WO1992017981A1 - Verfahren zur kompression von bilddaten - Google Patents
Verfahren zur kompression von bilddaten Download PDFInfo
- Publication number
- WO1992017981A1 WO1992017981A1 PCT/DE1992/000199 DE9200199W WO9217981A1 WO 1992017981 A1 WO1992017981 A1 WO 1992017981A1 DE 9200199 W DE9200199 W DE 9200199W WO 9217981 A1 WO9217981 A1 WO 9217981A1
- Authority
- WO
- WIPO (PCT)
- Prior art keywords
- ifs
- image
- stochastic
- fitness
- invariant
- Prior art date
Links
Classifications
-
- 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
- H04N19/90—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using coding techniques not provided for in groups H04N19/10-H04N19/85, e.g. fractals
- H04N19/99—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using coding techniques not provided for in groups H04N19/10-H04N19/85, e.g. fractals involving fractal coding
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T9/00—Image coding
- G06T9/001—Model-based coding, e.g. wire frame
Definitions
- the invention relates to a novel method for reducing the memory requirement when storing image data in electronic data processing systems or for reducing the data transmission rate when " transmitting image data in telecommunication systems or other data transmission devices.
- Such methods have a high and still strongly growing economic importance also with regard to corresponding services in connection with digital message transmission in public communication networks (ISDN), or in connection with the mass storage of image data, for example in medical technology and in other areas of applied research.
- ISDN public communication networks
- LZW Lempel-Ziv-Welch
- compression factors of generally not more than 20 (for still images) to 200 (for example for the coding of moving picture sequences for transmission via 64 kbit / s lines on video phones).
- the invention is based on the object of specifying a method for compressing image data, with the aid of which the automatic compression of image data is possible even without the involvement of a human surgeon, while achieving extremely high data compression factors in the range between 1000 and 10000. This object is achieved with the aid of a method for compressing image data with features according to claim 1.
- the image data are approximated by invariant densities of stochastic processes
- the carriers of these invariant densities ie those subsets of the express plane on which these invariant densities do not assume dwindling values, are also mostly of a fractal dimension.
- Wer ⁇ As with the method of the Barnsley the parameters that stochastic processes, whose invariant densities representing the image approximated used as there are suitable primêt representation of the image, or used in a compressed representation of the image '.
- the decisive difference to the Barnsley method is that the stochastic processes are not found by an experienced human operator, but with the help of an automatically running optimization method.
- This optimization method looks for the stochastic process that optimizes a quality function for evaluating the quality of the approximation of the image to be compressed by the invariant densities of stochastic processes.
- the values of the parameters of the stochastic process found in this way result in a compressed representation of the image to be compressed.
- the method is suitable for compression of any kind of digital raster images, in particular also for the compression of digital raster images from temporal sequences of such raster images.
- the image data to be compressed can in particular also have arisen when other bilodata compression methods are carried out, such as e.g. Image data in the transformation area of a transformation coding method, differential image data when a differential pulse code modulation method is carried out, images which are undersampled in space or time, or motion-compensated images.
- the compression method proposed here for image data can hereby be combined with a large number of known compression methods for express data.
- the method is also particularly suitable for compression of coloring image data, binary image data or gray image data. If the image data to be compressed is color image data, the method must be applied to each color component separately.
- Binary images can be regarded as special cases of gray value images, the gray values of which can only assume two (extreme) values, namely "black” or "white”.
- Image is present in different resolution levels of a resolution pyramid, and if the resolution is gradually increased with the progress of the optimization process which selects the stochastic process used for compression.
- the invention is based on two completely different, newer mathematical methods: first, the theory of iterated function systems, and second, the theory of genetic algorithms. Because both methods
- An affine mapping w (i) is a linear transformation of the image plane in itself, which converts every point x of the image plane into point w (i, x) of the image plane according to formula 2):
- a (i) is a 2 2 transform matrix belonging to w (i) and Bd) a translation vector belonging to w (i).
- the set of affine transformations forms a group in the algebraic sense. If affine transformations are carried out one after the other, the result is an affine transformation. If one carries out affine transformations w (i) in a random order with probabilities p (i) one after the other, one obtains a stochastic process, the properties of which depend on the parameters A (i), E (i) of the affine transformations and also depend on the probabilities or relative frequencies p (i) with which these transformations are selected. Under the in formula (3)
- a stochastic process of the type shown here has an invariant density, ie it gives a function on the Eild plane, that is, an image that is transformed into itself by this stochastic process or iterated function system.
- the carrier of this invariant density can be generated by the following simple experiment: For this purpose, the fixed point of one of the images w (i ), that is a point which is mapped onto itself from the illustration w (i). At this starting point x (0) one uses the images w (j) with the relative frequencies p (j) one after the other and thus obtains a set of points in the image plane which bear the invariant density of the stochastic associated with this parameter Process or iterated function system the better covered the longer you get this iteration process.
- the carrier of the invariant density of the stochastic process is generated as an image. This carrier is also known as the attractor of the iterated function system.
- the condition described in formula (3) is called the contractivity condition.
- Each point of the attraction works at a relative frequency.
- the stochastic P zeß which corresponds to the value of the invariant density of this stoc stical process at this point.
- the said attractor is invariant under the stochastic process, which corresponds to the parameters given by the sets W and P s.
- genotype is given by the sum of hereditary dispositions, the phenotype by the shape and behavior of the individual.
- the Ph notype is determined almost exclusively by the genotype.
- genotype When reproducing individuals, the genotype is passed on.
- the genotype of the offspring is modified by some processes which are only subject to chance and in particular are not dependent on the phano type: a) mutation due to incorrect copying of the genes or externally triggered processes; b) crossing genes (genetic crossover); c) Mutual exchange of individual paternal and maternal genes;
- a genetic optimization process is accordingly a stochastic, i.e. randomly controlled optimization process, in which a statistical set of suboptimal solutions to the optimization problem, represented by a population of individuals, is subjected to certain changes, controlled by the mechanisms of biological evulotion, with the aim of changing the statistical population to get from sub-optimal solutions, the members of which by an improved degree of adaptation, ie are characterized by better values of a quality criterion.
- Phanotype of invariant density of an iterated function system or stochastic process Individual member of a statistical population
- the 2nd generation is now created by "crossing" individuals of the 1st generation, whereby mutation, crossover and exchange of father and mother genes (each with a certain probability of occurrence) change the genotype.
- the individuals of the second generation will therefore differ from one another in their genotypes, and thus also in their phenotypes, and thus in their fitness. From now on, those parents who have a higher fitness compared to the rest of the population are preferred for the generation of offspring. From generation to generation, descendants are therefore preferred whose genotype was formed from the genotypes of the most adapted individuals of the parent generation.
- Gene exchange one or more of the parameters a. to p. that the individual has taken over from one parent are replaced by the corresponding parameter or parameters from the other parent.
- the set of probabilities p may need to be re-standardized.
- the number of individuals in a running population can either be kept constant or can be varied in accordance with the evolution (optimization). So you can e.g. increase the number of individuals in a population in order to increase the range of the gene pool with stagnating fitness values. On the other hand, it is appropriate to reduce the size of a population if the development proceeds sufficiently quickly in order to reduce the computational effort. In any case, only the best individuals in a population are adopted. Two approaches are possible for the temporal development of a population:
- Synchronous evulotion A certain number of offspring is bred from the population at time t, taking into account differential reproduction. A corresponding number of the best individuals, who can come from both the new offspring and the old population (ie from the older generation), are included in the population at time t + 1.
- Asynchronous Evulotion A progeny is generated from the population at time t, taking into account the differential reproduction. If the fitness level is high enough, it will immediately become part of the new population. One or more individuals with the weakest fitness may be removed from the population in order to adjust the population size to the desired value.
- the technical suitability of the described optimization method for image data compression depends crucially on the choice of a suitable fitness function.
- the fitness function that is to say the quality criterion, must, like any cost function, satisfy the usual conditions of consistency and monotony.
- common procedures e.g. the sum of all squares of differences between the gray values of the image to be compressed and the values of the invariant density, which extends over all pixels of the image plane.
- the procedure can possibly be if you squared differences in those image areas that are visually critical, e.g. (in the center of the image cde: in areas with high contrast) multiplied by higher weight actuators.
- the compression of an image can additionally be supported by the use of resolution pyramids:
- the image to be compressed is fed to the optimization process for searching for the best iterated function system in different resolution levels of a resolution pyramid: the fitness for the first generations of the evulotion is initially assisted a coarse image resolution has been tested (for example with the third stage of a quadtree display, or for example with the aid of an image which has been undersampled by a corresponding factor. solution at suitable times, for example when the fitness values are only slowly changing, are gradually improved until the original image is resolved. In this way, considerable savings in computing time and thus a faster execution of the method are possible.
- the decompression of an image compressed with this method is very simple: for this purpose, the invariant density of the iterated function system or stochastic process used to compress the image is simply calculated. This happens e.g. with the help of the method described in the section "iterated function systems", in which - starting from a starting point in the image plane, the attractor and thus the invariant density of an iterated function system by repeated application of the affine transformations belonging to it with the relative frequencies belonging to it on the resulting points of the image plane - is calculated.
- the company Iter ⁇ ted Systems Inc. already mentioned, has developed hardware with which images in the format 256 x 256 pixels can be reconstructed from their compressed representation in the form of the parameters of an iterated function system in 1/30 second. This shows that the image data compressed with the aid of this method can be decompressed at a speed which permits the use of the method described for decompression of temporal sequences of compressed images in real time.
- Both the compression of image data and the decompression of image data can be easily parallelized and can be carried out with simple hardware outlay.
- image decompression for example, it is necessary to calculate a random number, two multiplications and three additions per iteration of a pixel.
- Standard image processing hardware can be used to calculate the quality function.
- performing the image compression with a corresponding approximation quality generally requires considerably longer computation times than performing the decompression, so that the method described appears to be particularly suitable for use in cases in which the time used to compress the image data is not particularly critical , as long as the decompression of the image data is fast enough, e.g. is possible in real time.
- the method is therefore particularly suitable for use in the compression and decompression of medical image data or other image archives.
- the method can also be carried out with the aid of a computer program which is annexed to this patent application.
- mapping (xx (IFS-x IFS)) (yy (IFS-y IFS))) (cons (setf (IFS-x IFS ) (floor (+ (* (aref (IFS-a IFS) j) xx) (* (aref (IFS-b IFS) j) yy) (aref (IFS-e IFS) j)))))
- test-IFS current-IFS push current-IFS * IFS-pop *
- setq * IFS-pop * sort * IFS-pop * # ' ⁇ : key #' IFS-fitness
Landscapes
- Engineering & Computer Science (AREA)
- Multimedia (AREA)
- Signal Processing (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Image Processing (AREA)
- Image Analysis (AREA)
- Apparatus For Radiation Diagnosis (AREA)
- Measuring And Recording Apparatus For Diagnosis (AREA)
- Compression, Expansion, Code Conversion, And Decoders (AREA)
Abstract
Description
Claims
Priority Applications (3)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
EP92906343A EP0577631B1 (de) | 1991-03-28 | 1992-03-10 | Verfahren zur kompression von bilddaten |
JP4505474A JPH06506326A (ja) | 1991-03-28 | 1992-03-10 | 画像データの圧縮方法 |
DE59206430T DE59206430D1 (de) | 1991-03-28 | 1992-03-10 | Verfahren zur kompression von bilddaten |
Applications Claiming Priority (2)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
DE4110400A DE4110400A1 (de) | 1991-03-28 | 1991-03-28 | Verfahren zur kompression von bilddaten |
DEP4110400.5 | 1991-03-28 |
Publications (1)
Publication Number | Publication Date |
---|---|
WO1992017981A1 true WO1992017981A1 (de) | 1992-10-15 |
Family
ID=6428510
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
PCT/DE1992/000199 WO1992017981A1 (de) | 1991-03-28 | 1992-03-10 | Verfahren zur kompression von bilddaten |
Country Status (6)
Country | Link |
---|---|
EP (1) | EP0577631B1 (de) |
JP (1) | JPH06506326A (de) |
AT (1) | ATE138770T1 (de) |
CA (1) | CA2106974A1 (de) |
DE (2) | DE4110400A1 (de) |
WO (1) | WO1992017981A1 (de) |
Cited By (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
GB2269719A (en) * | 1992-02-28 | 1994-02-16 | British Tech Group | Fractal coding of image, speech, handwriting or shape data |
DE4415167A1 (de) * | 1994-04-29 | 1995-11-02 | Siemens Ag | Telekommunikationsanordnung zum Übertragen von Bildern |
WO1997033254A1 (en) * | 1996-03-04 | 1997-09-12 | Iterated Systems, Inc. | System and method for the fractal encoding of datastreams |
CN103927584A (zh) * | 2014-04-17 | 2014-07-16 | 湖北欣纬应急科技有限公司 | 一种基于遗传算法的资源调度优化方法 |
Families Citing this family (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US5651099A (en) * | 1995-01-26 | 1997-07-22 | Hewlett-Packard Company | Use of a genetic algorithm to optimize memory space |
JP7036704B2 (ja) * | 2018-11-16 | 2022-03-15 | ヤフー株式会社 | 情報処理装置、情報処理方法、及び情報処理プログラム |
-
1991
- 1991-03-28 DE DE4110400A patent/DE4110400A1/de not_active Withdrawn
-
1992
- 1992-03-10 EP EP92906343A patent/EP0577631B1/de not_active Expired - Lifetime
- 1992-03-10 JP JP4505474A patent/JPH06506326A/ja active Pending
- 1992-03-10 CA CA002106974A patent/CA2106974A1/en not_active Abandoned
- 1992-03-10 DE DE59206430T patent/DE59206430D1/de not_active Expired - Fee Related
- 1992-03-10 WO PCT/DE1992/000199 patent/WO1992017981A1/de active IP Right Grant
- 1992-03-10 AT AT92906343T patent/ATE138770T1/de not_active IP Right Cessation
Non-Patent Citations (1)
Title |
---|
BYTE Januar 1988, ST PETERBOROUGH (US) Seiten 215 - 223; M.F.BARNSLEY ET. AL.: 'A Better Way to Compress Images' in der Anmeldung erwähnt * |
Cited By (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
GB2269719A (en) * | 1992-02-28 | 1994-02-16 | British Tech Group | Fractal coding of image, speech, handwriting or shape data |
DE4415167A1 (de) * | 1994-04-29 | 1995-11-02 | Siemens Ag | Telekommunikationsanordnung zum Übertragen von Bildern |
WO1997033254A1 (en) * | 1996-03-04 | 1997-09-12 | Iterated Systems, Inc. | System and method for the fractal encoding of datastreams |
CN103927584A (zh) * | 2014-04-17 | 2014-07-16 | 湖北欣纬应急科技有限公司 | 一种基于遗传算法的资源调度优化方法 |
Also Published As
Publication number | Publication date |
---|---|
ATE138770T1 (de) | 1996-06-15 |
DE59206430D1 (de) | 1996-07-04 |
DE4110400A1 (de) | 1992-10-01 |
EP0577631B1 (de) | 1996-05-29 |
CA2106974A1 (en) | 1992-09-29 |
EP0577631A1 (de) | 1994-01-12 |
JPH06506326A (ja) | 1994-07-14 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
DE3005775C2 (de) | Kodierverfahren für ein Farbbild | |
DE3546136C2 (de) | ||
DE69333288T2 (de) | Verbesserte vektorquantisierung | |
DE19503571B4 (de) | Bildkompressionsverfahren und -vorrichtung unter Verwendung eines bedingten Viergruppenbaumunterteilungssystems | |
EP1797536B1 (de) | Codierschema für einen ein zeitlich veraenderliches graphikmodell darstellenden datenstrom | |
DE4339753C2 (de) | Vorrichtung zum Komprimieren und Dekomprimieren von Bilddaten | |
EP0368139B1 (de) | Verfahren zur Codierung von Restfehlerbildern | |
EP0309669A2 (de) | Verfahren zur szenenmodellgestützten Bilddatenreduktion für digitale Fernsehsignale | |
DE60020887T2 (de) | Optischer fluss und bildaufbau | |
DE19928985A1 (de) | Rekursive Online-Wavelet-Datenkompressionstechnik für die Verwendung bei der Datenspeicherung und bei Nachrichtenübermittlungen | |
DE69907917T2 (de) | Selektive verfeinerung der maschen | |
DE60033716T2 (de) | Verfahren und Vorrichtung zur Bilddatenkompression | |
DE69820148T2 (de) | Verfahren zur Kompression/Dekompression von Bilddaten | |
EP1374559B1 (de) | Verfahren zur komprimierung und dekomprimierung von bilddaten | |
DE69724891T2 (de) | Bildkodierung beruhend auf Unterabtastung, Klassifizierung von Pixelblöcken und Zuordnung von Abbildungskoeffizienten | |
WO1992017981A1 (de) | Verfahren zur kompression von bilddaten | |
DE3917362C2 (de) | Verfahren zum Vermindern visueller Artefakte aufgrund von Kompression über eine Transformierte in medizinischen Bildern | |
DE19941183B4 (de) | Pseudo-Halbton-Bildausdrucksverfahren | |
DE69628892T2 (de) | Digitales Drucksystem und -verfahren | |
EP1034511B1 (de) | Verfahren zur umsetzung digitaler daten im raster einer ersten auflösung in digitale zieldaten einer zweiten auflösung | |
DE10152612B4 (de) | Verfahren zur Komprimierung und Dekomprimierung von Bilddaten | |
DE19944213C1 (de) | Verfahren zum Komprimieren eines digitalen Bildes mit mehreren Bit-Ebenen | |
DE60002580T2 (de) | 3d maschenkompression und -kodierung | |
DE69828019T2 (de) | Verfahren und Vorrichtung zur iterativen Bildtransformation und -dekodierung | |
WO1998015125A1 (de) | Verfahren und anordnung zur vektorquantisierung und zur inversen vektorquantisierung eines digitalisierten bildes |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
AK | Designated states |
Kind code of ref document: A1 Designated state(s): CA JP US |
|
AL | Designated countries for regional patents |
Kind code of ref document: A1 Designated state(s): AT BE CH DE DK ES FR GB GR IT LU MC NL SE |
|
DFPE | Request for preliminary examination filed prior to expiration of 19th month from priority date (pct application filed before 20040101) | ||
WWE | Wipo information: entry into national phase |
Ref document number: 1992906343 Country of ref document: EP |
|
WWE | Wipo information: entry into national phase |
Ref document number: 2106974 Country of ref document: CA |
|
ENP | Entry into the national phase |
Ref document number: 1993 122443 Country of ref document: US Date of ref document: 19930927 Kind code of ref document: A |
|
WWP | Wipo information: published in national office |
Ref document number: 1992906343 Country of ref document: EP |
|
ENP | Entry into the national phase |
Ref document number: 1995 402152 Country of ref document: US Date of ref document: 19950309 Kind code of ref document: A |
|
WWG | Wipo information: grant in national office |
Ref document number: 1992906343 Country of ref document: EP |