mode with implicitly determined motion vectors (MV) = 0, this is mimicked by an INTER coding mode in which the parameters specifying the motion vector are explicitly set so that the motion vector is zero, and without encoding transform coefficients. This case can occur because nonzero motion vectors are implicitly derived from adjacent image blocks or due to other video format restrictions.

Instead of using the MSE, the sum of squared differences (SSD) can also be used as the error metric. The decision is then made according to

wobei th_1,SSD = N·th_1,MSE gilt. Dies erleichtert die Berechnung der Entscheidungsparameter, da eine Divisions-Berechnung vermieden werden kann. SSD is calculated according to

where th _{1, SSD} = N · th _{1, MSE} holds. This facilitates the calculation of the decision parameters, since a division calculation can be avoided.

wobei l_i und l_i,Ref die Level der Transformationskoeffizienten und k ein konstanter Multiplikator sind. Für eine gleichförmige Quantisierung mit dem Vielfachen einer Quantisierungsschrittgröße sind SSD und SSD_level identisch. In diesem Fall ist der Faktor k die quadratische Quantisierungsschrittgröße q, welche abhängig ist von einem Quantisierungsparameter QP, d.h. k = q² = [q(QP)]² =: k(QP). In an encoded video data stream, the transform coefficients are represented by the so-called level l _i . To approximate SSD, the following formula can be used:

where l _i and l _{i, Ref are} the levels of the transform coefficients and k is a constant multiplier. For uniform quantization with the multiple of a quantization step size, SSD and SSD _{levels are} identical. In this case, the factor k is the quadratic quantization step size q, which depends on a quantization parameter QP, ie k = q ² = [q (QP)] ² =: k (QP).

Here, q is the quantization step size in the image block of the image P11, P12 of the image data stream, and QP is the corresponding quantization parameter. To achieve a low implementation complexity, k (QP) can be stored in a table for different QP values for fast calculation of SSD _level .

wobei

die Summe der quadrierten Differenzen der Level (SSLD) und

ist. The decision process can be further simplified by

in which

the sum of the squared differences of the levels (SSLD) and

is.

Likewise, the threshold th _{1, SSLD} (QP) may be stored in a table for less complexity.

The thresholds th _{1, MSE} , th _{1, SSD} or th _{1, SSLD} may depend on the quality of the output data _stream , ie the encoded video data _stream . In particular, there is a dependence on the quantization parameter in the image block to be coded. The larger the quantization parameter, ie the larger the quantization step size, the greater the threshold value is selected. For a lower quality of the output data stream, therefore, the skip coding mode can be selected more frequently.

The choice between the INTER and the INTRA coding mode also takes place depending on the result of the error metric used, for example MSE, SSD or SSLD. If the result of the error metric (eg MSE) is below a second threshold th _{2, MSE} , where the second threshold is th _{2, MSE} greater than the first threshold th _{1, MSE} , then the INTER coding mode is selected without displacement , This means that the parameters representing the motion vector are implicitly or explicitly set to zero. In this case, further refinement of the transform coefficients can be performed. This is not possible if the skip coding mode has been selected as a function of the comparison result.

If the result of the error metric is greater than or equal to the second threshold value th _{2, MSE} , the INTRA coding mode is selected. Then, no prediction is made of an image block of the same location of the image.

The decision process can be expressed by the following formulas for MSE, SSD and SSLD:

The decision between the coding modes INTRA or INTER with a motion vector not equal to zero (MV ≠ 0) can for example be based on the rate distortion criterion with which the various coding options can be analyzed.

Similar to determining the absolute value of the first threshold value, the second threshold value (th _{2, MSE} , th _{2, SSD} or th _{2, SSLD} ) can be selected depending on the quality of the output data _stream , ie the coded video data _stream .

In the case where the INTER coding mode is selected, the pixel values of an image block can be refined by the coding of transform coefficients. The computation of this residual information can optionally take place in the pixel or the transformation domain. Since the computation in the transformation domain is less expensive, this is preferred.

The commonly used discrete cosine transform (DCT) in the encoding of images and video is a linear operation. Basically, it is therefore possible to calculate the transformation coefficients of the residual (R ~) in the transformation domain. The following signals are present in the input image data stream BDS: I ~ = T (I) and I ~ _Ref = T (I _Ref ), which are each INTRA coded. In this notation, I represents the image to be transcoded in the current step, while I _Ref serves as the reference image transcoded in a previous step.

In the output video data stream VDS, the following signals must be encoded:
I ~ = T (I) for INTRA coded picture blocks as well as the transformed residual R ~ = T (R) for INTER coded picture blocks. The temporary prediction process on a video decoder is performed in the pixel domain. After the residue R is encoded in the transform domain (and thus in the form of the signal R ~), an inverse transformation must be performed. Therefore, the prediction process becomes too I = P + T ^-1 (R ~).

By applying a transformation results: T (I) = T (P + T ^-1 (R -)) = T (P) + T (T ^-1 (R -)) = T (P) + R -.

It should be noted that for linear operations f, the correspondence f (a * A + b * B) = a * f (A) + b * f (B) holds, where in the derivative f ≡ T, a = 1 and b = 1 applies.

The above equation can be reworded to: R ~ = T (I) - T (P) = I ~ - P ~.

Since the predictor of the current image is equivalent to the same block image block in the reference image (P = I _Ref ), this can be rewritten as: R ~ = I ~ - I ~ _Ref .

Therefore, the residual to be coded in the output video data stream VDS can be calculated from the coded information of the input data stream, as the above derivation shows.

I:

Pixelwerte des Bildblocks des im gegenwärtigen Schritt zu codierenden Bildes,

I_Ref:

Pixelwerte des Bildblocks gleichen Orts im Referenzbild,

P:

Pixelwerte des Prädiktors für das gegenwärtige Bild, welche von dem Referenzbild errechnet sind,

R:

Pixelwerte des Residuum-Blocks im gegenwärtigen Bild,

I ~,I ~_Ref, P ~, R ~:

Transformationskoeffizienten der entsprechenden Signale I, I_Ref, P, R,

T(·):

Transformationsoperation, beispielsweise diskrete Kosinus-Transformation (DCT).

The symbols used here are as follows:

I:

Pixel values of the image block of the image to be encoded in the current step,

I _Ref :

Pixel values of the image block same location in the reference image,

P:

Pixel values of the predictor for the current image calculated from the reference image,

R:

Pixel values of the residual block in the current image,

I ~, I ~ _Ref , P ~, R ~:

Transformation coefficients of the corresponding signals I, I _Ref , P, R,

T (·):

Transformation operation, for example discrete cosine transformation (DCT).

In the event that the INTRA encoding mode for the frame of interest is selected in the output video stream, the coefficients of the input image stream can be used to compute the coefficients of the video stream without decoding the pixel values for the frame of interest considered. In the specific case that the output coding format is a superset of the input coding format, the same coding parameters can be chosen in the output video data stream. This is the case, for example, when transcoding from an H.264 / AVC INTRA profile to an H.264 / AVC INTER profile.

If in the coding loop additional operations, e.g. deblocking filtering or adaptive loop filtering, determining the residual in the compressed domain may result in image quality degradation, particularly if these operations are not performed in decoding the frame stream. Likewise, decoded pixel values of the output video data stream could be altered by these operations. Therefore, it is expedient to deactivate the execution of filter operations, which is possible, for example, by the corresponding coding of a corresponding flag which indicates the handling of the filtering for the decoder.

Some INTRA coding methods, such as e.g. JPEG, use a prediction of a DC coefficient to reduce the data rate. Typically, a simple differential pulse-coded modulation (DPCM) is used for this purpose. Here, the DC transformation coefficient of the current image block is coded as a difference to a corresponding value of the previous image block in the same image. This results in a prediction chain that starts in the first block in the image. The DC coefficient includes, for example, a DC component of the average brightness of the image.

In order to be able to correctly calculate the result of the error metric for the selection of the coding mode and the transformation coefficients of the residual (R ~), this operation must be reversed. This means that the DC transformation coefficients must be calculated from the difference values of the DC values. This can be done step by step according to the following formulas: DC _Blk2 = DC _Blk1 + ΔDC _Blk2 , DC _Blk3 = DC _Blk2 + ΔDC _Blk3 ....

multipliziert, bevor der SSLD und unter Umständen die Transformationskoeffizienten-Level des Residuums ermittelt werden. stepsize_Ref und stepsize_Cur sind die Quantisierungsschrittgrößen, die in dem Referenz- und dem gegenwärtigen Bildblock verwendet werden. Der SSLD kann dann nach folgender Formel ermittelt werden:

When computing the SSLD in the error metric, it is convenient to apply a scaling (normalization) to the transform coefficient levels to compute correct values of the SSLD for the encoding mode selection and the transformation coefficients of the residual. This is important if the quantization parameters QP and therefore the quantization step size vary from image to image or even from the image block of the reference image to the image block of the viewed image. Therefore, the transform coefficient levels of the reference picture block become s,

multiplied before the SSLD and possibly the transformation coefficient levels of the residuum are determined. stepsize _Ref and stepsize _Cur are the quantization step sizes used in the reference and current image blocks. The SSLD can then be determined using the following formula:

Moreover, it is possible to perform an aggregation, for example a simple averaging, of MSE or SSD or SSLD for a plurality of image blocks before the basic decision as to which coding mode is used. The image block to be encoded in the output video data stream could comprise a plurality of image blocks and / or sub-blocks in the same location of the input bitstream. This could for example be the case when using different image block sizes in the input and output data stream. In this case, the encoding mode selection can be done by aggregating MSE or SSD or SSLD from multiple image blocks. In the case of an image block in the input image data stream only partially covers the image block of the output video data stream, its weight may be set proportional to the range, ie the number of pixels.

bibliography

• [1] Thomas Wiegand et al., "Lagrange Multiplier Selection in Hybrid Video Coder Control", ICIP 2001, Thessaloniki, Greece, p. 1-4.

QUOTES INCLUDE IN THE DESCRIPTION

This list of the documents listed by the applicant has been generated automatically and is included solely for the better information of the reader. The list is not part of the German patent or utility model application. The DPMA assumes no liability for any errors or omissions.

Cited non-patent literature

• Standard H.264 [0032]

Claims (15)

Method for generating a coded video data stream (VDS) from a sequence of first pictures (P11, P12), wherein - the first pictures (P11, P12) are coded individual pictures (SP1, SP2) which have no data dependency on one another and are used as picture data stream (BDS ) are provided; - the coded video data stream (VDS) comprises an image sequence (BS) with interdependent second images (P21, P22); - The respective images each have a plurality of image blocks (BB) having a plurality of respective transformation coefficients (c _i , l _i ); in which the following steps are performed: - comparison of the transformation coefficients (c _i , l _i ) of at least one image block (BB11) of a reference frame (SP1) with the transformation coefficients (c _i , l _i ) of an image block (BB 12) of the same location a second frame to be encoded (SP2) based on an error metric; Selecting a coding mode for the second frame (SP2) to be encoded into a second frame (P22) of the video data stream (VDS) based on the comparison result; - Coding the image block (BB12) of the second frame to be coded (SP2) with the selected coding mode. Method according to Claim 1, in which two temporally adjacent first images (P11, P12) are selected as the first and second individual image (SP1, SP2) for the comparison of the transformation coefficients (c _i , l _i ). Method according to Claim 1 or 2, in which the skip coding mode or the INTER coding mode or the INTRA coding mode is used as the coding modes as a function of the comparison result. Method according to one of the preceding claims, in which the transformation coefficients (c _i , l _i ) are represented by their level (l) and a quantization step size (q). Method according to one of the preceding claims, in which the mean square error (MSE) or the sum of the squared differences (SSD) of the transformation coefficients (c _i , l _i ) is processed as an error metric. Method according to one of Claims 1 to 5, in which Skip is selected as the coding mode if the result of the error metric is smaller than a first threshold value. Method according to one of claims 1 to 5, _wherein when the result of the error metric is smaller than a first threshold (th _{1, MSE} , th _{1, SSD} , th _{1, SSLD} ) and the implicitly determined motion _vector (MV) of the image block being processed is not equal to zero, is selected as the encoding mode INTER, wherein a parameter representing the motion vector (MV) is explicitly set to zero. Method according to one of the preceding claims, in which a second threshold value (th _{2, MSE} , th _{2, SSD} , th _{2, SSLD} ) which is greater than the first threshold value (th _{1, MSE} , th _{1, SSD} , th _{1 , SSLD} ), where a) is selected as the encoding mode INTER, if the result of the error metric is greater than the first threshold value (th _{1, MSE} , th _{1, SSD} , th _{1, SSLD} ) and smaller than the second threshold value (th _{2, MSE} , th _{2, SSD} , th _{2, SSLD} ), b) is selected as the encoding mode INTRA or INTER with a motion vector not equal to zero, if the result of the error metric is greater than the second threshold (th _{2, MSE} , th _{2, SSD} , th _{2, SSLD} ).  Method according to Claim 8, in which, in variant a), one or more parameters of the INTER coding mode which represent the motion vector (MV) are explicitly selected to be zero. Method according to Claim 8 or 9, in which, with the INTER-coding mode selected in the transformation plane, a residual to be coded for the single picture to be coded in the video data stream is calculated from the difference of the transformation coefficients (c _i , l _i ) of the picture blocks (BB 11, BB12) same location of the first and the second frame (SP1, SP2) is determined. Method according to one of the preceding claims, in which a flag is set in the coded video data stream (VDS) which suppresses filtering of the video data stream (VDS) image data during decoding. Method according to one of the preceding claims, wherein for the processing of the error metric for the selection of the coding mode and / or for the calculation of the residual a transformation coefficient, the includes or represents a difference value of a DC value, is calculated back to an absolute value of the DC value. Method according to one of the preceding claims, in which the quantization step size (q) between the first and the second individual image (SP1, SP2) varies. Device for generating a coded video data stream (VDS) from a sequence of first images (P11, P12), wherein - the first images (P11, P12) are coded individual images (SP1, SP2) which have no data dependency on one another and are used as image data stream (BDS ) are provided; - the coded video data stream (VDS) comprises an image sequence (BS) with interdependent second images (P21, P22); - The respective images each have a plurality of image blocks (BB) having a plurality of respective transformation coefficients (c _i , l _i ); comprising the following units: a first unit (E1) for comparing the transformation coefficients (c _i , l _i ) of at least one image block (BB11) of a reference frame (SP1) with the transformation coefficients (c _i , l _i ) of an image block (BB12) same location of a second frame to be encoded (SP2) based on an error metric; A second unit (E2) for selecting a coding mode for the second frame (Sp2) to be encoded into a second frame (P22) of the video stream based on the comparison result; A third unit (E3) for coding the image block of the second frame (SP2) to be coded with the selected coding mode. Transcoding device (TVOR) for generating a transcoded video data stream (TVDS) from a coded video data stream (VDS), which can be generated according to one of claims 1 to 14, comprising the following units: - a first unit (E1) for comparing the transformation coefficients (c _i , l _i ) at least one image block of a first single frame serving as reference with the transformation coefficients (c _i , l _i ) of an image block same location of a second frame to be coded based on an error metric; A second unit (E2) for selecting an encoding mode for the second frame to be encoded into a second frame of the video stream based on the comparison result; - A fourth unit (E4) for generating the transcoded video data stream (VDS) by inserting the encoded with the selected encoding mode image block of the second frame to be encoded (SP2) in the transcoded video data stream (VDS).

Images