WO2020160991A1

WO2020160991A1 - Method and arrangement for selecting a color model conversion for an image compression

Info

Publication number: WO2020160991A1
Application number: PCT/EP2020/052255
Authority: WO
Inventors: Anko BÖRNER; Muhammad Asif Ali Rajput
Original assignee: Deutsches Zentrum für Luft- und Raumfahrt e.V.
Priority date: 2019-02-08
Filing date: 2020-01-30
Publication date: 2020-08-13
Also published as: DE102019201673B3

Abstract

The invention relates to a method for selecting a color model conversion for an image compression, comprising: a) selecting at least one partial image region (20) from an overall image (16), wherein the partial image region (20) has an initial color model; b) applying at least one color model conversion to the at least one partial image region (20), such that the partial image region (20) has at least one color model which is different from the initial color model; and c) selecting a color model on the basis of the results from step b) as the color model to be used for the compression of the overall image (16). Furthermore, the invention relates to an arrangement (10) for the image compression.

Description

Method and arrangement for selecting a color model conversion for a

Image compression

The invention relates to a method and an arrangement for selecting a

Color model conversion for image compression.

The compression of digital images or digital image files is widespread for the purpose of file size reduction. There are various standards, such as the JPEG standard and the JPEG2000 and JPEG-LS standards based on it. Such standards each define a procedure for the compression and decompression of image files, in particular in the form of a sequence of steps to be carried out. The steps can be conversions and transformations contained in an image file

Include image information.

A common step, especially in the context of JPEG-based compression, is a so-called color model conversion, also called color space conversion. In this case, an output color model of an image, which is typically an RGB color model, is transformed into another color model in order to reduce information contained and / or correlating multiple times in the output color model. This allows the file size to be reduced without any loss of information.

In other words, a correlation of color component-specific information, as is often the case in the RGB color model for the individual color components R, G and B, is intended to be reduced or not contribute more or at least less strongly to the file size. A typical example of a new color model that the

Information from the original color model is converted and thus transformed, is a YUV color model. Y is the brightness (or luminance) and U and V are each color components.

An important criterion for image compression is the required compression time. This depends at least in part on the selected color model conversion. Approaches have therefore been proposed in order to determine suitable and, in particular, rapid color model conversions on the basis of the initial color model. It was found, for example, that for images that have similar properties, for example with regard to their color balance or their (RGB) histogram (Brightness distribution of the individual RGB color components), similar color model conversions are suitable (Starosloski, R .: New simple and efficient color space transformations for losless image compression. Journal of Visual Communication and Image Representation, 25 (5): 1056-1063, 2014) . Such comparisons to

However, performing properties of other images in real time has not proven practical. There are also approaches for calculating an optimal

Color model conversion, which, however, is also not practical (Kamisetty Raman Rao, Patrick C Yip: The transform and data compression handbook, volume 1. CRC press, 2000).

It is an object of the present invention to improve and, in particular, accelerate the compression of digital image files.

This object is achieved by a method and an arrangement according to the attached independent claims. Advantageous further developments are given in the dependent claims. Furthermore, it is understood that the introductory

Features mentioned in the description can also be provided individually or in any combination in the presently disclosed solution, unless otherwise stated or evident.

A basic idea of the invention provides that at least one color model conversion is applied to at least one partial area of an image or an image file in order to select a suitable color model conversion based thereon.

Because the partial areas are smaller than the overall image, such a prior application of color model conversions can take place comparatively quickly. Thus, with a short time required, a

Color model conversion can be selected, for example a desired

Compression ratio made possible without, however, the calculation time for complete image compression (in particular a so-called calculation overhead) being unnecessarily increased.

In detail, a method for selecting a color model conversion for (digital) image compression is proposed with: a) selecting at least one partial image area from an overall image, the partial image area having an output color model;

b) Applying at least one color model conversion to the at least one image sub-area, so that the image sub-area is at least one of the

Output color model has different color model; and

c) Selecting a color model based on the results from step b) as the color model to be used for the compression of the overall image.

Compression or image compression can generally be understood here to mean the generation of an image file which has a reduced file size (i.e. requires a small amount of storage space) compared to an output image file. examples for

Image compression standards that can also be used in the present solution have been mentioned above (in particular the JPEG-LS standard).

The partial image area can have a reduced size compared to the overall image. For example, the overall image can be divided into any number of image sub-areas (tiles) or, in concrete terms, an image sub-area should typically take up between 0.1% and 5% of the area (and / or the number of pixels) of the overall image. This serves to a) minimize the computational effort and b) that relevant information is included. An image sub-area can comprise a contiguous set of pixels or a contiguous area of the overall image. In the case of a plurality of partial image areas, these are preferably defined to be the same size and can also adjoin one another and in particular completely cover the overall image.

The color model conversions can be carried out using standards known per se, for example using so-called RCT (software) modules (Reversible Color Transform). Examples of color models from which or into which a corresponding

Conversion can take place are shown in the table of the attached FIG. 3.

According to the invention, any combination and any conversion between at least two of these color models can be provided.

It can be provided that only a color model conversion is carried out and, for example, a compression ratio that can be obtained with this is compared with a compression ratio that can be achieved on the basis of the output color model. However, at least two color model conversions can also be carried out. General is it is preferred if at least two different color models are checked with regard to the image compression that can be achieved in each case and, for example, compared (for example an initial color model and a converted color model or two converted color models).

The result of step b) can in particular be defined as the compression ratio that can be achieved on the basis of the color models. In other words, complete image compressions can be carried out in each case in step b) on the basis of the various color models (for example, the starting color model and converted color model) and then based on the results (for example on the

Compression ratios) of these image compressions.

In general, it is preferred if one of the color model conversions (or a color model) is selected, which (or that) was actually also checked in step b). In other words, one of the color model conversions used in step b) can be selected as the color model conversion for the overall image in step c). However, it is also conceivable to specify a color model conversion as the standard conversion and only if one of the color model conversions from step b) delivers a better result (for example a compression ratio above a reference threshold value) to use this instead of the standard color model conversion. The standard color model conversion itself does not have to be used or carried out in step b), but this is also possible.

It goes without saying that the invention relates generally to digital image compressions and in particular can be carried out by means of digital image files or can be applied to such image files. All procedural steps can therefore be carried out with one

Computing unit (for example a microprocessor) are carried out, for example by executing associated program instructions on the computing unit.

In particular, the method can be carried out by means of a computer program product or defined by such, wherein the computer program product, when executed on the computing unit, can cause the latter to carry out or provide the steps, measures and functions described herein.

According to a development of the method and the arrangement, a compression of the partial image area is performed on the basis of a respective color model conversion and / or with performed at least two different color models. In other words, the partial image area is completely compressed (for example by means of a conventional standard such as the JPEG-LS standard), with a color model conversion being taken as a typically initial step. As explained below, however, several selected partial image areas can also form a new partial image

put together and this partial image compressed, which nevertheless includes a compression of the partial image contained in the partial image.

In summary, in this development, at least two compressions of the at least one image sub-area are carried out, each with a different color model and in particular on the basis of an optional plurality of color model conversions from step b).

In this context, it can also be provided that an achieved compression ratio is determined for each compression (from step b)). The

Compression ratio can be defined as a ratio of the image size (or image file size) before and after the compression was performed (or vice versa). However, a difference in these sizes could also be considered, or a different one that enables the image files to be compared before and after compression

especially quantifying quantity.

In addition, it can be provided in this context that that color model is selected in step c) with which an above-average

Can achieve compression ratio. In the case of only two color models, the larger than average compression ratio is the larger compression ratio (i.e. the one that allows a more significant file size reduction). In general, the largest or maximum compression ratio is always an above-average compression ratio and can therefore be selected according to the invention (i.e. the color model with which the largest can be selected can be selected

Compression ratio).

It is therefore not absolutely necessary to actually determine an average compression ratio within the scope of the method or to carry out comparisons of the individual compression ratios with this compression ratio. Instead, the maximum compression ratio can always be considered, so to speak is automatically above average, or a number of achievable high compression ratios can be considered (for example the three largest) if the corresponding number is greater than half of the total color models considered.

The above aspects, which provide for a compression of the partial image area and optionally a determination of the compression ratios that can be achieved, have the advantage that available algorithms can be used and the method described here can be implemented reliably and with little effort. In general, a reliable statement can also be made about which color model is actually suitable for compressing the overall image.

A further development of the method and the arrangement provides that a sub-area of the overall image is selected as the partial image area:

whose color components correlate above average; and or

which has a below average number of edges; and or

which has a below average number of sharp features; and or

which is above average smooth (e.g. in terms of a low variance of the color values) in at least one color channel

It should again be pointed out that according to the invention no corresponding average actually has to be calculated and / or comparisons to such an average

Average value are required, which is optional but also possible. Instead, maximum or minimum values can also be considered, which always

are above-average or below-average values. The same applies if a predetermined number of the largest or smallest values is selected (for example from a list sorted by size) and this number is less than half of the total available values.

On the other hand, it goes without saying that the terms “above average” and

“Below average” in the present context can relate to other partial image areas of the overall image or of a reference image (in the sense of a typical comparison image) and / or reference state, which are initially examined or preselected, for example in or before step a). Then from these

Subareas and by determining the properties mentioned above another

Selection can be made and in particular those sub-areas with the corresponding above and below average properties are selected and in particular those in which the corresponding properties are maximally or minimally pronounced.

By selecting appropriate partial image areas it can be achieved that the

Color model conversions and especially one based on them (optional)

Complete compression of the partial image area carried out reliably and optimally (in the sense of a compression ratio) succeeds and succeeds in particular with a short calculation time. It is therefore avoided that the initial selection of a suitable color model conversion significantly increases the total calculation time or the calculation overhead associated with this selection. Instead, partial areas of the image are selected that have a preferred

Enable compression behavior and also provide a high level of significance for the

Have the compression behavior of the overall image based on the corresponding color model.

In order to determine a correlation of color components, a variable, which is explained below, can be determined which shows a value distribution of at least one of the

Describes color components. This applies in particular to the standard deviation explained below.

Sharp features can generally be referred to as features in which there is a significant local transition between image properties of adjacent image areas and in particular pixels. This can involve significant changes in gray values or general values of a respective color component. In other words, sharp features can be formed by significant local deviations between neighboring pixels or pixel areas with regard to their pixel values (in particular per color component). Examples of sharp features are points

(zero-dimensional), edges (one-dimensional) or corners that can be formed by at least two edges that run at an angle to one another. A deviation of, for example, at least 5% or at least 10% or also at least 25% can be defined as a significant deviation, it also being possible for this deviation to be adjustable by a user. According to one embodiment of the method and the arrangement, a plurality of partial image areas are selected and combined to form a partial image. This partial image can then be used to select a small color model and, in particular, be used as the basis for step b). In other words, several image sub-areas of the overall image can be selected, which do not necessarily have to be positioned adjacent to one another, and these can be combined to form a new partial image (for example by defining a corresponding new image file in which the selected image sub-areas are adjacent to one another).

The at least one color model conversion can then be applied to this partial image, which is synonymous with this color model conversion being applied to at least one partial image area and, more precisely, to the plurality of

Sub-image areas contained in the sub-image is applied.

It goes without saying that the compressions of the

Image subareas for selection of a preferred color model comprise a compression of the partial image contained in the image subareas or can be achieved by such a compression.

Forming or, in other words, assembling or creating a partial image has the advantage that the informative value of the results is increased, since several areas may also be selected from different areas of the overall image

Sub-areas of the image can be viewed simultaneously. It can therefore with a higher

Probability to be assumed that a determined

Color model conversion is actually suitable for the overall picture, as if only a single and possibly non-representative part of the picture were considered.

In general, the selection of the partial image areas that are combined to form the partial image can take place in accordance with any of the criteria outlined herein and in particular by means of the aforementioned consideration of correlating color components, edges and / or sharp features.

In general, it goes without saying that the present invention can also be applied to black-and-white or gray-scale images. In this case, under a

Color component can be understood as a corresponding gray value. In addition, a plurality of color components cannot be present in this case. According to a further embodiment of the method and the arrangement, to select the partial image area (in step a)), a variable (distribution variable) is determined which describes a distribution of the values of at least one color component of this partial image area.

The distribution can be a statistical distribution. The values of the color component can be pixel values. The distribution can be a frequency distribution of pixel values of the color component that are present within the image area, and the size can describe this frequency distribution.

In particular, the above-explained correlation of

Color components are described, for example if the same size is determined for each color component and a comparison of these sizes is carried out.

As an example, the size can be a standard deviation of the corresponding distribution for the at least one color component. Alternatively, a scatter or bandwidth would come into consideration, i.e. the difference between a maximum and a minimum value of the color component.

If such variables are determined for a plurality of color components and in particular for all color components, these can be offset against one another in order to determine an overall characteristic value. In particular, this

Overall characteristic the above-mentioned correlation of the color components

describe. According to one variant, an average value is formed from the corresponding plurality of variables, and in particular an average

Standard deviation for all color components in the image sub-area. The lower the standard deviation of a color component and / or a correspondingly formed average value for all color components, the more uniform the image and the more clearly correlate any multiple color components in the partial image area.

As explained below in the context of (but not limited to) the exemplary embodiment (s), small standard deviations and / or average values may be preferred since the corresponding image areas can be compressed more reliably or enable greater informative value with regard to the compression that can be achieved for the overall image.

A further development provides that for selection in step a) a plurality of

Sub-image areas is checked at least partially in parallel with regard to at least one predetermined property. The predetermined property can be any of the properties mentioned herein, for example the correlation of color components, the presence of edges and / or the extent of sharp features in a partial image area. Due to the partially parallel processing or

Verification, the time required to select one or more suitable partial image areas can be reduced. The parallel processing can for example be made possible by means of a computing unit (for example a microprocessor) which is suitable for this and / or on which a parallel processing, e.g. is possible by means of threads running in parallel.

According to a further embodiment of the method and the arrangement, a sequence of several color model conversions to be carried out in step b) is specified in advance, this sequence being able to determine the sequence according to which the

Color model conversions are to be carried out (i.e. which color model conversion is to be carried out first, which second, which third, etc.). Preferably, it is further provided that when a desired compression ratio can be achieved with a color model conversion, step b) is ended (for example canceled) and this color model conversion is used in step c).

In particular, after each color model conversion, the

Determined compression ratio and this with a predetermined target value

(desired compression ratio) can be compared. If this desired compression ratio is reached or exceeded, the processing of the specified sequence can be dispensed with and step b) can be terminated. Figuratively speaking, it can then be assumed that a sufficiently suitable color model conversion has already been determined and it is accepted that a possibly even better color model conversion due to the abortion of step b) is not additionally checked. This is advantageous, for example, when a desired minimum compression duration is to be maintained, but it does not necessarily have to be minimized. The invention also relates to an arrangement for image compression, comprising at least one computing unit which is set up to carry out the following (for example starting from an image file defining the overall image):

a) selecting at least one partial image area from an overall image, the partial image area having an output color model;

Output color model has different color model; and

The arrangement can be a conventional PC and the

The computing unit can be a microprocessor or a computer processor. In general, the arrangement and / or the computing unit can be operated digitally.

The arrangement can comprise any further development and any further feature in order to achieve all of the above and below steps, operating states and

To provide or execute functions. In particular, the arrangement can be set up to carry out a method in accordance with any of the above and below aspects.

In the following, an embodiment of the invention will be explained with reference to the attached schematic figures. Features that match in terms of their type and / or function can be provided with the same reference symbols across the figures. They represent:

1 shows a schematic representation of the individual components and

Software modules or functions provided with an exemplary arrangement according to the invention;

Fig. 2 is a schematic flow chart together with the individual

Functional representations of a method according to the invention, associated with method steps, which can be carried out with the arrangement from FIG. 1; and 3 shows an overview of color models and color model conversions that can be used according to the invention.

In Fig. 1, an arrangement 10 according to an embodiment of the invention is shown schematically. The arrangement 10 comprises a computing unit 12 in the form of a computer processor, which has access to a memory device 14 of the arrangement 10. The computing unit 12 and the storage device 14 are shown in a structurally integrated manner in FIG. The computing unit 12 is generally set up to carry out all of the method steps described herein. For this purpose, it executes program instructions stored in software modules or defined by the software modules. These program instructions and software modules are indicated schematically in FIG. 1.

A first software module 1, which performs image compression, and a second software module 2, which performs image decompression, can be seen in detail. The

Software module 1 receives an output image or overall image 16 as an input variable, e.g. in TIFF format. The image information of this image (e.g. in the form of individual pixel values of this image 16 and in particular of pixel values per color component) are then fed to a first function group 3 of the software module 1. There, by means of a selection function block 4, at least one suitable

Partial image area 20 (see FIG. 2) is selected from the overall image 16 made available.

This at least one partial image area 20 is then fed to a color model function block 5. The color model function block 5 outputs a color model to be used and / or a color model conversion required to achieve this color model as an output variable. These serve as an input variable for a color model conversion module 6 (RCT module) known per se, which is an initial stage and, according to the illustration shown, a preparatory stage for an image compression function 7.

In the case shown, the latter contains instructions for performing a JPEG-LS image compression, the image having already undergone a color model conversion due to the color model conversion module 6.

The output variable of the software module 1 is thus a compressed one

Image file 8 which is stored on the storage device 14. For decompression can this image file 8 can be decompressed by means of the further software module 2 by reversing the compression steps according to the decompression function 7 * and then converting or re-converting the color model according to the module 6 *. Thus, the output variable of software module 2 is again obtained

Overall image 16. The direct connection of features 16 and 6 and also 8 and 6 * is intended to generally express that a selected color transformation is to be applied to the entire image.

FIG. 2 shows a flow diagram of a method according to the invention that can be carried out with the arrangement 10 from FIG. 1. The steps shown including the associated functional representations relate in particular to the selection function 4 from FIG. 1. More precisely, FIG. 2 shows the criteria and form according to which partial image areas 20 are selected, which can then be converted by applying color model conversion to get based on it

To determine the color model for the overall image 16.

In a step S1, a plurality of partial image areas 20 are generated from at least one overall image 16, which is an initial image. These can also be referred to as picture tiles. In the example shown, the partial image areas 20 are each of the same size and completely cover the overall image 16. In particular in steps S1 and S2, a corresponding perspective illustration in FIG. 2 indicates that the method can also be applied to a plurality of overall images 16 and that the individual intermediate results can result correspondingly for each overall image 16.

In the case shown, the overall image 16 is a colored RGB image or has as

Color model has an RGB color model and thus contains the three color components (or color channels) red (R), green (G) and blue (B).

In step S2, for each image sub-area 20 and for each contained therein

Color component determines a size that is a statistical distribution of this

Color component within the image sub-area 20 describes, in particular a

(Frequency) distribution of pixel values for the color component. More precisely, the standard deviation in a partial image area 20 is determined for each corresponding color component. The standard deviation is given as d and is calculated according to the formula block 21 shown in FIG. I is an index for the individual image sub-areas, n the number of generated image sub-areas 20, x a value (pixel value) of the currently considered color component and m a mean value (more precisely average value or arithmetic mean) of the pixel values of this color component in a respective image area 20.

In FIG. 2 it is shown for step S2 that the standard deviation d is determined for each color component (in this case, for example, R) and for each image area 20 (0... N). These are then offset to form a total value S per image area 20 (0... N), specifically with the further, not separately shown standard deviations d for the others

Color components green (G) and blue (B). For this purpose, the function block 22 is shown as an example for step S2, in which the corresponding arithmetic mean

average standard deviation S per image area 20 (i = 0 ... n) is calculated. This average standard deviation S forms the size finally obtained (distribution size), on the basis of which the distribution of the color components within a respective image area 20 is described.

If the average standard deviation S has a high amount, the associated image area 20 is irregular or has color components R, G, B that differ greatly from one another and / or are unevenly distributed. If, on the other hand, the size S has a small amount, the image area 20 is comparatively uniform with regard to the distribution of color components. In addition, the correlate

Color components then stronger together. Eventually forming one

The average value S of the standard deviations d enables such possible correlations between the color components to be taken into account (the lower the average value S, the lower the correlation).

In general, in the example shown, small amounts of the average standard deviation S are preferred, since correspondingly uniform partial image areas can be compressed more easily with the function block 5 from FIG. It has also been shown that

Color model conversions, which were carried out based on compressions of such uniform image areas, allow greater transferability to others

Have partial image areas or achieve a preferred compression for the overall image with a higher probability. In addition, common compression standards and in particular JPEG-based standards can produce above-average compressions in image areas that have uniform properties. Correspondingly, partial image areas 20 that have a high number of details and associated irregularities, such as low correlations, cannot be representative of a compressibility of the overall image that can actually be achieved on the basis of a specific color model conversion.

Finally, it goes without saying that with the standard deviations d considered, in particular the presence of edges or generally sharp features within partial image areas 20 can be detected at least indirectly. The higher the number of sharp features, the higher the standard deviation d of the individual

Color components. Thus, by means of the standard deviations d, the

Uniformity of the partial image areas 20 are closed.

All of the operations explained in step S2 and in particular the determination of the respective standard deviations can be carried out in parallel to at least one further operation. In particular, the standard deviations and also the average standard deviation S for a respective partial image area 20 can be determined in parallel for a plurality of partial image areas 20. This shortens the total time required to carry out step S2.

In the purely optional step S3, there is then the associated sub-area 20 for each image

mean standard deviation S. In the context of step S3, a Gaussian filter or another soft-character filter can then be applied, namely generally to the information or pixel values of the image subareas and / or to a standard deviation image file explained below.

More precisely, an image file can be generated by adding the respective associated average at the positions corresponding to the image areas 20

Standard deviations S are entered. On this average

Standard deviations S can then be applied using the soft-focus filter. As a result, high and low values of the variable S that are isolated in relation to their neighborhood are compensated or these (local) deviations are reduced. Generating an image file from the values of the size S has the advantage that this image file can be processed reliably and with little effort using existing soft-character algorithms. In particular, existing soft-character algorithms can be used, since from their point of view this is a common image file.

Filtered average standard deviations S * per image area 20 are obtained as the result. These can then be sorted according to their size, for example from small to large. Those values S * or the associated image areas 20 which have the lowest S * value can then be selected from a corresponding dated list. For example, the two, three or four image areas 20 with the lowest S * values can be selected.

A corresponding selection is shown in step S4 by highlighting two image areas mo and m, the size of which is only to be understood schematically. It can be seen that these image areas mo and rri are not interrelated, since the selection takes place purely on the basis of the S * values (or only the unfiltered S values if the optional step S3 is not carried out). According to step S5, however, these image areas mo and m can be combined to form a new image or partial image 30 (and / or a partial image file)

can be put together, which can be achieved, for example, by storing them together in an image file and without having to generate an actually visually represented or representable file. This has the advantage that this partial image 30 can be converted in a subsequent step by means of common color model conversions and

Compression algorithms can be edited.

The use of the composite partial image 30 newly created in step S5 to examine which color model conversion is suitable is not shown by means of a separate figure. This partial image 30 is used in function block 5 from FIG. 1 in order to carry out iterative color model conversions and subsequent complete compressions of this partial image 30. The term “iterative” refers to the fact that at least one color model conversion is initially only applied to the partial image 30 and not directly not to the overall image 16, in order to initially assess its suitability for overall image compression. With reference to FIG. 1, the partial image 30 can be used instead of the original image 16 there and according to the straight arrow and without going through the

Function group 3 in block 6 can be converted using a first color model conversion and then compressed using block 7. Subsequently, the partial image 30 (or a copy thereof) can according to the straight arrow and without going through the

Function group 3 can be converted in block 6 with a second color model conversion and then compressed using block 7. The each received

Compression ratios can then be compared by computing unit 12.

In general, it is initially possible to carry out only a single color model conversion including image compression and to determine whether this leads to a better compression ratio than if no color model conversion is carried out (i.e. than if the entire image or partial image 30 is only based on its

Source color model is compressed).

This is based on the idea that the overall output image 16 will typically be present in an RGB color model, which in a large number of cases already enables a satisfactory compression. In contrast, it was recognized according to the invention that the ORCT color model can be advantageous for certain cases, for example for uniform natural image recordings. It can therefore be provided that a color model conversion from RGB to ORCT is the only color model conversion of partial image 30 that is carried out. Subsequently, image compression of the partial image 30 can also be carried out with this (i.e. on the basis of the ORCT color model). In a subsequent step, the compression ratios that can be achieved on the basis of the corresponding ORCT and RGB color model are compared with one another and that for the compression of the overall image 16 is selected which enables a higher compression ratio.

If this is the RGB color model, the overall image 16 can be without

Color model conversion can be compressed. If, on the other hand, the ORCT color model is selected to compress the overall image, a color model conversion into the corresponding ORCT color model then initially takes place in the overall image 16.

For the conversion between RGB and ORCT, reference is made to the attached table 3, in which numerous other color models are named, as well as their conversions at the compression (forward conversion), which according to the invention can be carried out iteratively. For the sake of completeness, the backward conversions to be carried out are also given, which would have to be carried out when the image was decompressed.

Alternatively, provision can be made for at least two color model conversions to take place in color models other than the initial color model, including the associated image compression and subsequent consideration of the compression ratios that can be achieved. Again, an image compression based on the original color model (i.e. without color model conversion) can optionally be carried out. The color model that enables the greatest compression ratio can then be used as the color model for the

Overall image 16 can be selected and the overall image 16 can be compressed using the selected color model conversion.

In this case it can also be provided that the sequence of the color model conversions to be carried out, as defined, for example, in the table from FIG. 3 from top to bottom, is predetermined and that these are carried out successively one after the other.

In particular, a compression without color model conversion based on the output color model (RGB) can (optionally) first be carried out and that

Compression ratio compared to a desired compression ratio. If the desired compression ratio is not achieved, a color model is converted into the ORCT color model and the compression ratio is compared again with the desired compression ratio, etc. As soon as the desired

Compression ratio is reached or exceeded, that color model is selected with which this result was achieved.

However, it is also possible to use all or more or fewer color models than is shown in FIG. 3. A predetermined minimum number of corresponding color models could also be used and a suitable color model could then be selected from the totality of the compression ratios achieved therewith, in particular the one which enables the greatest compression ratio.

Claims

1. Procedure for selecting a color model conversion for image compression according to the JPEG-LS standard, with:

a) selecting at least one partial image area (20) from an overall image (16), the partial image area (20) having a size reduced compared to the overall image and an output color model;

b) applying at least one color model conversion to the at least one image sub-area (20), so that the image sub-area (20) is at least one of the

Output color model has different color model; and

c) Selecting a color model based on the results from step b) as the color model to be used for compressing the overall image (16).

2. The method according to claim 1,

characterized in that the at least one partial image area (20) is compressed with at least two different color models.

3. The method according to claim 2,

characterized in that for each compression a achieved

Compression ratio is determined.

4. The method according to claim 3,

characterized in that in step c) that color model is selected with which an above-average compression ratio can be achieved.

5. The method according to any one of the preceding claims,

characterized in that a partial area of the overall image (16) is selected as the partial image area (20):

whose color components correlate above average; and or

which has a below average number of edges; and or

which has a below average number of sharp features; and or

that is above average smooth in at least one color channel.

6. The method according to any one of the preceding claims,

characterized in that, in order to select the partial image area (20), a variable (S) is determined which describes a distribution of the values of at least one color component of this partial image area (20).

7. The method according to any one of the preceding claims,

characterized in that a plurality of partial image areas (20) is selected and combined to form a partial image (30) on which step b) is based.

8. The method according to any one of the preceding claims,

characterized in that for the selection in step a) a plurality of

Image subregions (20) is checked at least partially in parallel with regard to at least one predetermined property.

9. The method according to any one of the preceding claims,

characterized in that a sequence of several in step b)

color model conversions to be carried out is specified in advance and, in particular, if a color model conversion is desired

Can achieve compression ratio, step b) finished and this

Color model conversions in step c) is selected.

10. Arrangement for selecting a color model conversion for image compression according to the JPEG-LS standard,

comprising at least one processing unit which is set up to carry out the following:

b) applying at least one color model conversion to the at least one image sub-area (20), so that the image sub-area (20) has at least one color model that is different from the initial color model; and