DE102004019339A1 - Prediction method and associated method for decoding a prediction method, associated encoding device and decoding device - Google Patents

Abstract

The invention relates to a prediction method in which a prediction of a current image block of an original image is encoded in a first-resolution image coding method, wherein the first resolution of the prediction is adjusted such that the first resolution is adaptable to the current image block in arbitrary accuracy set first resolution is signaled. Furthermore, the invention relates to an encoding device for carrying out a prediction method. Furthermore, the invention relates to a decoding device for decoding a prediction method.

Description

The The invention relates to prediction methods according to the preamble of claim 1. The invention also relates to a method for decoding a prediction procedure according to the preamble of Claim 14. Furthermore, the invention relates to an encoding device to carry out a prediction procedure according to the generic term of claim 15. Furthermore, the invention also relates to a Decoding device for decoding a prediction method according to the preamble of Claim 17.

image coding have the task, among other things, that for a picture or a sequence needed from pictures To reduce data volume by suitable compression methods. Here is a possible Compression methods, such as standard H.263, ITU-T recommandation H.263, "Video Coding for Low Bitrate Communication ", February 1998, referenced. Use such compression methods among other things a temporal and / or local prediction for one or more image blocks. With Help of temporal and / or local Prediction is it is possible temporal or local Redundancy of an image or a sequence of images during compression exploit. By using the temporal and / or local prediction can the needed Data volume can be significantly reduced.

ever more precisely, the prediction for one Image block is, the smaller is a difference or error to the one Reference picture image block used for this prediction is used. As a result, through an encoding This image block either a coding error for this image block clotting ger fails or for one Residual error coding a smaller volume of data is needed.

The The problem underlying the invention is a prediction of an image block in a picture coding method in a simpler and more efficient manner Way to enable.

These Task is in each case based on the preamble of the claim 1 or 14 or 15 or 17 solved by its characterizing features.

at a prediction method be according to claim 1 a prediction a current image block of an original image in a picture coding method with a first resolution encoded, with the first resolution the prediction is set so that the first resolution in any accuracy is customizable to the current image block, and the set first resolution signaled.

By the inventive method is it possible the first resolution in the prediction of a to set the current picture block individually. It will only possible the first resolution to choose such that the prediction the current image block optimal, that is, for example, with a minimum Error, can be set. Thus, it is only possible by means of the method according to the invention, the first resolution in any accuracy and thus adaptive to the properties of the current image block to be encoded.

Thereby, that the first resolution for the prediction of the current image block is signaled by means of a resolution value, is a pre-set of possible resolution levels the prediction unnecessary. Much more can be any first resolution by means of the resolution value be signaled.

There the first resolution the prediction can be adjusted individually, it is by the inventive method possible, the first resolution also dependent from one available standing resource, such as Computing power, to scale. In the case, that, for example, little computing power to determine the prediction a current image block is available, the prediction by means of a coarse first resolution be determined. If, however, much computing power is available, so can also be a very fine resolution, possibly a very complex computational algorithm needed for the determination, for creation the first resolution the prediction to be used.

Preferably is represented by a horizontal and a vertical resolution component the first resolution set differently in the horizontal and vertical directions, causing the prediction for one current image block can be adjusted more accurately than when using an identical resolution in vertical and horizontal directions.

Preferably becomes the first resolution for many image blocks which only transmits one resolution value for multiple image blocks and therefore less data volume is needed for signaling, as for the case in the for each image block in each case a resolution value is used.

Preferably, the first resolution of the prediction of the current image block is determined as a function of a second resolution of the prediction of an adjacent second image block, which possibly results in an optimal resolution of the prediction of the current image block with less computation Is found.

Becomes the first resolution the prediction of the current image block depending on a quantization parameter of the current image block determined can an optimal resolution of prediction of the current image block can be found faster.

Becomes continue the first resolution the prediction of the image block depending on determined from a block structure of the current image block, so can determining an optimal resolution of the prediction of the current image block simplified and thus accelerated.

Preferably becomes the first resolution the prediction of the current image block depending on a Image format of the original image determined, for example, at larger image formats a larger resolution required will, especially for larger image formats occurring disturbing To avoid coding errors.

Becomes the first resolution the prediction of the current image block depending on from a temporal distance of the original image to one for the prediction used reference image determined, so can, for example, the first resolution the prediction from a distant time Reference image determined in a larger resolution than from a short time ago reference image. hereby can the needed Computing power to determine the first resolution can be reduced.

Becomes the first resolution the prediction of the current image block depending on of a length of one associated Motion vector determined, so can the computational effort for the implementation of the prediction be reduced.

Preferably becomes the first resolution with the help of a resolution value transmitted in a first compressed image signal, as such a simple assignment between the resolution value and its associated current image block allows becomes.

alternative or additionally becomes the first resolution with the help of a resolution value transmitted separately to a second compressed image signal, because hereby the resolution value with a first error protection and the second compressed image signal with a second error protection for a transmission can be provided for example, more errors can be corrected with the first error protection are as with the second error protection.

Preferably will reflect the resolution value Data is compressed, allowing a data volume to be transferred reproducing data is reduced.

Preferably will be at least one resolution value encoded differentially so that a data volume to be transmitted for the resolution value is reduced.

Of Furthermore, the invention also relates to a method for decoding a prediction procedure, with which it makes possible becomes, a prediction method decode.

Farther The invention also relates to an encoding device for carrying out a prediction method with a processing unit for encoding a prediction a current image block of an original image in a picture coding method with a first resolution, being the first resolution the prediction is set so that the first resolution in any accuracy is customizable to the current image block, and signaling the set first resolution. This makes it possible prediction perform.

In addition, can the invention an encoding device with a transmitting unit for Send the first resolution by means of a resolution value include.

Preferably The invention includes a decoding device for decoding a prediction method, whereby a decoding of a prediction method is feasible.

In addition, can the invention also a decoding device with a receiving unit for accepting a resolution value, being the first resolution with the help of the resolution value is illustrated.

Further details and advantages of the invention will become apparent from the background of the prior art 2 based on 1 . 3 . 4 and 5 explained. In detail show:

1 an encoding device, a decoding device, a camera, an output unit and a transmission medium for carrying out the method according to the invention;

2 a sequence of digitized images stored in a memory and a division of the respective digitized images into image blocks to perform a prediction;

3 a section of an image in which a first resolution in horizontal and vertical direction has been increased for carrying out the prediction method according to the invention;

4 a representation of a possible arrangements of a resolution value;

5 a detailed representation of a search area block, wherein the resolution of the search area block has been increased.

elements with the same function and mode of action are in the figures, respectively provided with the same reference numerals.

In 1 An encoding device EV, a transmission medium UEM and a decoding device DV are shown. The encoding device EV comprises a processing unit VE for carrying out the method according to the invention and for compressing at least one image. In addition, the encoding device EV has a transmitting unit SE for sending one or more compressed video signals, such as a first and second compressed video signal BS1 or BS2, as well as for transmitting at least one resolution value AW. In addition, the encoding device EV comprises a first memory module SP1, which is used by the processing unit VE and / or the transmission unit SE for temporarily storing data. In addition, a camera K can be connected to the encoding device EV, which receives an image or a sequence of images and transfers them to the processing unit VE for further processing. The processing unit VE, the transmission unit SE, the first memory module SP1 and the camera K are interconnected by means of an encoding-side connection network EX for the purpose of exchanging data.

The Decoding device DV initially comprises a receiving device ED for receiving one or more compressed video signals, such as. a first and second compressed video signal BV1 or BV2, as well as at least one resolution value AW. In addition, the decoder has DV over a processing unit VD having the task received and to analyze compressed first and second video signals BV1 and BV2, respectively and from this using the method according to the invention an image or to reconstruct a sequence of images. In addition, received resolution values AW analyzed and for image reconstruction by the processing unit VD are used. possibly One or more reconstructed images are sent to the decoding device via the DV connected output unit DD, e.g. a monitor, to View be brought. additionally the decoding device DV comprises a second memory module SP2, that from the processing unit VD and / or the receiving unit ED is used for caching data. The processing unit VD, the second memory module SP2, the receiving unit ED are by means of a decoding-side interconnection network DX for the purpose of replacement connected by data.

The transmission medium UEM is for transmission one or more compressed video signals, e.g. a first or second compressed video signal BS1 or BS2, as well as for transmission one or more resolution values AW, from the encoder EV to the decoder DV.

The Encoding device EV and / or the decoding device DV can in a mobile device according to GSM standard (GSM - Global System for Mobile Communications) or UMTS standard (UMTS - Universal Mobile Telecommunications System) as well as in a computer unit, possibly in a portable device is integrated, housed. To transmit the compressed first and second video signals BS1 and BS2 and the resolution values, respectively AW between the encoding device EV and the decoding device DV may, for example, be a wireless radio network, such as e.g. after this GSM standard, as well as a wired transmission medium, such as an IP-based network (IP - Internet Protocol) or IDSN (ISDN - Integrated Services Digital Network).

Next The possibility, the first and second compressed video signals BV1 and BV2 together with the corresponding ones Resolution values AW from the encoder EV directly to the decoder DV can be useful in practice, the compressed first and second video signals BS1 and BS2 and the associated resolution values AW after encoding on a storage medium, as example, a CD (Compact Disc Disc) or a video server, for later use.

According to the present invention, a picture coding method is a method in which a prediction in the processing of at least one picture is required in at least one method step. The term "picture coding method" particularly encompasses the compression of an image or a sequence of images, as can be found, for example, in the standard H.263, ITU-T Recommendation H.263, "Video Coding for Low Bit-Rate Communication", February 1998. With the help of 2 the term prediction is explained in more detail below.

2 shows two images stored in a memory module, such as the first memory module SP1. On the right side there is an original picture OB, which consists of 6 lines of 6 pixels each. Thus, the original image OB comprises 6 × 6 pixels. On the left side, a reference image RB, such as a predecessor image, is seen, which was taken in a sequence of images directly in front of the original image OB. A pixel is the smallest pixel that contains part of the image information of the image. For example, a pixel represents a gray value with an accuracy of 8 bits. In a further step, a number of pixels in the horizontal and vertical direction are combined into image blocks. So be in 2 Image blocks of 2 × 2 pixels each, such as a first image block BB1, a second image block BB2 or an m-th image block BBM, formed.

to prediction but not a recorded uncoded predecessor image the sequence of images used. For the prediction of the original image OB is a reference image RB is used, which from the decoding an encoded predecessor image is produced. This reference image RB is after the transmission the first or second compressed video signal BS1 or BS2 also in the processing unit VD of the decoding device DV for Available. It is also possible, that one or more decoded image blocks of the original image, the have been decoded after their encoding, for prediction be used.

In order to determine the optimal prediction for an image block BB1... BBM in the original image OB, an attempt is made to find an image block in the reference image RB which is most similar to the searched image block. The following example will follow 2 attempts to find an image block in the reference image RB for a current image block BBN of the original image OB, for which the prediction can best be realized. For this purpose, first the position of the current image block BBN in the reference image RB is determined. A reference picture block BBNV of the reference picture RB is in the same place as the current picture block BBN in the original picture OB. In a further step, a search area SB is placed around the reference image block BBNV, which is stored in the 2 is indicated by a dashed rectangle. In this searchable SB SB is searched for the optimal prediction for the current image block BBN. Thus, an error criterion may possibly be used as a selection criterion for the optimal prediction, in which case the summation over the square of the differences of the respective pixels results in the smallest value for two image blocks to be compared. In a further step, the error measure for each possible combination of a 2 × 2 image block of the reference image RB, which is located within the search area SB, is then determined using the current image block BBN of the original image OB. Such an image block in the reference image RB is referred to as a search area block, such as a first search area block SBB1. Thus, for example, first the error measure for the first search area block SBB1 of the reference image RB is formed with the current image block BBN of the original image OB. After the determination of all possible combinations, for example, the optimal search area block SBBX is found, which has the smallest error measure compared to the current image block BBN.

In practice, it is expedient not to transmit now the current image block BBN of the original image OB, but only an image vector BV which indicates the position of the optimal search area block SBBX. The motion vector BV can be formed from a horizontal and vertical subvector BVH or BVV. The use of the motion vector BV allows, for example, a significant data reduction in image coding methods, which are intended to compress images. In the example below 2 the image vector BV indicates the following values: BV (BVH, BW) = (1, 1);

in the Generally, an image block can be made up of any number of horizontal and vertical pixels, for example, 8 × 8 pixels or 12x8 pixels. Besides the possibility that a picture block is square or rectangular shaped is it also possible that an image block consists of an arbitrarily bordered object. According to the inventive method For example, a pixel may have different forms of presentation, such as for example, a pixel is a red component R of an RGB color point, as well as a black and white color value equivalent. additionally The color depth can be set according to the desired color resolution such as 8-bit or 12-bit. Besides the possibility a full search (English: "Full Search Method ") to determine the prediction a variety of other search strategies are known. This is to the reference Shi & Sun, "Image and Video Compression for Multimedia Engineering ", CRC Press, 2000, Chapter 11.

In 2 For example, the optimal prediction is searched only at every possible position of a pixel within the search area SB of the reference image RB. In order to increase the search accuracy and thus to improve the prediction, the first resolution A1 within the search range SB is increased according to the present invention. This is exemplified by 3 explained in more detail by way of example.

In 3 is shown on the right side of the current image block BBN with 2 × 2 pixel elements. To clarify the numbering of each individual pixel, an XY coordinate system is drawn for this purpose. For example, a current pixel BPN located at the top right of the current image block BBN has the coordinates (2,1). On the left half of 3 only the search area SB of the reference picture RB is off 2 displayed. The pixels with the circles correspond exactly to the pixels of the search area SB 2 , The additional pixels which are added to increase the accuracy of the prediction are designated below with the symbol "X." The accuracy of the prediction is specified with the aid of a first resolution A1 For the easier identification of individual pixels, the search range SB is an associated one For example, a first pixel BPV1 located at the second position in the first row has the coordinates (2,1) and a second pixel BPV2 which is at the third position in the second row. the coordinates (3,2).

The additional pixels marked "X" are generated, for example, by means of a linear interpolation, so that the first and second pixels BPV1 and BPV2 can be calculated as follows: BPV1 = 2/3 · BPV3 + 1/3 · BPV4; BPV2 = 2/3 · BPV7 + 1/3 · BPV8 = 2/9 · BPV3 + 4/9 · BPV4 + 1/9 · BPV5 + 2/9 · BPV6; with BPV7 = 1/3 · BPV3 + 2/3 · BPV4;
BPV8 = 1/3 · BPV5 + 2/3 · BPV6;

In contrast to 2 Now includes a search area block 4x4 pixels. For example, the first search area block SBB1 each includes the first four pixels of the first four lines of the search area SB. In order to determine the error measure, only those pixels of the search area block which are located at the corners of the respective search area block are used in the present example. For example, for the first search area block SBB1, these are the pixels with the coordinates (1,1), (1,4), (4,1), (4,4). After the error measure for this first search area block SBB1 has been formed, the position of the next search area block is determined by shifting the search area block by one pixel. In the following, the error measure for all search area blocks possible in the search area is determined with the original image block. For example, an error measure for a sixth search area block SBB6 is formed with the current image block BBN. After the respective error measure has been formed for all possible search area blocks within the search area SB, that search area block which generates the optimum prediction for the current image block BBN of the original image OB is selected. For example, the optimal search area block SBBX gives the optimal prediction for the current image block BBN of the original image OB. For the optimum search area block SBBX found, an image vector BV is specified which consists of a horizontal and vertical subvector BVH or BVV. To 3 the image vector BV indicates the following values: BV (BVH, BVV) = (2, 2);

in this connection is taken into account, that the image vector BV the horizontal and vertical sub-vector BVH or BVV in the elevated first resolution A1 indicates.

Since the first resolution A1 in the example after 3 was increased, a prediction can be found in which the error measure is equal to or less than the determined error Lermaß from the example according to 2 in which the first resolution A1 was not increased. In order to achieve an optimal prediction, the first resolution A1 is set individually according to the method according to the invention. Thus, no fixed first resolution A1 is specified, but these are individually adapted depending on the requirements of the accuracy of the prediction. In addition, the method according to the invention makes it possible to set different resolutions horizontally and vertically. This is illustrated by means of a horizontal and vertical resolution component AH or AV. For example, a possible first search area block SBB1 in the horizontal consists of 6 pixels and in the vertical of 5 pixels. By the individual setting of the first resolution A1, the prediction can be adapted adaptively to the present properties of the original image OB and of the reference image RB. For example, the first resolution A1 in the horizontal direction may be increased until the corresponding error measure no longer indicates an improvement in the prediction.

at the increase the first resolution However, in practice, A1 notes that it may be possible to increase by increasing the first resolution A1 a data volume required for signaling the image vector BV increases. So it may be useful in practice, in optimizing the prediction in addition to the minimization of the error measure, that also by the chosen first resolution A1 needed Data volume for signaling the respective image vector BV consider.

In the following, possible forms of representation for the first resolution A1 and the associated resolution value AW, with which the used first resolution A1 is signaled, are explained in more detail. In the event that both in the horizontal and in the vertical direction, the identical first resolution solution A1 is used, only a value for specifying the first resolution A1 is needed. As in 3 in the first search area block SBB1, additional pixels marked by the symbol "X" are inserted between the existing image points indicated by circles The first resolution A1 of the first search area block SBB1 is inserted by inserting two each Thus, between the original pixels, two new pixels each are inserted, and a possible representation of the first resolution A1 or the resolution value AW indicates how many additional columns or rows have been inserted to achieve the increased first resolution A1. In the present example, the first resolution A1 or the resolution value AW is assigned the value "2", which indicates that in each case two columns and rows are inserted between the existing pixels in order to increase the first resolution A1. Thus, the resolution value AW can be given as follows: AW = "2";

In the case, that different resolutions are selected in the horizontal and vertical directions, it is necessary that the resolution value AW both the horizontal and vertical resolution components AH and AV indicates. For example, in a first search area block, SBB1 becomes three Columns and five Inserted rows. A possible resolution value AW or the resolution Thus, the duple comprises three and five or AW = "3.5".

A For example, alternative representation indicates how many pixels after the increase the first resolution A1 comprises a search area block. For example, a first one includes Search area block SBB1 5 pixels in the horizontal plane and 7 pixels in the vertical plane. Thus, e.g. a resolution value AW or the first resolution A1 the duple five and seven or AW = "5.7".

According to the inventive method can a first resolution A1, in addition for several image blocks, like beispielswei se the first and second image block BB1 and BB2, signaled and used. This is useful in practice because therefore only a first resolution A1 must be specified or transmit a resolution value AW for several image blocks and thus the needed Data volume is reduced. In a further advantageous embodiment the method according to the invention becomes the first resolution A1 or the associated resolution value Keep AW until a new resolution value A is signaled.

• a) Für einen benachbarten Bildblock des aktuellen Bildblocks BBN, zum Beispiel dem zweiten Bildblock BB2, wurde die dazugehörige zweite Auflösung A2 bereits ermittelt. Beispielsweise wird für die erste Auflösung A1 die gleiche Auflösung verwendet, wie für die zweite Auflösung A2. Wird beispielsweise für die zweite Auflösung A2 = „7, 4" verwendet, so wird damit auch für die erste Auflösung A1 = „7, 4" benutzt.
• b) Bei einem Bildcodierverfahren wird der Bildblock BBN einer Quantisierung mit einem Quantisierungsparameter QP unterzogen. Beispielsweise kann bei einer hohen Quantisierung der Unterschied zwischen den benachbarten Bildpunkten sehr gering sein, so dass unter Umständen bereits mit einer geringen Auflösung eine gute adaptive Prädiktion erreicht werden kann. So kann auch für jeden möglichen Quantisierungsparameter QP jeweils eine eindeutig zuordenbare erste Auflösung A1 festgelegt sein. Beispielsweise wird für den Quantisierungsparameter QP = „15" die erste Auflösung A1 = „15" benutzt.
• c) Weist ein Bildblock eine bestimmte Blockstruktur BK auf, so kann daraus abgeleitet werden, dass eine bestimmte erste Auflösung A1 eine sehr gute Prädiktion ergibt. Umfasst beispielsweise sowohl der aktuelle Bildblock BBN des Originalbildes OB als auch der erste Suchbereichsblock SBB1 des Referenzbildes RB lediglich schwarze Bildpunkte, so kann damit ein Fehlermaß gefunden werden, das den Wert Null annimmt. Somit kann bereits ohne eine Erhöhung der Auflösung eine optimale Prädiktion gefunden werden. Es kann auch für vorgebbare Blockstrukturen BK jeweils eine eindeutig zuordenbare erste Auflösung A1, festgelegt werden.
• d) Weiterhin kann die gewählte erste Auflösung A1 des aktuellen Bildblocks BBN in Abhängigkeit vom gewählten Bildformat BF des zu bearbeitenden Originalbildes OB ermittelt werden. Beispielsweise umfasst das Originalbild OB 1024 × 1024 Bildpunkte und soll auf einem großflächigen Ausgabemedium DD wiedergegeben werden. Hierbei kann es in der Praxis vorteilhaft sein, durch die Wahl einer hohen Auflösung eine sehr gute Prädiktion zu erreichen. Dadurch kann erreicht werden, dass ein Kompressionsverfahren für eine Sequenz von Bildern trotz hoher Kompression eine sehr gute Bildqualität nach der Decodierung der komprimierten ersten Videosignals BS1 durch die Decodiervorrichtung DV ermöglicht. Es kann auch für jeweils eine oder mehrere Bildformate BF eine dazugehörige erste Auflösung A1 eindeutig festgelegt sein. Möglicherweise wird für das Bildformat BF = „1024 × 1024" die erste Auflösung A1 = „12" und für das Bildformat BF = „ 176 × 144" die erste Auflösung A1 = „16, 12" gewählt.
• e) Des Weiteren kann die erste Auflösung A1 der Prädiktion des aktuellen Bildblocks BBN von einem zeitlichen Abstand des Originalbilds OB zu dem für die Prädiktion verwendeten Referenzbild RB ermittelt werden. Beispielweise liegen zwischen dem Originalbild OB und dem Referenzbild RB eine Sekunde Abstand. Aufgrund des großen zeitlichen Unterschieds wird eine geringere erste Auflösung A1 zugelassen, als bei einem Referenzbild, welches nur eine Viertelsekunde Abstand aufweist. Es kann auch die erste Auflösung A1 anhand des zeitlichen Abstandes zwischen Originalbild OB und Referenzbild RB eindeutig festgelegt werden, wie z.B. mit einer Tabelle, die bestimmten zeitlichen Abständen eine vordefinierte erste Auflösung A1 zuweist.
• f) Schließlich kann die erste Auflösung A1 der Prädiktion des aktuellen Bildblocks BBN in Abhängigkeit von der Länge des zugehörigen Bewegungsvektor VB ermittelt werden. Wird bei der Ermittlung einer optimalen Prädiktion in einer hohen ersten Auflösung A1 ein langer Bewegungsvektor gefunden, so kann es in der Praxis zweckmäßig sein, anstelle des langen Bewegungsvektors BV einen kürzeren Bewegungsvektor BV in einer niedrigeren ersten Auflösung A1 zu übermitteln. Hierdurch wird unter Umständen das Fehlermaß ansteigen, jedoch wird das zur Übermittlung des Bewegungsvektors BV benötigte Datenvolumen reduziert. Möglicherweise wird an Hand einer Tabelle langen Bewegungsvektoren BV mit einer hohen ersten Auflösung A1 in kürzere Bewegungsvektoren BV mit einer geringeren ersten Auflösung A1 überführt. So werden beispielsweise die hohe erste Auflösung A1 = „12" und der lange Bewegungsvektor BV = „25, 38" zur ersten Auflösung A1 = „4" und zum kurzen Bewegungsvektor BV = „8, 13".

In a possible extension of the method according to the invention, the first resolution A1 of the current image block BBN can be determined as a function of one of the following parameters:

• a) For an adjacent image block of the current image block BBN, for example the second image block BB2, the associated second resolution A2 has already been determined. For example, the same resolution is used for the first resolution A1 as for the second resolution A2. If, for example, A2 = "7, 4" is used for the second resolution, then A1 = "7, 4" is also used for the first resolution.
• b) In an image coding method, the image block BBN is subjected to quantization with a quantization parameter QP. For example, with a high quantization, the difference between the adjacent pixels may be very small, so that a good adaptive prediction may already be achieved with a low resolution under certain circumstances. Thus, a clearly assignable first resolution A1 can also be defined for each possible quantization parameter QP. For example, the first resolution A1 = "15" is used for the quantization parameter QP = "15".
• c) If an image block has a specific block structure BK, then it can be deduced that a particular first resolution A1 results in a very good prediction. If, for example, both the current image block BBN of the original image OB and the first search region block SBB1 of the reference image RB comprise only black pixels, then an error measure which assumes the value zero can be found. Thus, an optimal prediction can already be found without increasing the resolution. It is also possible to specify a uniquely assignable first resolution A1 for predefinable block structures BK.
• d) Furthermore, the selected first resolution A1 of the current image block BBN can be determined as a function of the selected image format BF of the original image OB to be processed. For example, the original image OB comprises 1024 × 1024 pixels and is to be reproduced on a large-area output medium DD. It may be advantageous in practice to achieve a very good prediction by choosing a high resolution. It can thereby be achieved that a compression method for a sequence of images despite high compression enables a very good image quality after the decoding of the compressed first video signal BS1 by the decoding device DV. It can also be uniquely determined for each one or more image formats BF associated first resolution A1. Possibly the first resolution A1 = "12" is selected for the image format BF = "1024 × 1024" and the first resolution A1 = "16, 12" for the image format BF = "176 × 144".
• e) Furthermore, the first resolution A1 of the Prediction of the current image block BBN from a time interval of the original image OB to the reference image RB used for the prediction. For example, there is one second between the original image OB and the reference image RB. Due to the large difference in time, a lower first resolution A1 is permitted than with a reference image which is only a quarter of a second apart. The first resolution A1 can also be unambiguously determined on the basis of the time interval between the original image OB and the reference image RB, for example with a table which assigns a predefined first resolution A1 to specific time intervals.
• f) Finally, the first resolution A1 of the prediction of the current image block BBN can be determined as a function of the length of the associated motion vector VB. If a long motion vector is found in the determination of an optimal prediction in a high first resolution A1, it may be expedient in practice to transmit a shorter motion vector BV in a lower first resolution A1 instead of the long motion vector BV. As a result, under certain circumstances, the Fehleraßaß increase, but the data volume required to transmit the motion vector BV is reduced. Possibly, using a table of long motion vectors BV with a high first resolution A1, it is possible to convert into shorter motion vectors BV with a lower first resolution A1. For example, the high first resolution A1 = "12" and the long motion vector BV = "25, 38" become the first resolution A1 = "4" and the short motion vector BV = "8, 13".

In a possible Extension of the method according to the invention can the first resolution A1 the current image block BBN also with the help of several of the above Parameters are determined. Since with the mentioned parameters the first resolution A1 in the method for encoding as well as for decoding the prediction method can be determined implicitly, is a transmission of the first determined resolution A1 by means of a resolution value AW unnecessary. However, for the case by an implicitly determined first resolution A1 through another first resolution A1 is to be replaced, the other first resolution A1 using the resolution value AW be communicated.

In a further alternative execution step the method according to the invention becomes a needed one Data volume for signaling at least one resolution value AW reduced. For example, with the aid of the Huffman coding method those resolution values AW, which occur frequently a shorter one Codeword, and those resolution values AW, which rarely occur, a longer one Codeword to be assigned.

According to the present Invention can several resolution values AW, AW1, AW2 are differentially coded. For example, for a second resolution value AW2 formed only the difference to a first resolution value AW1 and signaled.

According to the method of the invention, the first resolution A1 is signaled with the aid of the resolution value AW. As in 1 At least one resolution value AW can be transmitted in the first compressed image signal BS1 or separately to a second compressed image signal BS2. For example, in the first compressed image signal BS1, the resolution value AW is inserted before the signaling of the respective image vector BV into a data stream DS to be transmitted. For this purpose, as in 4 In the data stream DS of the first compressed image signal BS1, the resolution value AW comprising the horizontal and vertical resolution components AH and AV, respectively, followed by the motion vector BV, possibly including a horizontal and vertical motion vector BVH and BW, respectively, is inserted.

alternative can be at least one resolution value AW separately transmitted to the second compressed image signal BS2 become. This may be about a unique identifier, that in the resolution value AW, a referencing of the resolution value AW for a given Image block within the original image OB enabled. For example, includes this unique identifier is a number that contains a Image block within the image, such as a current image block BBN of an original image OB, clearly identified.

To Receiving the first or second compressed image signal BS1 or BS2 by the decoder DV, is processed by the processing unit VD first a reconstructed image block of the current image block BBN based the first resolution A1 and its associated Image vector BV determined. possibly will also be a resolution value AW receive and from this the corresponding first resolution A1 to reconstruct the current image block. After generating all reconstructed image blocks a reconstructed image can be output on the output medium DD. Depending on The image encoding method used may be the reconstructed image over the Original image deviations, such as those in the professional world known artifacts have.

In addition to the possibility to search the prediction within a reference image RB, the Prediction can also be determined in a decoded successor image. The successor image is an image which was taken at a later time than the original image OB.

Claims (18)

prediction, in which - one prediction a current image block (BBN) of an original image (OB) in one Image coding method with a first resolution (A1) is encoded, - in which the first resolution (A1) the prediction is set such that the first resolution (A1) in any accuracy customizable to the current image block (BBN), - the set first resolution (A1) is signaled. Method according to claim 1, characterized in that that by a horizontal and a vertical resolution component (AH or AV) the first resolution (A1) set differently in the horizontal and vertical directions becomes. Method according to one of the preceding claims, characterized in that the first resolution (A1) for a plurality of image blocks (BB1, BB2, BBN) is used. Method according to one of the preceding claims, characterized characterized in that the first resolution (A1) of the prediction of the current image block (BBN) as a function of a second resolution (A2) the prediction an adjacent second image block (BB2) is determined. Method according to one of the preceding claims, characterized characterized in that the first resolution (A1) of the prediction of the current image block (BBN) as a function of a quantization parameter (QP) of the current image block (BBN) is determined. Method according to one of the preceding claims, characterized characterized in that the first resolution (A1) of the prediction of the current image block (BBN) as a function of a block structure (BK) of the current image block (BBN) is determined. Method according to one of the preceding claims, characterized characterized in that the first resolution (A1) of the prediction of the current image block (BBN) depending on an image format (BF) of the original image (OB) is determined. Method according to one of the preceding claims, characterized characterized in that the first resolution (A1) of the prediction of the current image block (BBN) as a function of a time interval of the original image (OB) to a reference image used for the prediction (RB) is determined. Method according to one of the preceding claims, characterized characterized in that the first resolution (A1) of the prediction the current image block (BBN) as a function of a length of an associated motion vector (BV) is determined. Method according to one of the preceding claims, characterized characterized in that the first resolution (A1) by means of a resolution value (AW) in a first compressed image signal (BS1) becomes. Method according to one of the preceding claims, characterized characterized in that the first resolution (A1) by means of a resolution value (AW) transmitted separately to a second compressed image signal (BS2) becomes. Method according to one of the preceding claims, characterized characterized in that the resolution value (AW) are compressed data. Method according to one of the preceding claims, characterized characterized in that at least one resolution value (AW) differentially is encoded. Method for decoding a prediction method, characterized in that in this case the prediction method according to a the claims 1 to 13 is decodable. Encoding device (EV) for performing a prediction method in particular according to one of claims 1 to 13, with a processing unit (VE) for encoding a prediction a current image block (BBN) of an original image (OB) in one Image coding method with a first resolution (A1), wherein the first resolution (A1) the prediction is set such that the first resolution (A1) in any accuracy is customizable to the current image block (BBN) and signaling the set first resolution (A1). Encoding device (EV) according to claim 15, having a Sending unit (SE) for transmitting the first resolution (A1) by means of a resolution value (AW). Decoding device (DV) for decoding a prediction method, characterized in that in this case a decoding of a prediction method according to claim 14 feasible is. Decoding device (DV) according to claim 17, comprising a receiving unit (ED) for receiving a resolution value (AW), whereby the first resolution (A1) is represented with the aid of the resolution value (AW).