Classifications

• • H—ELECTRICITY
  • H03—ELECTRONIC CIRCUITRY
  • H03M—CODING; DECODING; CODE CONVERSION IN GENERAL
  • H03M7/00—Conversion of a code where information is represented by a given sequence or number of digits to a code where the same, similar or subset of information is represented by a different sequence or number of digits
  • H03M7/30—Compression; Expansion; Suppression of unnecessary data, e.g. redundancy reduction
  • H03M7/3002—Conversion to or from differential modulation
• • H—ELECTRICITY
  • H04—ELECTRIC COMMUNICATION TECHNIQUE
  • H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
  • H04N19/00—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals
  • H04N19/50—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using predictive coding
  • H04N19/503—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using predictive coding involving temporal prediction
  • H04N19/51—Motion estimation or motion compensation
  • H04N19/573—Motion compensation with multiple frame prediction using two or more reference frames in a given prediction direction

Definitions

• the invention relates to a method for coding a sequence of digitized images comprising a plurality of pixels with associated pixel values, and a corresponding decoding method and a corresponding encoder and decoder.
• Digitized sequences of images occur, inter alia, in medical applications in which, for example, a sequence of two-dimensional sectional images of organs of the x-rayed patient are generated in X-ray images by means of a computer tomograph, these sectional images being recorded at a certain spatial distance from one another.
• Various compression methods for the lossless compression of such image data are known from the prior art, and these methods are also used for any other digitized images.
• the standards JPEG-LS and JPEG-2000 are used for compression, in which the individual images of the image sequence are compressed independently of each other. In this case, an existing correlation between the individual successive images is not utilized.
• video coding standards which are used to compress moving image content, such as video. In the H.264 / AVC coding standard, image blocks of successive images are predicted by means of motion compensation. The resulting prediction error is then coded.
• the object of the invention is to provide a method and a device for coding or decoding images, which allow good compression rates for images of any image content.
• a sequence of digitized images comprising a multiplicity of pixels with associated pixel values is coded, wherein pixels of the images to be coded are predicted and the prediction errors resulting from the prediction are coded.
• the prediction of a pixel to be coded is carried out for at least a part of the pixels by means of trajectories along coded pixels.
• the prediction based on trajectories is preferably always carried out whenever possible, i. E. if there are enough already coded pixels.
• a plurality of trajectories is first of all determined, the trajectories being respectively from the pixel to be coded and further pixels from the image of the pixel to be coded and / or from one or more temporally to the image of the pixel to be coded run the pixel adjacent images.
• an evaluation measure is determined which is configured in such a way that the smaller the fluctuations of the uncoded pixel values of the further pixels along the trajectory are, the higher the evaluation of a trajectory according to the evaluation criterion.
• a prediction value for the pixel value of the pixel to be coded is then determined.
• the method according to the invention is based on the knowledge that pixels with similar pixel values often continue in a predefined direction, so that a particularly good prediction is achieved with the aid of trajectories whose pixels have similar pixel values. As a result, a high coding efficiency, in particular for lossless or almost lossless coding, can be achieved.
• the evaluation measure here depends on the sum of the absolute deviations or the quadratic deviations between the pixel values of the further pixels of the trajectory and the mean value of these pixel values, with a trajectory being rated the higher the smaller the absolute value deviations or quadratic deviations are.
• the evaluation measure also depends on the distances between the further pixels of the trajectory, with a trajectory being rated the higher, the larger these distances are. This achieves a suitable normalization of the evaluation measure.
• the score may depend on the distance between the pixel to be coded and the further pixel closest to the pixel to be coded, the smaller the distance the higher the trajectory.
• a reliability factor flows in the evaluation of the trajectory It can be assumed that the closer the further pixels of the trajectory are to the pixel to be coded, the more reliable is the statement that a pixel to be coded belongs to a structure with similar pixel values.
• trajectories with the same length and / or trajectories with different lengths can be taken into account. All that is decisive is that a trajectory comprises at least two further, already coded pixels.
• the permissible length of the trajectories used in the prediction is increased by one pixel each time the image is coded. This takes into account that the longer the number of images already encoded, the longer trajectories can be formed from pixels of different images. If appropriate, the permissible length of the trajectories can also be fixedly selected at least for a predetermined number of successive images.
• the prediction value for the pixel value of the image point to be coded is determined by means of an extrapolation. Any extrapolation methods can be used, in particular a linear extrapolation and / or a spline extrapolation and / or a polynomial extrapolation.
• the mean value of the pixel values of the further pixels can be determined as the prediction value for the pixel value of the pixel to be coded.
• bi-points are predicted by means of trajectories comprising pixels from the same image and / or by means of trajectories comprising pixels from different images.
• at least for the first two images are predicted by means of trajectories from the same image and / or coded based on a different coding from the inventive method.
• the deviating coding is used in an image if there are not yet sufficient number of pixels for prediction based on trajectories.
• the prediction errors resulting from the prediction are at least partially represented by the deviation between the pixel value of the pixel to be coded and the predicted pixel value.
• the deviation is signed in order to correct the decoded pixel correctly by means of the deviation in a later decoding.
• a loss of information in the coded pictures can be accepted.
• a threshold value is preferably specified, wherein the prediction errors resulting from the prediction when the threshold value is exceeded by the deviation between the pixel value of the pixel to be coded and the predicted pixel value are represented and are otherwise set to the value zero.
• an error bound is taken into account, with all prediction errors less than or equal to the error bound being set to zero, thereby increasing the coding efficiency.
• the pixel value of a pixel with a prediction error of zero is replaced by the predicted pixel value, thereby avoiding a drift between the encoding and the decoding of the images.
• the actual prediction error can be coded lossless as well as lossy in the method according to the invention.
• methods known from the prior art are used for coding the prediction error, which in particular involves a transformation (preferably a DCT transformation) and / or a quantization and / or include entropy coding.
• Quantization implies information loss, whereas entropy coding is lossless.
• the pixel value of the pixel for which the prediction error was determined is preferably replaced by the predicted pixel value corrected with the decoded prediction error. This, in turn, ensures that there are no drifts between the encoding and the decoding of the images.
• a particularly preferred field of application of the invention is the coding of digitized images in the form of medical images, in particular medical radiographs.
• structures with the same or similar pixel values continue in those directions in which the captured object (e.g., the organ of a patient) is or is a boundary of the organ to its environment.
• the invention further comprises a method for decoding a sequence of digitized pictures which are coded with the coding method according to the invention.
• the prediction errors of the respective pixels are decoded and the respective pixels to be decoded are predicted, whereby a decoding error occurs.
• the pixel results from a correction of a predicted pixel with the decoded prediction error.
• the prediction of a pixel to be decoded is again based on trajectories.
• a plurality of trajectories is determined, wherein the trajectories each extend through the pixel to be decoded and further, already decoded pixels from the image of the pixel to be decoded and / or from one or more temporally adjacent to the image of the pixel to be decoded Border ,
• an evaluation measure is determined, which is designed in such a way that the smaller the fluctuations of the decoded pixel values of the further pixels along the trajectory are, the higher the evaluation of a trajectory according to the evaluation criterion.
• a prediction value for the pixel value of the pixel to be decoded is then determined.
• the invention further comprises a method for transmitting a sequence of digitized pictures in which the pictures of the sequence are coded using the coding method according to the invention, transmitted over a transmission link and subsequently decoded using the decoding method according to the invention.
• the invention furthermore relates to an encoder for coding a sequence of digitized images, which comprises a prediction means for predicting pixels of the images to be coded and an encoding means for coding the prediction errors resulting from the prediction.
• the prediction means is designed in such a way that the prediction according to the invention described above can be carried out by means of trajectories with the prediction means.
• the invention further relates to a corresponding decoder for decoding a sequence of signals coded pictures, wherein the decoder comprises a decoding means for decoding the prediction errors of the respective pixels as well as a prediction means for predicting the pixels to be decoded, wherein the decoded dot results from a correction of a predicted pixel with the decoded prediction error.
• the prediction means is designed in such a way that the prediction takes place by means of trajectories based on the decoding method according to the invention.
• the invention further relates to a system comprising an encoder according to the invention and a decoder according to the invention.
• Fig. 1 is a schematic representation of the coding of
• FIG. 2 is a diagram which illustrates the calculation of a prediction value for a pixel to be coded according to an embodiment of the method according to the invention.
• FIG. 3 shows a schematic representation of a system of encoder and decoder for coding and decoding a digitized image sequence according to an embodiment of the method according to the invention.
• the principle of the coding according to the invention based on a section of a sequence of digitized images comprising three temporally successive images II, 12 and 13 will be described below with reference to FIG.
• the sequence of images may be, for example, act of medical imaging systems.
• the individual images can in particular represent sectional images of organs of a patient recorded using a computer tomograph.
• Such systems generate very large amounts of data which have to be stored for lossless or virtually lossless archiving.
• Each individual image II, 12 or 13 of the digitized image sequence according to FIG. 1 contains a multiplicity of pixels in the form of pixels, each pixel being associated with a pixel value which has a brightness value in black-and-white images and color images in color images Normally a lumi- nance value.
• the individual pixels of the images are represented by dots in FIG. 1, wherein a distinction is made between three different types of pixels.
• white pixels to be coded pixels P are reproduced, by black and hatched pixels P 'and P "already coded pixels.
• the hatched pixels P ' serve in this case for the formation of trajectories for the prediction of the pixel to be coded in the third row and third column of the image 13.
• FIG. 1 Each individual image II, 12 or 13 of the digitized image sequence according to FIG. 1 contains a multiplicity of pixels in the form of pixels, each pixel being associated with a pixel value which has a brightness value in black-and-white images and color images in color images Normally
• FIG. 1 also shows a coordinate system with x, y and z direction.
• the x-y plane represents the two-dimensional extent of the corresponding images, and the z-direction reproduces the temporal sequence of the images, wherein the image II was taken in front of the image 12 and the image 12 in front of the image 13.
• the coding of the pixel in the third column and third row of the image 13 is shown.
• trajectories are determined starting from the pixel to be coded by at least two further, already coded pixels in the same image 13 or in the adjacent images II and 12.
• the trajectories which pass through pixels in the same image 13 are denoted by T 'in FIG.
• the trajectories, which run through pixels in the temporally preceding images II and 12 are denoted by the reference symbol T.
• T For reasons of clarity, again only a few trajectories are provided with the corresponding reference symbols T and T '.
• a suitable prediction of the pixel to be coded is made possible with the coding according to the invention. The individual steps for prediction run as explained below.
• n of pixels is selected, which are selected per trajectory.
• the value n of the number of pixels varies in the course of the method or is chosen differently for trajectories with pixels within the same image than for trajectories with pixels from different images. If necessary, the number of permissible pixels per trajectory can always be increased by one after each encoding of a new image.
• the trajectories running through already coded pixels are determined.
• trajectories are used both in the image 13 and in the two preceding images II and 12.
• Fig. 1 there are four trajectories T 'with pixels from the same image 13 and nine trajectories T, each with a pixel from the image Il and the image 12. It is determined an error measure f ⁇ now for each of the trajectories which the deviation of the Pixel values of the pixels P 'of the respective trajectory represented by the average value of these pixel values.
• the means that the error measure is defined in the embodiment described here as follows:
• the above error measure corresponds to a specific variant of a rating measure according to claim 1, wherein a lower error measure corresponds to a higher rating.
• the error measure can also be defined differently if necessary, the only decisive factor is that the error measure represents a measure of the fluctuations in the pixel values along the trajectory. For example, instead of the deviation in the form of the absolute difference between the pixel value and the mean value, the quadratic deviation and thus the variance can be used to calculate the error measure.
• error measure can also be defined such that it is normalized by the distance d of the pixels on the trajectory.
• a corresponding error measure f 2 then reads as follows:
• ⁇ x, Ay and ⁇ z are the distances of two pixels on the trajectory in the x, y and z directions. In the scenario of FIG. 1, these distances are constant along a trajectory.
• the above equation may also be be modified so that changing distances between the individual pixels are taken into account.
• the trajectory with the lowest error measure i. with the highest rating selected. This takes into account the fact that, especially in medical images, structures with an identical or similar brightness value continue in one direction.
• the calculation of the predictor with linear extrapolation is shown in FIG.
• the values Pl and P2 denote the positions of the already coded pixels along the trajectory used for the prediction and the point P3 is the position of the bi-dot which is to be predicted.
• the corresponding pixel values of the pixels are plotted, for example as corresponding brightness values.
• the pixel value of the pixel at the position Pl is denoted by W 1 and the pixel value of the pixel at the position P2 by W 2 .
• the prediction value can also be determined in another way, for example by spline interpolation or polynomial extrapolation. Likewise, if appropriate, the mean value of the pixels can also be used as a prediction value.
• the selection of the trajectory can be modified by a reliability factor g (a), which is dependent on a distance a of the pixel to be predicted from the nearest pixel on the trajectory.
• This modified distance measure / ⁇ can for example be as follows:
• trajectories comprising pixels with a small distance to the predicted pixel are preferably selected.
• the function g (a) is monotone increasing in this case.
• the reliability factor takes into account that the closer the pixels of a trajectory are to the predicted pixel, the higher the probability of a correct prediction.
• an error signal (also called a prediction error) is calculated, which corresponds to the difference between the predictor and the original pixel value. In the case of lossless coding, this error signal is coded.
• an error barrier it is also possible for an error barrier to be specified for coding, whereby the originally determined error signal is only coded if the magnitude difference of the error signal does not lie within the error barrier. Should the difference be within the error barrier, a prediction error of zero is transmitted.
• the error barrier can also be varied, wherein the error barrier is set to zero in the case of lossless coding.
• Pixel value of a predicted pixel is replaced by the pixel value of the predicted pixel if the difference is within the error bound. This ensures that encoders and decoders work on the same data base, avoiding drift.
• this error is coded.
• This coding can be done in any way with methods known in the art.
• a transformation preferably a DCT transformation
• quantization and / or entropy coding take place.
• the steps of quantization and entropy coding are well known in the field of video coding. The quantization leads to a loss, whereas the entropy coding results in lossless compression.
• the images of the image sequence are to be coded loss-free, only the entropy coding and not the quantization is used in the coding of the prediction error. If the prediction error is changed by the encoding (i.e., in particular by quantization), it is convenient to change the original pixel value of the coded pixel by the
• the coded error is transmitted to a decoder.
• decoding is implemented in an analogous manner based on the determination of trajectories described above.
• the decoder decodes prediction errors and performs a prediction according to the invention based on trajectories, the individual pixel values.
• the predicted pixels are corrected with the decoded prediction errors, thereby obtaining the original pixel value. In the case of loss-free coding, exactly the original pixel value is then also obtained.
• the coding of the first two images does not necessarily take place on three-dimensional trajectories from different images, since initially no coded pixels from different images for forming the trajectories are present. Instead, a two-dimensional direction prediction of a pixel based on pixels of the same image is used. Furthermore, a completely different coding is used for the first pixels of the first or second image to be coded (for example, an intra-coding without prediction). tion), since initially no two-dimensional trajectories can be formed in the same image. A direction prediction based on three dimensions, ie taking into account the z direction, takes place only when the third image is coded after the method has been initialized.
• only two-dimensional trajectories within the same image can be used in the third image.
• the three-dimensional prediction can now also be continued for the further images to be coded, it being possible for the maximum length of the three-dimensional trajectories to be increased by one with each additional image.
• the inventive method has a number of advantages. By selecting a preferred prediction direction per pixel based on corresponding trajectories, a good prediction can be achieved in particular for structures in which the same brightness value of an object continues in one direction, as is the case in particular with medical image data. Thus, a higher coding efficiency can be achieved, in particular for lossless or almost lossless coding.
• the prediction direction is determined according to the invention using already coded pixels. This determination takes place in the decoder in the same way as in the encoder, so that no additional prediction information from the encoder to the decoder must be transmitted. This additionally increases the coding efficiency. Shown schematically in FIG. 3 is a system for coding and decoding an image sequence based on an embodiment of the method according to the invention.
• the encoder for encoding the image sequence is denoted by reference numeral 1 and the encoder supplied image data stream comprising images II, 12 and 13 is first fed to a prediction means 2, which based on the inventive method by means of trajectories a prediction value and a corresponding prediction error between Predictive onswert and original pixel value of the individual pixels determined.
• the prediction errors are then supplied to a coding means 3 which codes the prediction error based on correspondingly known coding methods, such as quantization or entropy coding.
• the coded prediction error is finally transmitted to a decoder 4 via a transmission path indicated by the arrow P.
• the transmission path can be wired or wireless.
• the received prediction error is decoded into a corresponding decoding means 5.
• the prediction of the pixel values is carried out in a prediction means 6, the prediction proceeding analogously to the prediction means 2 based on trajectories.
• the predicted pixels are then corrected with the decoded prediction error, so that in the case of loss-free coding the original pixel value and in the case of lossy coding an approximated pixel value is obtained.

Landscapes

• Engineering & Computer Science (AREA)
• Multimedia (AREA)
• Signal Processing (AREA)
• Theoretical Computer Science (AREA)
• Compression, Expansion, Code Conversion, And Decoders (AREA)
• Compression Or Coding Systems Of Tv Signals (AREA)
• Compression Of Band Width Or Redundancy In Fax (AREA)

Applications Claiming Priority (3)

Publications (1)

Family

ID=41228720

Family Applications (1)

Country Status (7)

Families Citing this family (21)

Family Cites Families (9)

• 2008
  • 2008-11-21 DE DE102008058489A patent/DE102008058489A1/de not_active Ceased
• 2009
  • 2009-08-14 CN CN200980132218.XA patent/CN102124495B/zh active Active
  • 2009-08-14 WO PCT/EP2009/060570 patent/WO2010020592A1/de active Application Filing
  • 2009-08-14 EP EP09781871A patent/EP2313864A1/de not_active Withdrawn
  • 2009-08-14 JP JP2011523399A patent/JP5284471B2/ja active Active
  • 2009-08-14 US US12/737,814 patent/US8472739B2/en active Active
  • 2009-08-14 KR KR1020117006482A patent/KR101618080B1/ko active IP Right Grant

Non-Patent Citations (1)

Also Published As

Similar Documents

Legal Events