DE102006032021A1 - A method and encoding device for encoding an image area of an image of an image sequence in at least two quality levels, and a method and decoding device for decoding a first encoded data stream and a second encoded data stream - Google Patents

Abstract

The invention relates to a method and an encoding device for coding an image area of an image, wherein for the image area in a first quality level a first coded data stream with a basic quality and in a second quality level, a second coded data stream with a coded together with the first data stream compared to the base quality improved quality, in which a first intermediate information representing the image area in the first quality level and a second intermediate information representing an estimated image area of the image area in at least one temporally preceding or subsequent image or already reconstructed image of the second quality level, providing weighted first and second intermediate information by respectively weighting the first and second intermediate information with at least one respective respective first and second weighting factor, wherein the at least one first weighting factor is formed by a predeterminable weighting function as a function of at least one of the second weighting factors, creates a third intermediate information by subtracting the weighted intermediate information from an image information representing the image area to be encoded of the second quality level, and generating the second coded data stream based on the third intermediate information are formed. Furthermore, the invention comprises ...

Description

The The invention relates to a method and an encoding device for Encoding an image area of an image of an image sequence in at least two quality levels and a method and a decoding device for decoding a first and a second coded data stream.

Currently New video coding and related standards are being developed with which a coding of images of a picture sequence in at least two quality levels, a basic quality and at least one extension quality, can be encoded. For this counts, for example a solution in the current Joint Scalable Video Module (JSVM-Joint Scalable Video Model) [1] for a future one Standard SVC (SVC Scalable Video Coding) for H.264 or MPEG4, Part 10. Here are temporal and local predictions treated separately. First is a time decorrelation, i. a temporal prediction from an already encoded image. Then it will be a local one Decorrelation applied, i. a local prediction of a next lower local Layer.

at the use of two or more quality levels, such as in the JSVM, a coding of the image in a basic quality of a first quality level and in a too basic quality improved quality a second quality level instead of. It represents a first coded data stream the basic quality of the image and the second coded data stream improved quality, the second coded Data stream together with the first coded data stream the improved quality reproduces. The different quality levels represent the image, for example, in different spatial and / or quality resolutions (SNR scalability, SNR signal to noise ratio).

Of the Invention is based on the object, a method and an encoding device for coding and a method and a decoding device for Decoding indicating an improvement in coding efficiency achieve.

other Further developments of the invention are given in the dependent claims.

• a) Bereitstellen einer ersten Zwischeninformation, die den Bildbereich in der ersten Qualitätsstufe repräsentiert;
• b) Bereitstellen einer zweiten Zwischeninformation, die einen geschätzten Bildbereich des Bildbereichs in zumindest einem zum Bild zeitlich vorangegangenen oder zeitlich nachfolgenden Bild oder bereits rekonstruierten Bild der zweiten Qualitätsstufe repräsentiert;
• c) Erzeugen einer gewichteten ersten Zwischeninformation durch Gewichten der ersten Zwischeninformation mit zumindest einem ersten Gewichtungsfaktor und einer gewichteten zweiten Zwischeninformation durch Gewichten der zweiten Zwischeninformation mit zumindest einem zweiten Gewichtungsfaktor, wobei der zumindest eine erste Gewichtungsfaktor durch eine vorgebbare Gewichtungsfunktion in Abhängigkeit von zumindest einem der zweiten Gewichtungsfaktoren gebildet wird;
• d) Erzeugen einer dritten Zwischeninformation durch Subtraktion der gewichteten Zwischeninformationen von einer den zu kodierenden Bildbereich der zweiten Qualitätsstufe repräsentierenden Bildinformation;
• e) Erzeugen des zweiten kodierten Datenstroms auf Basis der dritten Zwischeninformation.

The invention relates to a method for encoding an image area of an image of an image sequence in at least two quality levels, wherein for the image area in a first quality level, a first coded data stream with a basic quality and in a second quality level, a second coded data stream with a coded together with the first data stream improved quality compared to the basic quality, in which the following steps are carried out:

• a) providing a first intermediate information representing the image area in the first quality level;
• b) providing a second intermediate information representing an estimated image area of the image area in at least one temporally preceding or temporally subsequent image or already reconstructed image of the second quality level;
• c) generating a weighted first intermediate information by weighting the first intermediate information with at least a first weighting factor and a weighted second intermediate information by weighting the second intermediate information with at least one second weighting factor, wherein the at least one first weighting factor by a predeterminable weighting function in dependence on at least one of the second Weighting factors is formed;
• d) generating a third intermediate information by subtracting the weighted intermediate information from an image information representing the image area of the second quality level to be coded;
• e) generating the second coded data stream based on the third intermediate information.

The Method achieves an improvement in the compression efficiency consistent image quality of first and second quality level as conventional Method. Using weighting improves one Gesamtprädiktion, since those in each higher quality level conducted Single prediction with Help the weighting factors with a representative of the encoded image content the first quality level connected become. In doing so, the temporal prediction and the prediction become from a lower, i. first quality level, to a next higher quality level, i.e. second quality level, improved. In addition, an increase in compression efficiency by change the weighting factors of the image area to be encoded to the image area be achieved. This change In particular, it is carried out adaptively. By using the weighting function will be a reduction of to be transferred Reduces information in the first and / or second coded data stream, because the first weighting factor using the weighting function on Basis of the second weighting factor, also at a receiver, reconstructed is.

• a) Dekodieren des ersten kodierten Datenstroms zum Rekonstruieren einer ersten Zwischeninformation, wobei die erste Zwischeninformation einen rekonstruierten ersten Bildbereich der ersten Qualitätsstufe für den Bildbereich repräsentiert;
• b) Bereitstellen einer zweiten Zwischeninformation, die einen geschätzten Bildbereich des Bildbereichs in zumindest einem zum Bild zeitlich vorangegangenen oder zeitlich nachfolgenden Bild oder bereits rekonstruierten Bild der zweiten Qualitätsstufe repräsentiert;
• c) Dekodieren des zweiten kodierten Datenstroms zum Rekonstruieren eines rekonstruierten Restfehlers der zweiten Qualitätsstufe;
• d) Erzeugen einer gewichteten ersten Zwischeninformation durch Gewichten der ersten Zwischeninformation mit zumindest einem ersten Gewichtungsfaktor und einer gewichteten zweiten Zwischeninformation durch Gewichten der zweiten Zwischeninformation mit zumindest einem zweiten Gewichtungsfaktor, wobei der zumindest eine erste Gewichtungsfaktor durch eine vorgebbare Gewichtungsfunktion in Abhängigkeit von zumindest einem der zweiten Gewichtungsfaktoren gebildet wird;
• e) Bereitstellen eines rekonstruierten zweiten Bildbereichs der zweiten Qualitätsstufe für den Bildbereich durch Summation der gewichteten ersten und zweiten Zwischeninformationen und des rekonstruierten Restfehlers der zweiten Qualitätsstufe.

The invention further relates to a method for decoding a first encoded data stream and a second encoded data stream, wherein the first encoded data stream comprises an image area of an image of an image sequence in a first quality level with a basic quality and the second encoded data stream combines the image area in a second quality level with one represents quality improved with the first coded data stream compared to the base quality, wherein the coded data streams were generated in particular according to the coding method in which the following steps are performed:

• a) decoding the first encoded data stream to reconstruct a first intermediate information, the first intermediate information representing a reconstructed first image area of the first quality level for the image area;
• b) providing a second intermediate information representing an estimated image area of the image area in at least one temporally preceding or temporally subsequent image or already reconstructed image of the second quality level;
• c) decoding the second encoded data stream to reconstruct a reconstructed residual error of the second quality level;
• d) generating a weighted first intermediate information by weighting the first intermediate information with at least one first weighting factor and weighted second intermediate information by weighting the second intermediate information with at least one second weighting factor, wherein the at least one first weighting factor is determined by a predeterminable weighting function depending on at least one of the second Weighting factors is formed;
• e) providing a reconstructed second image area of the second quality level for the image area by summing the weighted first and second intermediate information and the reconstructed residual error of the second quality level.

This Method of decoding is capable of using the method for Encoding generated data streams to decode and play, for example, on a monitor. A mission the coding and decoding method is not specific to one Coding method limited, such as. H.264. Rather, it is with block-based and / or transformation-based Video coding method, e.g. MPEG-2, or MPEG-4 SVC (MPEG Motion Picture Expert Group, SVC Scalable Video Coding), or wavelet-based compression techniques used.

Becomes generates the weighting function such that the first weighting factor a coefficient of the first intermediate information at a first Position and the second weighting factor of a coefficient of second intermediate information at the first position in total one given value, in particular to one, with the respective Weighting factors especially in the range between zero and one [0; 1] can be a low complexity implementation of the Procedure can be achieved. By normalizing the sum to one will achieve that overvaluation intermediate information is avoided.

Preferably be at least two coefficients of the first and / or second Intermediate information with different weighting factors be weighted, thereby increasing the compression efficiency is reachable.

wobei a_{2_min} einem minimalen Wert von a2, a_{2_max} einem maximalen Wert von a2, |MV2| einem Betrag einer Länge des zweiten Bewegungsvektors MV2 und |MV2max| einem maximalen Betrag einer Länge des zweiten Bewegungsvektors MV2 entspricht.In a preferred embodiment, the second weighting factor of the second quality level is formed by dividing a length of a motion vector of the second quality level belonging to the image area to be coded by a maximum length of the motion vector, in particular by the following formula:

where a _{2_min is} a minimum value of a2, a _{2_max is} a maximum value of a2, | MV2 | an amount of a length of the second motion vector MV2 and | MV2max | corresponds to a maximum amount of a length of the second motion vector MV2.

Herewith a low complex calculation formula can be realized. Further this configuration gives a good compression efficiency. Furthermore can the receiver the second weighting factor based on the already transmitted Determine second motion vector, which eliminates additional bandwidth needed to create the second weighting factor.

wobei fg_min eine Minimalfrequenz, fg_max eine Maximalfrequenz, |MV2| einen Betrag einer Länge des zweiten Bewegungsvektors zur Erzeugung der zweiten Zwischeninformation und |MV2max| einen Betrag einer maximalen Länge des zweiten Bewegungsvektor beschreibt.In an alternative embodiment, the weighting factors are a low, high or low Bandpass filter formed, whereby an individual adaptation to a content of the image area to be encoded is made possible. In particular, the first weighting factor will be formed by the low-pass filter and the second weighting factor by the high-pass filter. In addition, a cutoff frequency fg of the high and low pass filters can be formed by dividing a length of a second motion vector of the second quality level belonging to the image area to be coded by a maximum length of the motion vector, in particular by the following formula:

where fg_min is a minimum frequency, fg_max is a maximum frequency, | MV2 | an amount of a length of the second motion vector for generating the second intermediate information and | MV2max | describes an amount of a maximum length of the second motion vector.

hereby is in a simple way the determination of the cutoff frequency dependent on the second motion vector of the second quality level achievable. Preferably For example, the coefficients of the filters are selected such that an addition of the filter coefficients at an identical coefficient position the selected one Filters, e.g. High pass and low pass filters, one over one or more filter coefficients constant value, e.g. to one, results. This will result in an over- and / or undervaluation of individual frequency components of the first or second intermediate information avoided.

In In another alternative, the weighting function is generated in such a way that by setting the first weighting factor to zero a1 = 0 and of the second weighting factor to unity a2 = 1 a coupling of the first Intermediate information of the first quality level for coding or Decryption of the second coded data stream is prevented. As a result, the method of coding can be switched off specifically so that coupling in the coding of the first and second quality level is prevented.

Preferably the weighting function is generated such that if one of the Weighting factors One is, the other weighting factor too unequal One chosen becomes. This ensures that at least one of the two weighting Intermediate information muted or reinforced is used, i. the weighting factor, e.g. the first weighting factor a1, unequal to one a1 ≠ 1 is.

alternative or additionally can uses the intermediate information in a location or frequency domain be that, for example, a filtering both in the local, as well as in the frequency domain possible is. This allows the methods for coding and / or decoding coding methods be used, either in the local, or in the frequency domain work.

Preferably the first intermediate information in a reduced spatial or frequency resolution across from the second intermediate information and the first intermediate information before a weighting in one for the location or frequency resolution of second intermediate information identical location or frequency resolution transformed becomes. This ensures that processing the weighted Intermediate information in the third intermediate information at identical Location and / or frequency resolution takes place.

The Procedures can Insert an image area that is defined by an image block, any outlined image object and / or an entire image is described.

In an extension of the coding and / or decoding methods comes with more than two quality levels another coded data stream of quality levels starting with the lowest further quality level encoded or decoded such that the further quality level according to the second quality level is encoded or decoded, with the further quality level lower quality level like the first quality level is treated. Thus, you can the procedures for more than two quality levels be used. Thereby every further quality level becomes analogous to the second quality level processed, with a coupling between the other and the another directly lower quality level analogous to the coupling the second and the first quality level.

• a) Ein erster Enkoderblock zum Bereitstellen einer ersten Zwischeninformation, die den Bildbereich in der ersten Qualitätsstufe repräsentiert, und zum Erzeugen des ersten Datenstroms und zum Erzeugen des zweiten kodierten Datenstroms auf Basis der dritten Zwischeninformation;
• b) Ein zweiter Enkoderblock zum Bereitstellen einer zweiten Zwischeninformation, die einen geschätzten Bildbereich des Bildbereichs in zumindest einem zum Bild zeitlich vorangegangenen oder zeitlich nachfolgenden Bild oder bereits rekonstruierten Bild der zweiten Qualitätsstufe repräsentiert;
• c) Erzeugen einer gewichteten ersten Zwischeninformation durch Gewichten der ersten Zwischeninformation mit zumindest einem ersten Gewichtungsfaktor und einer gewichteten zweiten Zwischeninformation durch Gewichten der zweiten Zwischeninformation mit zumindest einem zweiten Gewichtungsfaktor, wobei der zumindest eine erste Gewichtungsfaktor durch eine vorgebbare Gewichtungsfunktion in Abhängigkeit von zumindest einem der zweiten Gewichtungsfaktoren gebildet wird;
• d) Eine Additionseinheit zum Erzeugen einer dritten Zwischeninformation durch Subtraktion der gewichteten Zwischeninformationen von einer den zu kodierenden Bildbereich der zweiten Qualitätsstufe repräsentierenden Bildinformation.

Furthermore, the invention comprises an encoding device for coding an image area of an image of an image sequence in at least two quality levels, wherein for the image area in a first quality level a first coded data stream with a basic quality and in a second quality level a second coded data stream with one coded together with the first Improved data stream compared to the base quality quality, in particular for carrying out the method of coding, in which the following units are used:

• a) a first encoder block for providing a first intermediate information representing the image area in the first quality level and generating the first data stream and generating the second encoded data stream based on the third intermediate information;
• b) a second encoder block for providing a second intermediate information representing an estimated image area of the image area in at least one temporally preceding or temporally subsequent image or already reconstructed image of the second quality level;
• c) generating a weighted first intermediate information by weighting the first intermediate information with at least a first weighting factor and a weighted second intermediate information by weighting the second intermediate information with at least one second weighting factor, wherein the at least one first weighting factor by a predeterminable weighting function in dependence on at least one of the second Weighting factors is formed;
• d) An addition unit for generating a third intermediate information by subtracting the weighted intermediate information from an image information representing the image area to be encoded of the second quality level.

hit With the aid of the encoding device, the method for coding can be implemented and executable.

• a) Ein erster Dekodierblock zum Dekodieren der ersten kodierten Datenstroms zum Rekonstruieren einer ersten Zwischeninformation, wobei die erste Zwischeninformation einen rekonstruierten ersten Bildbereich der ersten Qualitätsstufe für den Bildbereich repräsentiert;
• b) Ein zweiter Dekodierblock zum Dekodieren des zweiten kodierten Datenstroms zum Rekonstruieren eines rekonstruierten Restfehlers der zweiten Qualitätsstufe und zum Bereit stellen einer zweiten Zwischeninformation, die einen geschätzten Bildbereich des Bildbereichs in zumindest einem zum Bild zeitlich vorangegangenen oder zeitlich nachfolgenden Bild oder bereits rekonstruierten Bild der zweiten Qualitätsstufe repräsentiert;
• c) Eine Gewichtungseinheit zum Erzeugen einer gewichteten ersten Zwischeninformation durch Gewichten der ersten Zwischeninformation mit zumindest einem ersten Gewichtungsfaktor und einer gewichteten zweiten Zwischeninformation durch Gewichten der zweiten Zwischeninformation mit zumindest einem zweiten Gewichtungsfaktor, wobei der zumindest eine erste Gewichtungsfaktor durch eine vorgebbare Gewichtungsfunktion in Abhängigkeit von zumindest einem der zweiten Gewichtungsfaktoren gebildet wird;
• d) Eine Additionseinheit zum Bereitstellen eines rekonstruierten zweiten Bildbereichs (B2') der zweiten Qualitätsstufe für den Bildbereich durch Summation der gewichteten ersten und zweiten Zwischeninformationen und des rekonstruierten Restfehlers der zweiten Qualitätsstufe.

Finally, part of the invention is a decoding apparatus for decoding a first encoded data stream and a second encoded data stream, the first encoded data stream comprising an image area of an image of an image sequence in a first quality level with a base quality and the second encoded data stream having the image area in a second quality level represents quality enhanced with the first encoded data stream relative to the base quality, the encoded data streams being generated in particular according to the encoding method, in particular for performing the decoding method using the following units:

• a) a first decoding block for decoding the first encoded data stream for reconstructing a first intermediate information, the first intermediate information representing a reconstructed first image area of the first quality level for the image area;
• b) A second decoding block for decoding the second encoded data stream for reconstructing a reconstructed residual error of the second quality level and providing a second intermediate information comprising an estimated image area of the image area in at least one temporally preceding or temporally succeeding image or already reconstructed image of the second Quality level represented;
• c) A weighting unit for generating a weighted first intermediate information by weighting the first intermediate information with at least a first weighting factor and a weighted second intermediate information by weighting the second intermediate information with at least one second weighting factor, wherein the at least one first weighting factor by a predeterminable weighting function as a function of at least one of the second weighting factors is formed;
• d) An addition unit for providing a reconstructed second image area (B2 ') of the second quality level for the image area by summing the weighted first and second intermediate information and the reconstructed residual error of the second quality level.

hit The help of the decoding device implements the method for decoding. and executable.

The Invention and its developments are described below with reference to Drawings closer explained.

It demonstrate:

1 an encoder for coding an image area in two quality levels (prior art);

2 a decoder for decoding two coded data streams in two quality levels (prior art)

3 an encoder for encoding the image area in two quality levels with a weighting of intermediate information;

4 a decoder for decoding two coded data streams in two quality levels, taking into account the weighting of the intermediate information;

5 an alternative embodiment of an encoder in which the weighting of values of the intermediate information is performed by means of a high-pass filter or a low-pass filter;

6 an alternative embodiment of an encoder in which the weighting of the intermediate information in the frequency domain is carried out;

7 Representation of a procedure for the interpolation of a frequency coefficient-filled matrix;

8th an encoder for coding an image area in three quality levels with the weighting of the intermediate information;

9 a decoder for decoding three coded data streams in three quality levels, taking into account the weighting of the intermediate information.

Elements with the same function and mode of action are in the 1 9 with the same reference numerals.

In 1 By way of example, an encoder for a scalable video coding with two quality levels Q1, Q2 is shown, wherein these two quality levels encode an image P (n) to be encoded in different resolutions. The reference numeral n is an index for selecting the image P (n) from an image sequence P1, ..., PN, ie, 1 ≤ n ≤ N. A construction according to 1 is known from a document [1]. The coding takes place in each case for an image area PB of the image P (n) to be coded.

In the lower part of 1 is a first encoder block of the encoder for generating a first coded data stream DS1, which represents the image P (n) to be coded encoded in a basic quality. First, by subsampling the image area of the image PB (P (n)) to be coded in the local area by a factor of 2, a first image area B1 (n) is generated by a scanning unit 2DDOWN. The first image area B1 (n) is transferred to a feedback coding loop of the first encoding block. A first motion vector MV1 is determined on the basis of an already coded and reconstructed first image P1 '(n-1) (in 1 not shown). The reconstructed first image P1 'is stored in a first memory module FB1. A first residual error RB1 (n) is formed by subtracting an estimated first image area δ ₁ * MC (P1 '(n-1)) from the first image area B1 (n) at an adder unit AE. The first residual error RB1 (n) is subjected to a first transformation unit T1 for transformation, and the resulting frequency coefficients are supplied to a first quantization unit Q1 for quantization. The quantized frequency coefficients are processed in an entropy coding unit EC and subsequently inserted into the first coded data stream CD1.

By using selection parameters δ ₁ , δ ₂ δ ₃ , a selection of the coding mode to be used can be set. The selection parameters assume the following values:

The selection parameters δ ₁ , δ ₂ are multiplicatively linked to a respective estimated first and second image area MC (P1 '(n-1)), MC (P2' (n-1)).

In the feedback loop, the quantized frequency coefficients are inversely quantized by a first inverse quantization unit IQ1 and subsequently converted by a first inverse transformation unit IT1 into a reconstructed first residual error RB1 '(n). By adding the reconstructed first residual error RB1 '(n) and the estimated first image area δ ₁ * MC (P1' (n-1)) in an addition unit AE, a reconstructed first image area B1 '(n) of the first quality level Q1 is determined. This is then stored organized in the first memory module FB1. The procedure for generating the reconstructed first image area B1 '(n) is applied to different image areas of the image P (n) to be coded so that the reconstructed first image P1' is described by a plurality of reconstructed first image areas.

In the upper part of 1 A second encoding block of the encoder for determining a second encoded data stream DS2 of the second quality level Q2 can be seen. First, an estimate is made for the image area PB of the image P (n) to be coded on the basis of a reconstructed second image P2 '(n - n). 1) performed (not in 1 shown), whereby a second motion vector MV2 is determined. In the following, a reconstructed interpolated first residual error RB1 * (n) is generated by interpolation of the reconstructed first residual error RB1 '(n) by means of an interpolation unit 2DUP. By subtracting an estimated second image area δ ₂ * MC (P2 '(n-1)) by means of the second motion vector MV2 on the basis of the reconstructed second image P2' (n-1), the reconstructed interpolated first residual error RB1 * from the image area PB in an addition unit AE, a second residual error RB2 (n) of the second quality level is formed. For this second residual error, quantized frequency coefficients are generated by a second transformation unit T2 and a second quantization unit Q2. These are transferred to an entropy coding unit EC and subsequently inserted in coded form into the second coded data stream. In the feedback loop of the second coding block, the quantized frequency coefficients are fed to a second inverse quantization unit IQ2 and a second inverse transformation unit IT2, so that a reconstructed second residual error RB2 '(n) is generated. By adding the reconstructed second residual error, the estimated second image area δ ₂ * MC (P2 '(n-1)) and the reconstructed interpolated first residual error RB1 * in an addition unit AE, a reconstructed second image area B2' (n) of the second quality level is formed , which is stored in a second memory module FB2. The procedure for generating the reconstructed second image area B2 '(n) is applied to different image areas of the image P (n) to be coded so that the reconstructed second image P2' is described by a plurality of reconstructed second image areas.

In 2 are a decoder, according to the document [1], to see from the first and the second coded data stream respectively the reconstructed first image P1 '(n) of the first quality level and the reconstructed second images P2' (n) of the second quality level to obtain.

In the lower part of the 2 FIG. 2 shows a first decoder block of the decoder of the first quality level, in which the reconstructed first residual error RB1 '(n) is produced by supplying the first encoded data stream of an entropy decoding unit ED, the inverse first quantization unit IQ1 and the inverse first transformation unit IT1. By adding the reconstructed first residual error RB1 '(n) and the first image area δ ₁ * MC (P1' (n-1)) estimated by means of the transmitted first motion vector MV1 on the basis of the already reconstructed first image P1 '(n-1) generated by an addition unit AE, the reconstructed first image area B1 '(n). The reconstructed first image area is stored in the first memory module FB1. In the following, the reconstructed interpolated first residual error RB1 * (n) is generated by interpolation of the reconstructed first residual error RB1 '(n) in the interpolation unit 2DUP.

In the upper section of the 2 a second decoder block of the decoder of the second quality level Q2 is shown. The second coded data stream CD2 is transferred to the entropy decoding unit ED, the second inverse quantization unit IQ2 and the second inverse transformation unit IT2, so that the reconstructed second residual error RB2 'is generated. By adding an estimated second image area δ ₂ * MC (P2 '(n-1)) by means of the second motion vector MV2 on the basis of the reconstructed second image P2' (n-1), the reconstructed interpolated first residual error RB1 * and the reconstructed second residual error RB2 'in an addition unit AE, the reconstructed second image area B2' (n) of the second quality level is formed.

Become created all reconstructed first and second image areas, thus, the picture P (n) to be coded is rendered in a first quality, i. Basic quality, through the reconstructed first image P1 'and in the second quality reproduced the reconstructed second image P2 '.

The following is based on 3 a first embodiment of the invention explained in more detail. The structure of the encoder for generating the first and second coded data stream according to 3 is similar to the structure of the encoder according to 1 , Therefore, in the following on the peculiarities of the encoder according to 3 compared to the encoder 1 discussed in more detail. The first coded data stream is generated here in the first encoder block EB1 analogously to the embodiments according to FIG 1 , In this case, the reconstructed first image P1 '(n) is referred to as the first intermediate information ZI1. The first intermediate information ZI1 is interpolated, for example by a factor of 2 in the horizontal and vertical image direction, and then each value of the interpolated first intermediate information ZI1 is weighted, eg multiplied, by a first weighting factor a1 in a weighting unit GE. An exact determination of the weighting factors a1, a2 will be discussed in more detail later. The weighting with the first weighting factor a1 produces a weighted first intermediate information GZI1.

An image area δ ₂ * MC (P ₂ '(n-1)) of the second quality level estimated by the motion estimation, ie the estimated second image area, is referred to below as second intermediate information ZI 2 net. This second intermediate information ZI2 is weighted by means of a second weighting factor a2 in a weighting unit GE. In this case, for example, each pixel of the second intermediate information ZI2 is multiplied by the second weighting factor a2. The weighting produces the weighted second intermediate information GZI2. In order to generate the second residual error RE (n), hereinafter referred to as third intermediate information ZI3, a subtraction of the weighted first and second intermediate information GZI1, GZI2 from an image information PI, represented in the image area PB of the image P (n), is subtracted in the second encoder block EB2 subtracted an addition unit AE. In this embodiment, the image information PI is identical to the image area PB. The third intermediate information ZI3 is processed by the second transformation unit, second quantization unit and by the entropy coding unit and the result obtained is inserted into the second coded data stream CD2. If the third intermediate information ZI3 is transformed, quantized, inversely quantized and inversely transformed, the reconstructed second residual error RB2 '(n) results. By adding this reconstructed second residual error RE2 '(n), and the weighted first and third intermediate information GZI1, GZI2, the reconstructed second image area B2' (n) of the reconstructed second image P2 '(n) is generated.

wobei α₁, α₂ ∊ [0; 1] ist. Das bedeutet, dass eine Gewichtungsfunktion GF(a1) = 1 – a2 ist. Die Verwendung der Gewichtungsfaktoren kann nach folgender Vorschrift durchgeführt werden:

• • a2 = 0, a1 = 1: Falls eine Kodierung des Bildes im Intra-Kodiermodus erfolgt, beispielsweise wenn neue Bildteile im Bild auftreten;
• • a2q = 1, a1 = 0: Falls der zu kodierende Bildbereich keine bedeutende Änderungen gegenüber einem korrespondierenden Bildbereich in einem zuvor kodierten Bild aufweist;
• • 0 < a2 < 1, a1 = 1 – a2: Sonst, z.B. bei Verwenden eines Inter-Kodiermodus für bewegte Bildteile.

Through experiments it has been found that good compression can be achieved with good image quality if:

where α ₁ , α ₂ ε [0; 1] is. This means that a weighting function GF (a1) = 1 - a2. The use of the weighting factors can be carried out according to the following rule:

• • a2 = 0, a1 = 1: If the picture is encoded in intra-coding mode, for example when new parts of the picture appear in the picture;
• A2q = 1, a1 = 0: if the image area to be coded has no significant changes from a corresponding image area in a previously coded image;
• • 0 <a2 <1, a1 = 1 - a2: Otherwise, eg when using an inter coding mode for moving picture parts.

wobei a_{2_min} einem minimalen Wert von a2, a_{2_max} einem maximalen Wert von a2, |MV2| einem Betrag, d.h. einer Länge, des zweiten Bewegungsvektors MV2 und |MV2_max| einem maximalen Betrag des zweiten Bewegungsvektors MV2 entspricht.The weighting factors are determined in one embodiment on the basis of the motion vectors MV1, MV2 as follows:

where a _{2_min is} a minimum value of a2, a _{2_max is} a maximum value of a2, | MV2 | an amount, ie, a length, of the second motion vector MV2 and | MV2 _max | corresponds to a maximum amount of the second motion vector MV2.

With a _{2_min} = 0.2, a _{2_min} = 1.0, | MV2 | = 32, | MV2 _max | = 64 results in a2 = 0.6. Here, a _{2_min is} a constant which causes the value of the second weighting factor a2 to become not larger than the constant itself. Thus, a1 = 1 - a2 = 0.4.

A decoding according to 3 generated first and second coded data streams CD1, CD2 is based on an embodiment according to 4 shown in more detail. Here, the generation of the reconstructed first image area B1 'and the first and second residual errors RB1', RB2 'in the first decoder block DB1 is analogous to the embodiments according to FIG 2 , The first intermediate information ZI1 is interpolated by the interpolation unit 2DUP, for example by a factor of 2 in horizontal and vertical image direction, and then each coefficient of the interpolated first intermediate information ZI1 is multiplied by the first weighting factor a1 in the weighting unit GE. The weighting with the first weighting factor a1 produces the weighted first intermediate information GZI1. The second coded data stream CD2 is then decoded by means of the second decoder block DB2 and the reconstructed second image area B2 '(n) is created. For this purpose, the image range δ ₂ * MC (P ₂ '(n-1)) of the second quality level estimated by the motion estimation, ie

the estimated second image area or the second intermediate information ZI2, weighted by the second weighting factor a2 in the weighting unit GE, so that the weighted second intermediate information GZI2 is calculated. The reconstructed second image area B2 '(n) is added by adding the weighted second intermediate information GZI2, the reconstructed second residual error RB2 'and the weighted first intermediate information GZI1 obtained in the addition unit AE.

The Weighting factors a1, a2 are given with the respective coefficients the intermediate information multiplied by the coefficients For example, brightness and / or color values of the respective pixels represent the intermediate information. The respective weighting factor can change from one to another coding image area and / or within the image area to be coded changed become.

In the above embodiment according to the 3 and 4 a fixed first and second weighting factor a1, a2 was used. In an extension or alternative embodiment to the above example, the weighting factors are replaced by filters, eg high pass or low pass filters. 5 shows by way of example the use of such filters, wherein the first weighting factor a1, ie the weighting factor which is used after the interpolation of the reconstructed first image area B1 '(n), is replaced by a low-pass filter. The second weighting factor a2 is replaced by the high-pass filter HP. A use of a filter, for example a high-pass filter HP, for filtering coefficients in the local area, such as the estimated first image area, is well known to a person skilled in the art, so that will not be discussed further. For decoding with the encoder according to 5 generated first and second coded data streams is analogous to the embodiments according to 4 proceeding, wherein the first weighting factor a1 is replaced by the low-pass filter TP and the second weighting factor a2 by the high-pass filter HP.

In a further embodiment, a weighting of the first and second intermediate information is used instead of in the local area, such as in 5 , executed in the frequency domain. This is based on 6 explained in more detail. 6 generates the first and second coded data streams in two quality levels.

Of the lower part of the picture shows the first encoder block. Here is the to be encoded image area PB or the image to be encoded P (n) in the weighting unit GE low - pass filtered TP and then in Local area by a factor of 2 in horizontal and vertical direction sub-sampled, see sample unit 2DDOWN, so that the first image area B1 is generated. The first motion vector MV1 is based on previously subsampled images, ie. by searching in at least one filed in the first memory FB1 Image. The first image area B1 is organized in the first memory FB1 stored. By subtracting the estimated first image area MC (B1 (n-1)) of the first image area B1 is in the addition unit AE of first residual error RB1 (n) generated, which is the first transformation unit T1 supplied becomes. This gives rise to the first transformation coefficients TK1. The coded first data stream CD1 is used in this embodiment by applying the first quantization unit Q1 to the first transformation coefficients TK1 and a subsequent execution the Entropiekodierungseinheit EC formed. The first transformation coefficients TK1 correspond to the first intermediate information ZI1.

Through the upper part of 6 the second data stream CD2 is generated by the second encoder block. The image area PB to be coded or the image P (n) of the higher quality level to be coded, ie the second quality level Q2, is referred to as image information PI. For this image information PI, a second motion vector MV2 is searched in uncoded image information of the second memory module FB2. The second image area MC (P (n-1)) estimated with the aid of the second motion vector MV2 is transform-coded, see T2, whereby the second intermediate information ZI2 is created. By weighting the second intermediate information ZI2 with the high-pass filter HP in the weighting unit GE, the weighted second intermediate information GZI2 is generated.

Parallel to do this, the first transform coefficients TK1 (n), i. the first intermediate information, by the interpolation unit 2DUP interpolated, e.g. by a factor of 2, and then with the low-pass filter TP in the weighting unit GE coefficient-wise weighted. This becomes the weighted first intermediate information GZI1 created.

By Subtraction of the first and second weighted intermediate information GZI1, GZI2 of second transformation coefficients TK2 in the addition unit AE, wherein the second transformation coefficients TK2 by application the second transformation unit T2 to the image area to be coded PB, the weighted third intermediate information GZI3 determined. The third weighted intermediate information will be the second quantization unit and the entropy coding unit EC supplied, whereby the second data stream CD2 is generated.

The filters HP and TP are formed such that when weighting a coefficient of the first Intermediate information ZI1 at a first position with the first weighting factor a1 and a coefficient of the second intermediate information ZI2 at the first position with the second weighting factor a2 is a sum of the two weighting factors constant, in particular equal to one.

wobei fg_min eine Minimalfrequenz, fg_max eine Maximalfrequenz, |MV2| einen Betrag einer Länge des zweiten Bewegungsvektors zur Erzeugung der zweiten Zwischeninformation (ZI2) und |MV2max| einen Betrag einer maximalen Länge des zweiten Bewegungsvektor beschreibt.Furthermore, a cutoff frequency fg of the high and low pass filters HP, TP is formed by dividing a length of a motion vector MV2 belonging to the image area to be coded, ie the second motion vector, of the second quality level by a maximum length of the motion vector MV2max, in particular by the following formula:

where fg_min is a minimum frequency, fg_max is a maximum frequency, | MV2 | an amount of a length of the second motion vector for generating the second intermediate information (ZI2) and | MV2max | describes an amount of a maximum length of the second motion vector.

In the statements according to 6 The first transformation coefficients TK1 were subjected to interpolation. In the context of this invention, this interpolation is performed in such a way that the frequency range in each dimension is extended by the interpolation factor, for example a factor 2, and the newly added transformation coefficients are pre-assigned a value of 0. For example, if the first transform coefficients TK1 represent a two-dimensional frequency matrix having the maximum frequencies FX and FY, the frequency matrix changes such that the maximum frequencies are 2xFX, 2xFY. In this case, all transformation coefficients are set to 0, except for the first transformation coefficients TK1 adopted. In 7 this relationship is represented symbolically, wherein the newly added by the Inpolation transformation coefficients, which are set to 0, hatched have been marked.

The presented embodiments can for one Image area, for example, a block, a macroblock or corresponds to an arbitrarily bordered image object of the image P, or for the entire image P are used. It can be used for different image areas different weighting factors and / or filters used become.

In a further extension or alternative embodiment, the method according to the invention may, in addition to a local scalability, as shown in FIGS 3 and 4 , also for quality scalability (SNR scalability). For this purpose, analogous to the statements according to 3 and 4 proceeded, wherein the processing steps subsampling and interpolating are omitted. The SNR scalability is characterized in that the quantization factors used in the first and second quantization units Q1, Q2 or in the inverse quantization units IQ1 IQ2 are tuned such that in the first quality level a coarsely quantized and in the second quality level fine quantized image or image area can be generated or reconstructed. The use of SNR scalability is known to a person skilled in the art, for example, by coarse grainularity scalability (CGS) from [1].

In the previous embodiments, two quality levels were considered. However, the present invention is not limited to two quality levels, but can be used for several quality levels, each with two quality levels by means of a pair of weighting factors, for example, the first and second weighting factors a1, a2 according to 3 be coupled.

In the following 8th and 9 By way of example, the use of the method according to the invention is carried out on three quality levels Q1, Q2, Q3. In this case, the image area PB to be coded is subsampled once, resulting in a second image area B2. For generating the first and second coded data streams of the first and second quality levels, the second image area represents an image area of the higher quality level, ie the second quality level Q2, to be coded, this image area to be coded corresponding to the image information PI. The first and second quality level is analogous to the statements according to 3 carried out.

The third quality level Q3 works analogously to the second quality level according to 3 in which the original non-subsampled image area PB to be coded is used as the uncoded image information PI. Here, a weighted fourth intermediate information GZI4 is obtained by multiplying a third weighting factor a3 by the reconstructed and interpolated second image B2 '(n), ie a fourth intermediate information formation ZI4, formed. Further, a weighted fifth intermediate information GZI5 is prepared by weighting a fourth weighting factor a4 with an estimated third image area MC (P3 '(n-1)), where P3' is a reconstructed third image. A sixth intermediate information ZI6 is generated by subtracting the fourth and fifth weighted intermediate information GZI4, GZI5 from the image area PB, ie the image information PI. The third coded data stream CD3 is formed by applying a third transformation unit T3, a third quantization unit Q3 and the entropy-coding unit based on the sixth intermediate information ZI6. In the feedback loop, the sixth intermediate information after passing through the third transformation unit and the third quantization unit is formed by executing a third inverse quantization unit IQ3 and an inverse third transformation unit IT3, a reconstructed third residual error RB3 '. By adding the reconstructed third residual error RB3 'and the weighted fourth and fifth intermediate information in the addition unit AE, a reconstructed third image area B3' (n) can be obtained.

In 9 is a decoder with three decoding blocks to watch, each decode block decodes each one of the encoded data streams in one of the three quality levels. The procedure for decoding the first and second coded data streams CD1, CD2 is analogous to the embodiments according to FIG 4 , The reconstructed second image area B2 'corresponding to the fourth intermediate information ZI4 is interpolated by the interpolation unit 2DUP and multiplied by the third weighting factor a3. This results in the weighted fourth intermediate information GZI4. In the third quality level Q3, the weighted fifth intermediate information GZI5 is analogous to the exemplary embodiment according to FIG 8th generated. The reconstructed third image area B3 'is obtained by summing the weighted fourth and fifth intermediate information GZI4, GZI5 and the reconstructed third residual error RB3' in the addition unit AE. The reconstructed third residual error results after carrying out the entropy decoding unit ED, the inverse third quantization unit IQ3 and the inverse third transformation unit IT3 of the third encoded data stream CD3 (n).

• – Bestimmen des zweiten Gewichtungsfaktors a2 wie bei Verwendung von zwei Qualitätsstufen, d.h.
  
• – Zur Bestimmung des dritten Gewichtungsfaktors a3 wird analog zu den Ausführungen gemäß dem zweiten Gewichtungsfaktors a2 vorgegangen, a_{3_min} einem minimalen Wert von a3, a_{3_max} einem maximalen Wert von a3, |MV3| einem Betrag, d.h. einer Länge, des dritten Bewegungsvektors MV3 und |MV3max| einem maximalen Betrag des dritten Bewegungsvektors MV3 entspricht,
  
• – Für die Ermittlung des ersten Gewichtungsfaktors gilt gemäß der vorgegeben Gewichtungsfunktion GF weiterhin:
  

Using three quality levels, the following procedure is used to determine the weighting factors a1, a2, a3:

• Determining the second weighting factor a2 as when using two quality levels, ie
  
• The procedure for determining the third weighting factor a3 is analogous to the statements in accordance with the second weighting factor a2, _{a3_min} a minimum value of a3, _{a3_max} a maximum value of a3, | MV3 | an amount, ie, a length, of the third motion vector MV3 and | MV3max | corresponds to a maximum amount of the third motion vector MV3,
  
• - For the determination of the first weighting factor according to the predetermined weighting function GF continues to apply:
  

in this connection The inventors have recognized that a coding result is very good results supplies, if any, the one motion vector who becomes the higher resolution level belongs, i.e. the motion vectors of the second and third resolution levels.

• – Zumindest einer der Gewichtungsfaktoren wird in zumindest einem der kodierten Datenströme übertragen, sodass der zumindest eine Gewichtungsfaktor durch Auslesen des dazugehörigen kodierten Datenstroms gewonnen werden kann;
• – Die Bestimmung der Gewichtungsfaktoren kann anhand der in den kodierten Datenströmen kodierten Bewegungsvektoren erfolgen.

In determining the weighting factor (s) for use in the method of decoding in a receiver, the following can be done:

• - At least one of the weighting factors is transmitted in at least one of the coded data streams, so that the at least one weighting factor can be obtained by reading the associated coded data stream;
• The determination of the weighting factors can take place on the basis of the motion vectors coded in the coded data streams.

In an alternative embodiment, the determination of the weighting factors is carried out by means of a rate-distortion optimization. It is for the second weighting factor a2 in several Stu in the range [0; 1] with a1 = 1 - a2 the encoding of the second encoded data stream is performed and the one selected for a2, which has the best combination of image quality, for example in PSNR (PSNR = Peak Signal-to-Noise Ratio), versus a low bit rate of achieved second coded data stream.

In An extension of the invention is a coupling of the first intermediate information (ZI1) of the first quality level (Q1) for encoding or decoding the second encoded data stream (CD2) preventable. By setting those weighting factors to zero, e.g. a1 = a3 = 0, which is the coupling between the quality levels can be achieved that the respective encoder or decoder blocks independently work from each other.

The Coding method is by means of an ENC encoding device and the method for decoding is by means of a decoding device feasible and executable. Both devices are available in hardware, or software, on one Processor with memory and input and / or output units has, or in a mix of hardware and software implemented. The devices are usable as part of a broadcasting and / or streaming application, wherein the encoding device is in a network element, e.g. one Broadcasting server, and the decoding device in a terminal, in particular a portable device, like a mobile phone or a PDA (PDA - Personal Digital Assistant), can be integrated. In this case, a transmission of the first and second streams by means of an MBMS protocol (MBMS - Multimedia Broadcast / Multicast Service) or over DVB (DVB - Digital Video broadcast).

bibliography

• [1] ITU-T, ISO / IEC JTC1, "Joint Scalable Video Model 6 ", JVT-R202, January 2006

Claims (16)

Method for coding an image area (PB) of an image (P (n)) of an image sequence (P1, ..., PN) in at least two quality levels, wherein for the image area (PB) in a first quality level (Q1) a first coded data stream (CD1) are generated with a basic quality and in a second quality level (Q2) a second coded data stream (CD2) with a quality improved with respect to the base quality together with the first coded data stream (CD1), characterized in that the following steps are carried out: a ) Providing a first intermediate information (ZI1) representing the image area (PB) in the first quality level (Q1); b) providing a second intermediate information item (ZI2) which stores an estimated image area (MC (P2 '(n-1)) of the image area (PB) in at least one image (P (n)) which precedes or follows in time (P (P) n - 1), P (n + 1)) or already reconstructed image (P2 '(n - 1)) of the second quality level (Q2); c) generating a weighted first intermediate information by weighting the first intermediate information (ZI1) with at least a first weighting factor (a1) and a weighted second intermediate information (GZI2) by weighting the second intermediate information (ZI2) with at least one second weighting factor (a2), the at least one first weighting factor (a1) being determined by a predeterminable weighting function (GF (a2)) is formed as a function of at least one of the second weighting factors (a2); d) generating a third intermediate information (ZI3) by subtracting the weighted intermediate information (GZI1, GZI2) from an image information (PI) representing the image area (PB) of the second quality level (Q2) to be coded; e) generating the second coded data stream (CD2) on the basis of the third intermediate information (ZI3). Method for decoding a first encoded data stream (CD1) and a second encoded data stream (CD2), wherein the first encoded data stream (CD1) comprises an image area (PB) of an image (P (n)) of an image sequence (P1, ..., PN ) in a first quality level (Q1) of a basic quality and the second coded data stream (CD2) represents the image area (PB) in a second quality level (Q2) with a quality improved with respect to the base quality together with the first coded data stream (CD1), wherein the coded data streams (CD1, CD2) have been generated in particular according to claim 1, or claim 1 and at least one of claims 3 to 14, characterized in that the following steps are carried out: a) decoding the first coded data stream (CD1) to reconstruct a first one Intermediate information (ZI1), wherein the first intermediate information comprises a reconstructed first image area (B1 ') of the first quality level (Q1) for the image area (PB) r epräsentiert; b) providing a second intermediate information item (ZI2) which stores an estimated image area (MC (P2 '(n-1)) of the image area (PB) in at least one image (P (n)) which precedes or follows in time (P (P) n - 1), P (n + 1)) or already reconstructed image (P2 '(n - 1)) of the second quality level (Q2); c) decoding the second encoded data stream (CD2) to reconstruct a reconstructed residual error (RB2 ') the second quality level (Q2); d) generating a weighted first intermediate information by weighting the first intermediate information (ZI1) with at least one first weighting factor (a1) and one weighted second intermediate information (GZI2) by weighting the second intermediate information (ZI2) with at least one second weighting factor (a2), wherein the at least one first weighting factor (a1) is formed by a predefinable weighting function (GF (a2)) as a function of at least one of the second weighting factors (a2); e) providing a reconstructed second image area (B2 ') of the second quality level (Q2) for the image area (PB) by summing the weighted first and second intermediate information (GZI1, GZI2) and the reconstructed residual error (RB2') of the second quality level. Method according to one of the preceding claims, characterized characterized in that the weighting function (GF) generates such is that the first weighting factor of a coefficient of the first Intermediate information (ZI1) at a first position and the second Weighting factor (a2) of a coefficient of the second intermediate information (ZI2) at the first position in total a predetermined value, in particular to one, where the respective weighting factors (a1, a2) in particular in the range between zero and one [0; 1] can be selected. Method according to one of the preceding claims, characterized characterized in that at least two coefficients of the first and / or second intermediate information (ZI1, ZI2) each having different ones Weighting factors are weighted. Method according to one of the preceding claims, characterized in that the second weighting factor (a2) of the second quality level (Q2) is formed by dividing a length of a motion vector of the second quality level belonging to the image area to be coded by a maximum length of the motion vector, in particular by the following Formula:

where a _{2_min is} a minimum value of a2, a _{2_max is} a maximum value of a2, | MV2 | an amount of a length of the second motion vector MV2 and | MV2max | corresponds to a maximum amount of a length of the second motion vector MV2. Method according to one of claims 1 to 4, characterized that the weighting factors (a1, a2) by a low, high or Bandpass filter (TP, HP) are formed. Method according to Claim 6, characterized the first weighting factor (a1) is determined by the low-pass filter (TP) and the second weighting factor (a2) is formed by the high-pass filter become. A method according to claim 7, characterized in that a cutoff frequency (fg) of the high and low pass filter (HP, TP) by dividing a length of belonging to the image portion to be encoded second motion vector (MV2) of the second quality level by a maximum length of the motion vector ( MV2max) is formed, in particular by the following formula:

where fg_min is a minimum frequency, fg_max is a maximum frequency, | MV2 | an amount of a length of the second motion vector for generating the second intermediate information (ZI2) and | MV2max | describes an amount of a maximum length of the second motion vector. Method according to one of claims 1 to 4, characterized in that the weighting function (GF) is generated such that by setting the first weighting factor (a1) to zero a1 = 0 and the second weighting factor (a2) to Eins a2 = 1 a coupling the first intermediate information (ZI1) of the first Quali tätsstufe (Q1) for encoding or decoding the second encoded data stream (CD2) prevented becomes. Method according to one of the preceding claims, characterized characterized in that the weighting function (GF) generates such is that if one of the weighting factors (a1) is one, the other weighting factor (a2) is chosen to be unequal to one. Method according to one of the preceding claims, characterized characterized in that the intermediate information (ZI1, ZI2) in one Local or frequency range can be used. Method according to one of the preceding claims, thereby marked that the first intermediate information (ZI1) in a reduced spatial or frequency resolution compared to the second intermediate information (ZI2) is provided, the first intermediate information (ZI1) before a weighting in one for the location or frequency resolution of second intermediate information (ZI2) identical location or frequency resolution transformed becomes. Method according to one of the preceding claims, characterized in that through the image area (PB) an image block, an arbitrarily bordered image object and / or an entire image described becomes. Method according to one of the preceding claims, characterized characterized in that with more than two quality levels (Q1, Q2, Q3) another coded data stream (CD3) of quality levels starting with the lowest further quality level (Q3) is encoded or decoded such that the further quality level according to the second quality level is encoded or decoded, with the further quality level (Q3) lower quality level (Q2) as the first quality level is treated. Encoding device (ENC) for encoding an image area (PB) of an image (P (n)) of an image sequence (P1, ..., PN) in at least two quality levels, being for the image area (PB) in a first quality level (Q1) a first coded Data stream (CD1) with a basic quality and in a second quality level (Q2) a second coded data stream (CD2) with one together with the improved the first coded data stream (CD1) compared to the base quality quality are generated, in particular for carrying out a method according to claim 1 or claim 1 and at least one of claims 3 to 14, thereby marked that the following units are used: a) A first encoder block (EB1) for providing a first intermediate information (ZI1) representing the image area (PB) in the first quality level (Q1) represents, and generating the first data stream (CD1) and generating the second coded data stream (CD2) based on the third intermediate information (ZI3); b) A second encoder block (EB2) for providing second intermediate information (ZI2) representing an estimated image area (MC (P2 '(n - 1)) of the Image area (PB) in at least one of the image (P (n)) temporally preceding or temporally subsequent image (P (n-1), P (n + 1)) or already reconstructed image (P2 '(n - 1)) of the second quality level (Q2) represents; c) Generating a weighted first intermediate information by weighting the first intermediate information (ZI1) with at least a first Weighting factor (a1) and a weighted second intermediate information (GZI2) by weighting the second intermediate information (ZI2) with at least a second weighting factor (a2), wherein the at least one first Weighting factor (a1) by a predeterminable weighting function (GF (a2)) dependent on formed by at least one of the second weighting factors (a2) becomes; d) An addition unit (AE) for generating a third one Intermediate information (ZI3) by subtracting the weighted intermediate information (GZI1, GZI2) of an image area (PB) of the second quality level to be coded (Q2) Image information (PI); Decoding device (DEC) for decoding a first coded data stream (CD1) and a second coded data stream (CD2), the first coded data stream (CD1) comprising an image area (PB) of an image (P (n)) of a picture sequence (P1, .. ., PN) in a first quality level (Q1) with a basic quality and the second coded data stream (CD2) the image area (PB) in a second quality level (Q2) with a quality improved together with the first coded data stream (CD1) compared to the base quality wherein the coded data streams (CD1, CD2) have been generated in particular according to claim 1 or claim 1 and at least one of claims 3 to 14, in particular for carrying out a method according to one of claims 2 to 14, characterized in that the following units are used a) A first decoding block (DB1) for decoding the first encoded data stream (CD1) for reconstruction first intermediate information (ZI1), the first intermediate information representing a reconstructed first image area (B1 ') of the first quality level (Q1) for the image area (PB); b) a second decoding block (DB2) for decoding the second encoded data stream (CD2) for reconstructing a reconstructed residual error (RB2 ') of the second quality level (Q2) and for providing a second intermediate information (ZI2) comprising an estimated image area (MC (P2 '(n-1)) of the image area (PB) in at least one image (P (n-1), P (n + 1)) temporally preceding or temporally following the image (P (n)) or already reconstructed image (P2 c) A weighting unit (GE) for generating a weighted first intermediate information by weighting the first intermediate information (ZI1) with at least a first weighting factor (a1) and a weighted second intermediate information (Z1) GZI2) by weighting the second intermediate information (ZI2) with at least one second weighting factor (a2), wherein the at least one first weighting factor (a1) is determined by a predeterminable weighting function (GF (a2)) in A dependence of at least one of the second weighting factors (a2) is formed; d) an addition unit (AE) for providing a reconstructed second image area (B2 ') of the second quality level (Q2) for the image area (PB) by summing the weighted first and second intermediate information (GZI1, GZI2) and the reconstructed residual error (RB2') the second quality level.