WO1996027977A1

WO1996027977A1 - Process for storing an image having a predeterminable data size

Info

Publication number: WO1996027977A1
Application number: PCT/EP1995/002155
Authority: WO
Inventors: Klaus Rindtorff
Original assignee: International Business Machines Corporation
Priority date: 1995-03-07
Filing date: 1995-06-06
Publication date: 1996-09-12

Abstract

In a process for storing an image having a predeterminable data size, the image is coded in a sequential, recursive manner. In a first step of the process, a representative value of the image is determined. In a second step, the image is subdivided into several sections and a representative value is determined for each section. In a third step, at least one of the thus obtained sections is recursively subdivided into further sections and a representative value for each of said sections is determined until a predetermined interruption criteria is reached. The coded values required to reconstruct the image are then sequentially stored in the same order in which the image was coded.

Description

DESCRIPTION

Method for storing an image with a predeterminable data size

Field of the Invention

The invention relates to a method for storing and

Compression of images.

State of the art

Data carrier cards, such as memory cards, with a memory area and intelligent data carrier cards which, in addition to a memory area, have at least one integrated circuit (IC), such as Smart cards, multi-functional smart cards, or smart cards, offer a wide range of applications. Among them is the area of personal identification, e.g. that of the cardholder. Other applications require, for example to exclude unauthorized access to the data carrier card, an identification method to prove the access authorization to the card.

One of the easiest ways to identify a person is to compare the person himself with an image of that person. It is therefore also intended that, for example, future machine-readable ID cards can also save an image in addition to information. This can be done, for example, using a data carrier card. The use of this technology also helps, among other things, to make these ID cards counterfeit-proof. In this context, it is then necessary to store an image in the memory of the card. However, the storage space of today's maps is limited. It is generally not sufficient to save a normal picture with the usual quality. This is especially true when rewritable memory is to be used.

Numerous compression methods are known from image processing, such as run length coding, dictionary coding, arithmetic coding and frequency component coding (JPEG). However, almost all known methods have at least one of the following disadvantages.

The data size of the (compressed) image depends on the image properties and cannot be predicted. This means that when reserving a memory space on the card, enough space must be reserved that the reserved memory size is sufficient for a sufficiently large part of the (compressed) images to be expected.

With a limited storage space, a small image with a correspondingly smaller amount of data is often chosen. Typically, however, the usual image compression methods have worse compression factors, especially for small amounts of data. The compressed data may even be larger than the original data.

The decompression process (restoration) of the image is usually done line by line and must be completed before the original image is restored. If it is canceled beforehand, parts of the picture have been restored, if possible. In addition, decompression is computationally intensive and the image cannot be viewed during this time. Summary of the invention

It is an object of the invention to provide a way to e.g. to store in data carrier cards, which offers simple coding of the image data and overcomes the disadvantages of the prior art mentioned.

The object of the invention is achieved as described in the independent claims.

The method according to the invention enables simple and compressible storage of digitized images, even in very small, fixed storage areas, such as, for example, in intelligent data carrier cards, such as memory cards or smart cards. It offers good data compression with lossless coding and a predeterminable compression factor with lossy coding. It still works effectively for very small images.

During the reconstruction of an image, the image is visible with increasing details. The process can be interrupted at any time, whereby an overall picture is always retained and the previously restored image information is incomplete, not spatially, but only in detail. It is therefore particularly suitable for the sequential reconstruction of images if only low data transfer rates are available when the stored image data is required to be transferred to a playback unit. Even after a small number of reconstruction steps, an image can be recognized by the viewer, so that, for example, a person can be recognized quickly.

The method according to the invention is also particularly suitable for storing images when only a small amount of fixed-length memory is available. With a sequential If the coded image data is saved in a storage space with a smaller data size than would be necessary to record the total amount of encoded image data, image details are lost, but the image is retained as an overall image and, if the data size is sufficient, also with a smaller detail resolution than the original image recognizable.

Images in the sense of the invention are not limited to photographic reproductions or video images, but can be images of any template. Photographs, fingerprints, iris or retina images or the like are particularly suitable as images for the identification of the card holder of a data carrier card, for example.

Further advantageous embodiments of the invention can be found in the subclaims.

Description of the drawings

To further explain the invention, exemplary embodiments are described in the following with reference to the drawings. Show it:

Fig. 1 shows a fixed scheme according to the invention for dividing the image sections;

Fig. 2 shows an example of an adaptive division of an image section with an essentially homogeneous color design in one area.

DETAILED DESCRIPTION OF THE INVENTION The basis of the invention is the decomposition of an original image with the aid of a recursive division of the image area. It is preferably a binary tree, four-tree (quadtree) or pyramid Data structure as known, for example, from "Algorithms for Graphics and Image Processing", Theo Pavlidis, Heise Verlag 1990, ISBN 3-88229-124-9. The trunk of this tree represents the entire picture. The leaves of the tree represent individual pixels that cannot be further divided.

Coding method

Before encoding an image, the original size or the aspect ratio of the image is saved. However, this can also be omitted if the exact, true-to-scale replication is not important in a subsequent reconstruction of the image, or if these data are fixed and therefore do not have to be stored separately.

During the subsequent coding, starting with the complete image as the first image section, an analysis of the colors or gray values used is carried out for the respective image section viewed. A color value or a gray value, the so-called representative color or the representative gray value, is determined from this analysis. Only the representative value (color or gray value) of an image section is saved for it. However, the method steps presented below for color values are not limited to these, but can be transferred analogously to gray values of a gray or black / white image or corresponding other values representing an image or image detail.

The representative value can be derived from various calculation rules. Such calculation rules are possible, for example, which use the mean value of the color values, the median of the color values or the determination of the most common color value. All points of the respective image section, or just samples thereof, can be used.

After determining the representative value of the currently viewed image section (parent-section) in several, not necessarily Alike _h large parts (child-outs) is broken. These children _A usschnitte then the parents-neck treated as before, that is, an analysis of the colors used or gray values performed for the respective Considered _h ended picture and determined from this analysis, the respective representative color or representative gray value.

The division continues recursively until no further division is possible, i.e. until the sections consist of a single pixel. However, the subdivision can also be terminated accordingly if a further, more precise subdivision is no longer desired or necessary.

The division of the image sections can be carried out adaptively, ie on the basis of the image information lying in the viewed image section, or according to a fixed scheme. Fig. 1 shows a fixed scheme for the division of the image sections with a four-tree structure. This has the advantage that the selected positions for a division are implicitly known and do not have to be coded. The current image section is preferably divided along its central axes, whereby any other division rule can also be used. In the case of a four-tree, this results in four new image sections per division. In each disassembly phase, all four sections are viewed in a fixed order. This repeats itself with each division phase, the number of cutouts quadrupling. All sections are preferably disassembled before recursively proceeding with another disassembly, but other disassembly orders can also be specified.

It is to be understood that the invention is not limited to the use of a four-tree structure, but that any decomposition structure can be used.

The entire image section 0 shown in FIG. 1 a is divided into the image sections 1-4 (FIG. 1 b) in a first division step. The representative value is determined for each of the image sections 1-4. Each of the image sections 1-4 is then divided into further image sections in a subsequent division step. Thus, section 1 is divided into sections 1.1, 1.2, 1.3 and 1.4, section 2 into sections 2.1, 2.2, 2.3 and 2.4, section 3 into sections 3.1, 3.2, 3.3 and 3.4, and finally section 4 into sections 4.1 , 4.2, 4.3 and 4.4. The representative value is determined accordingly for each of the subdivided image sections 1.1-4.4 and then each of the subdivided image sections 1.1-4.4 is subdivided into further image sections in a next division step. This division process continues until one of the abort criteria mentioned is met.

In the adaptive division of the image sections in accordance with the image information lying in the viewed image section, the optimal section for the division is based on the resultant result Coding effort determined. The division rule of each image section takes into account the content of this image section. In addition to the four-tree, binary tree structures are particularly suitable for adaptive divisions.

FIG. 2 shows an example of an adaptive division of an image section 10 with an essentially homogeneous color design in an area 20. A fixed division, e.g. a division into the child cutouts 30, 32, 34 and 36 shown in FIG. 2a leads to further divisions in the child cutout 30 containing the area 20. An adaptive division, as shown in FIG. 2b, in accordance with the recognized color design of the area 20, reduces the need for further divisions. If the resolution of area 20 is sufficient for the required resolution, further subdivision can already be made at this point. The adaptive division rule must be coded with the respective area.

It is to be understood that in the case of an adaptive subdivision, as shown in FIG. 2, the child cutout 30 that is created does not necessarily have to contain the entire area 20 or consist exclusively of a partial area of the area 20. With lower demands on the level of detail of the encoded image, e.g. if the area 20 is essentially contained or if the child section essentially consists of the area 20.

With the coding method described, images can thus be encoded in such a way that, in the case of a subsequent, sequential reconstruction of the image, which is explained below, the image is always an overall image represents increasing detail resolution. On the one hand, this allows the reconstruction method to be terminated when the reconstructed image has a sufficient resolution, and on the other hand, this method can also be used to limit the available storage space for storing the coded image to a fixed, predeterminable value. A sequential storage of the coded picture in a sequence corresponding to the coding, starting from the trunk, allows the storage of a full picture, regardless of the specified size of the storage space provided. If there is not enough storage space to store the amount of data for the entire image, the information from branches and leaves above that represent the detailed resolution of the image is omitted. If the image is then reconstructed from the limited amount of data, a resolution corresponding to the limited storage space has to be accepted, but a full image is always obtained. According to the invention, the method shown can therefore be used to store images with a predetermined, fixed data size. The resolution of the reconstructed image then depends on the image properties of the individual (original) image.

However, the encoding method described does not reduce (compress) the amount of data in the image itself, but even increases it, because not only the data of the individual leaves (pixels) of the tree must be determined and stored, but also the data of the respective associated nodes, right down to the trunk. For example, to save 64 pixels for a four-tree, 1 + 4 + 16 + 64 = 85 representative values must be saved. In general, the number of four-tree results representative values R to be stored at k quarters of the image sections for:

Thus ergi _b, a storage "overhead" Δ t of about _30%, or more specifically, at a divisible by 4 Number of pixels:

1 = 0

_D it alb _h emp _f iehlt for many applications a Mo _d ifikation the above encoding method with a compression of the image data.

Compression method

Possibilities for data compression arise, for example, from the compression methods known in data processing, such as frequency-dependent coding ₍ for example the Huffman coding) or the like.

Data compression can also be achieved for images with large contiguous single-color or structured areas. A further decomposition of the image sections can be terminated, for example, as soon as all the pixels (pixels) of the section have the same color or gray values.

Compression can also be achieved by determining and storing so-called decomposition information. This provides information about the type and progress of the dismantling and consists of a code that states whether or not to continue with the division (a section ₎ . Can a lossy compression of the image is accepted, the division of a section is terminated if a sufficient number of pixels have the same representative color value.

When decomposing the image, it is not necessary to save every representative color or gray scale. Instead of storing the representative color value for each decomposition, it is only specified and saved, for example, if there is a change compared to the previous parent section. So-called color information indicates whether the representative color has changed compared to the parent section (delta coding) and is optionally supplemented by the coding of the new representative color value if a change in the representative color is recognized by that of the parent section. This delta coding is chosen to be shorter than the color coding used (coding the representative color value) and provides information about the representative color value of a section, i.e. whether the representative color has changed compared to the parent section.

Starting from the spatial relationship of the distribution of the different colors in an image, a further improvement of the coding method can be achieved by further compressing the image. For this purpose, in a first step an analysis of the color change of the representative parent color to the representative child color is carried out by splitting the image as described above. The frequency of all color changes between the respective representative parent color and the representative child color is determined. The most frequently occurring color changes are determined and coded for each color. If one of the determined, most frequently occurring color changes are recognized, ie if one of the stored coded subsequent values matches the actually determined representative color value, this can be stored with a shorter coding, for example instead of the color value in the color information. A table of the most common subsequent colors for each of the colors in the image can be determined during image analysis and saved before the subsequent coding of the image begins.

The most frequent color changes determined will often be the color values of the immediately adjacent pixels in color gradients with the respective color. It was also shown that coding the two most frequent color changes for each color already leads to suitable compression.

In a further step, the image is coded with the help of the most frequent color changes of the respective colors determined and coded in the previous image analysis. Each newly determined representative color of a child section is compared to the most frequently occurring representative subsequent colors of the respective representative color of the parent section. If one of the coded, most frequently occurring color changes is recognized, the shorter coding of the color change is saved instead of the specific representative color of the child's image section.

From the possible sequence of decomposition and color information, there is a further possibility for improving data compression. For the coded data record of the respective child picture sections of a parent picture section, the delta codes are made from the respective coding of the new one Color values separated and added to the coded decomposition information. This combination is now regarded as a code word and the correspondingly required color values are appended to the code word formed. In the case of four-tree coding, a maximum of four color values occur if the representative color value deviates from that of the parent image section in all four child image sections.

When splitting an image, only a part of all possible code words occurs, but some of them are particularly common. A suitable Huffman coding is therefore preferably selected for the code words. This can be determined during an image analysis as described above. In general, however, a fixed coding is sufficient, so that the storage of the Huffman codes with the compressed image data is not necessary.

reconstruction

When an image is reconstructed from the encoded, stored data, the size of the image is first read out if it has been stored. For the simplest case of coding an image by successively dividing and determining the respective representative value, the reconstruction of the image is then carried out analogously to the coding by sequentially restoring the image, starting with the display of the representative value of the overall image and a step-by-step construction of the image by displaying the representative color values of the respective child image sections.

When reconstructing an image from the encoded and compressed stored data, the correspondingly stored tables, for example the subsequent colors, are first read in addition to the image size. This is followed by a sequence of code words with possibly additional color values. The sections to be considered are determined on the basis of the read decomposition information. If there are changes in the representative color, the respective section is colored with the corresponding color. This process continues until all code words have been processed. The order in which the sections are viewed is preferably the same as in the coding described above.

When reconstructing an image, a full image is always displayed with sequential improvement in the detail resolution of the image. During the reconstruction of the image, the displayed image already satisfies a required detail resolution, i.e. is e.g. a person already sufficiently recognizable from the displayed image, the reconstruction process can be canceled.

Disassembly order

In order to obtain a steady progress of the detail resolution over the entire image area during the reconstruction, the order of the viewing of sections during the coding and the reconstruction can be modified. Instead of strictly ordered processing, when using a four-tree structure, groups of four cutouts belonging together are preferably processed in a mixed order. Mixing can be achieved by using a suitable step size for a number n of cutouts. Since n will always be a multiple of 4, i.e. of the form n = 2 * 2 * k, the step size results from the smallest prime number p with p> k. The prime factor p is used as a step size and the calculation of the indices for the current section is carried out modulo n. This guarantees that everyone Excerpt is viewed exactly once in a fixed, but not ascending, order.

application areas

Data carrier cards, such as memory and smart cards, currently have a significantly limited storage space. This also often has to be permanently partitioned. The transfer of large amounts of data from the chip, e.g. to a reader, also takes a relatively long time. The method according to the invention can e.g. be used to store a, preferably relatively small, image on a card. The usable storage space can be determined as desired, it can also be smaller than the storage space required for lossless reconstruction. It is preferably determined in such a way that a sufficiently high quality in the reconstruction can be expected for a large number of images to be stored. Possible losses in the reconstructed image are only noticeable in detail and not spatially, since a full image is always displayed. When the image is called up, the entire image information is built up with increasing details. The retrieval can also be interrupted as soon as the desired level of detail has been reached or the image has been recognized or rejected.

Further applications are in all areas in which a memory area with a fixed size is to be used to take an image. For example, this is also the case for databases with a fixed record length.

Claims

EXPECTATIONS

1. Method for storing an image with a predeterminable data size, consisting of:

sequential, recursive coding of the image with:

a first step of determining a representative value of the image,

a second step of dividing the image into several child sections and determining a representative value in each of the child sections,

a third step of recursively dividing at least one of the multiple child clippings obtained so far into further child clippings, and determining a representative value in each of the newly received child clippings until a predetermined termination criterion is reached; and

sequential storage of the coded values necessary for a reconstruction of the picture in an order corresponding to the coding of the picture.

2. The method according to claim 1, wherein as

The termination criterion is such that a further division of a child section is no longer possible or is no longer desired.

3. The method of claims 1 or 2, wherein the encoding of the image comprises a step of determining the size of the image.

4. The method according to any one of the preceding claims with an additional step of specifying the data size available for storage.

5. The method according to any one of the preceding claims, wherein the division of a plurality of image sections is carried out according to a fixed scheme.

6. The method according to any one of the preceding claims, wherein the division of a plurality of image sections is performed adaptively on the basis of the image information lying in the viewed image section.

7. The method according to any one of the preceding claims, wherein the step of storing the values necessary for a reconstruction comprises a step of storing division rules used.

8. The method according to any one of the preceding claims, wherein a further division of an image section is terminated as soon as a predeterminable number of pixels of the image section have the same representative values.

9. The method according to any one of the preceding claims with a compression of the encoded data by determining and storing one

Disassembly information that provides information about the type and progress of the disassembly.

10. The method of claim 9, wherein the decomposition information indicates whether or not to continue the division.

11. The method according to any one of the preceding claims, wherein the determined representative values are only stored if there is a change compared to the previous parent section.

12. The method of claim 11, wherein encoded color information indicates whether the representative value has changed compared to the parent section.

13. The method of claim 12, wherein the color information is supplemented by encoding the new representative value when a change in the representative value from that of the parent section is detected.

14. The method according to claim 12 or 13, wherein the coding of the color information is chosen to be shorter than the coding used for the representative color value.

15. The method according to any one of the preceding claims with the additional steps:

the analysis of the changes between the representative values of the parent sections and the representative values of the respective child sections by decomposing the image and determining the frequencies of the changes, and

determining and coding the most frequently occurring changes; wherein the step of sequential, recursive coding of the image is carried out by means of the change of the representative values.

16. The method according to claim 15, wherein only a predeterminable number of the most frequent changes is encoded for each value.

17. The method of claim 15 or 16, wherein only two of the most frequent changes are encoded for each value.

18. The method according to any one of the preceding claims, wherein a frequency-dependent coding, preferably a Huffman coding, is applied.

19. The method according to any one of the preceding claims, wherein, in the case of a data size insufficient for storing the complete image, the last coded values are terminated.

20. A method for the reconstruction of an image encoded by a sequential, recursive encoding according to one of the preceding claims, comprising:

a first step of reading out and displaying a representative value of the image;

a second step of dividing the image into a plurality of child sections and reading out and displaying a representative value in each of the child sections;

a third step of recursively dividing at least one of the multiple child clippings obtained so far into further child clippings, and reading out and displaying a representative value in each of the newly obtained children's sections until a predetermined termination criterion is reached.

21. The method of claim 20, with an additional step of reading out and determining the size of the image.

22. The method of claim 20 or 21, wherein the third step is terminated when sufficient detail resolution of the reconstructed image is reached.

23. The method according to any one of the preceding claims, wherein at least one of the representative values represents a color value or a gray value or a black / white value of the respective section.

24. The method of claim 23, wherein the representative value is determined by an average, a median or by the most common value of the color or gray value.

25. The method according to any one of the preceding claims, wherein the predeterminable data size for storing the image is determined so that sufficient detail resolution is to be expected when reconstructing the image for a plurality of images to be stored.

26. Use of the method according to one of the preceding claims for storing an image in a data carrier card, preferably for identifying the cardholder.

27. Use of the method according to one of the preceding claims for storing an image in a database with a fixed record length.