FR2805640A1 - Generation of thumbnail image from an image stored in compressed form in a database, uses storage of multiple sub-images of differing size and selects image nearest optimal display size for scaling - Google Patents

Abstract

The thumbnail images are stored in the database as compressed multiple sub-images allowing extraction of images of different size. An optimum size is pre-determined (E115) and images extracted (E119). One sub-image is selected (E121) using a criterion of proximity of image size to optimal thumbnail, and the size of the selected sub-image is modified (E123) to obtain the optimal image size.

Description

The present invention relates to the field of storing digital images. More particularly, the present invention relates to a method for generating a reduced image, called an image, from an image stored in compressed form in a database.

The invention also relates to a device adapted to implement such a method.

conventionally, a database is queried through an interface software that uses a screen as a means of dialogue with a user. The latter typically uses a keyboard or a mouse to define the query on the screen.

When the database contains digital images, it is common for the reduced images, referred to herein as thumbnail, each representing a reduced image stored, to be displayed on the screen so that the user can see simultaneously multiple images and possibly make a choice among these.

When the database is not too large, the thumbnails are predefined and stored in memory, each thumbnail being associated with a stored image. On the other hand, when the database contains a large number of images, typically several thousand, the memory space required for storing images, indexes and thumbnails becomes too large.

The known solution is then to store the images in compressed form using, for example, a known coding standard such as JPEG (Joint Photographic Experts Group). Each piece of data stored in the database is then made up of a bit stream representing a compressed version of an original image.

A thumbnail associated with an image of the base is then generally stored, in compressed form or not, in connection with the index of the image.

However, the above solution is not totally satisfactory. Indeed, on the one hand, if the database contains a very large number of compressed images, the memory space required storage of the images, whether compressed or not, remains important.

On the other hand, when a thumbnail is stored with the corresponding compressed image, its size, expressed in number of pixels per row and column or equivalently in number of column lines, is fixed once and for all. This has the disadvantage of allowing the display of thumbnails of various sizes, determined example according to the dimensions of the viewing window on a screen, or according to the desired display resolution.

The present invention aims to overcome the aforementioned drawbacks. To this end, the present invention relates, in a first aspect, to a method of generating a thumbnail image from an image stored in a database, said database containing a plurality of images compressed according to a coding method. which allows the extraction, from each stored image, of a plurality of sub-images of different sizes, less than or equal to that of the stored image considered, characterized in that it comprises the following steps - predetermining a size optimal thumbnail; extracting the extractable sub-images of the stored image; selecting a sub-image from the sub-images extracted according to a size proximity criterion with the optimal size of a thumbnail; - Change the size of the selected sub-image to obtain the optimum size of the image, the resulting image thus constituting the thumbnail associated with said stored image.

By generating thumbnails from extractable sub-images from the stored images on demand, it is no longer necessary to store thumbnails (in compressed form or not) in the database with the corresponding stored images. This therefore reduces the memory space used in the database. On the other hand, by making it possible to predetermine a so-called optimum image size, the invention makes it possible to change the size of the images.

According to an advantageous characteristic of the invention, the optimum image size is predetermined so that the computation time necessary to obtain an image of this size from any image stored in the database is minimized.

By thus predetermining an optimal size of a thumbnail, the response time to a request for viewing a thumbnail image is minimized.

According to a preferred embodiment, the size proximity criterion consists in selecting the sub-image whose number of lines is the closest by excess or lack of the number of lines corresponding to said optimum image size, and whose number number of columns is the nearest excess or default number of columns corresponding to said optimal thumbnail size. modifying the size of the sub-image then consists in interpolating or sampling the number of lines of the selected sub-image, according to whether this number of lines is strictly less than or strictly greater than that corresponding to said optimum image size, and interpolating or sampling the number of columns of the selected sub-image, depending on whether this number of columns is strictly less than or strictly greater than that corresponding to said optimal thumbnail size. The definition of such a proximity criterion makes it possible to minimize the cost related to the interpolation sampling process.

In a preferred embodiment, the predetermination of an optimum thumbnail size includes the following steps: (A) Predefine a plurality of thumbnail sizes suitable for display; (B) selecting one of the predefined thumbnail sizes, referred to as the current preset size, and for each of at least a plurality of images stored in the database, (b1) calculating the size of the different extractable sub-images and selecting the closest one, according to the size proximity criterion, of the current predefined size; (b2) calculating a modification cost, associated with the stored image considered, indicative computation time to reduce the selected sub-image size to a current predefined size; (C) summing the calculated modification costs for each of the stored images considered in relation to the current predefined size and saving said sum of costs associated with the current predefined size; (D) repeat steps (B) and (C) for all predefined thumbnail sizes; (E) choosing the optimal thumbnail size, the predefined thumbnail size with the associated sum of costs is minimal.

The predetermination thus defined of an optimal size of thumbnails is simple to implement, and makes it possible to combine the possibility of changing the size of the thumbnails while minimizing the calculation time required for their generation.

According to a preferred embodiment, the coding method used for the compression of the images stored in the database is of the type using a spatio-frequency transformation by which image is represented a set of representative coefficients, said coefficients allowing to decompose the image in subbands with several levels of resolution.

Each image of the database is thus represented by a binary train said scalable resolution, that is to say from this bit stream, it is possible to extract images whose size is less than or equal to to that of the original image.

According to a second aspect, the present invention relates to a device for generating a thumbnail image from a stored image in a database, said database containing a plurality of images compressed according to a coding method that allows the extracting, from each stored image, a plurality of sub-images of different sizes, smaller than that of the stored image considered. According to the invention device comprises means adapted to implement a method of generating a picture as defined above.

The present invention also relates to a device for searching images in a database, characterized in that it comprises a device for generating a picture as defined above.

The present invention also relates to a computer, comprising means adapted to implement a method of generating a picture as defined above.

The invention also relates to a computer program comprising one or more sequences of instructions adapted to implement a method of generating a thumbnail according to the invention when the program is loaded and executed in a computer.

The invention also relates to an information carrier, such as a diskette or compact disc (CD), characterized in that it contains such a computer program. Advantages of this device, this computer, this computer program this information carrier are identical to those of the method as succinctly described above.

Other features and advantages of the invention will become apparent in the description below, made with reference to the accompanying drawings in which - Figure 1 is a block diagram illustrating the general architecture of an image search device incorporating the present invention; FIG. 2 represents a flowchart illustrating a sequence of steps of the image search method implemented in a device as represented in FIG. 1; FIG. 3 represents a flowchart illustrating a method of generating an image according to the invention; FIG. 4 illustrates the result of a multi-resolution spatio-frequency decomposition applied to a digital image; FIG. 5, composed of FIGS. 5a and 5b, represents a flowchart illustrating a method for determining an optimum size of an image according to the invention; - Figure 6 schematically shows a computer adapted to implement a method of generating a thumbnail according to the invention.

will firstly be described with reference to FIG. 1, a schematic block illustrating the general architecture of an image retrieval device incorporating a device for generating a thumbnail according to the present invention, according to a preferred mode. of realization.

In FIG. 1, a database 11 contains images stored in compressed form. Each image is therefore represented by a binary train. The images are compressed according to a coding method or compression algorithm, of a type that allows the extraction, from each stored image, of a plurality of sub-images of different sizes, less than or equal to that of the image. corresponding stored image. note here that we mean by size of an image, the dimensions this image in terms of number of pixels per row and column, or equivalently, in number of rows and columns. For example, an image of size 512 x 256 includes 512 pixels per line and 256 pixels per column, or 512 columns and 256 lines.

Returning to FIG. 1, an image data output unit 10 makes it possible to enter 11 new images to be stored in a database and to retrieve a stored image and / or indexed data (index) from an image. stored.

An example image input unit 30 is associated with an input device 31 to allow a user to input an example image to serve as a reference for the search. Typically, the input device 31 includes a pointing device such as a mouse and a keyboard.

image data storage unit 20 is intended to temporarily store the data recovered from the database 11 or data associated with the example image obtained by the units 30 and 31.

A unit 40 has the function of recovering indexed data from the images stored in the database or of generating this data for the new images to be stored in the database.

Similarly, a unit 50 is responsible for generating or retrieving the index data associated with the example image.

Each stored image of the database 11 is indexed as follows. When a new image must be stored in the database, it is previously extracted from it a data representative of its visual content. This data, for example characteristic of the distribution of colors in the image (color histograms), then constitutes the index of the image.

A similarity calculating unit 60 is responsible for evaluating the similarity of the example image with the images in the database. image selection unit 70 is responsible for sorting the images of the database according to their similarity with the example image and selecting one or more images as a result of the search.

An image generation unit 95 has the function of generating and temporarily saving thumbnails corresponding to the images previously selected from the unit 70. The unit 95 implements a method for generating thumbnails in accordance with FIG. present invention, which will be described later in the connection with FIG. 3.

display unit 90 typically comprising a screen, to display dialogue windows with the user as well as thumbnails, previously generated by the unit 95, corresponding to the images selected after a search, or, at the request of the user, to images directly extracted from the database.

 Finally, a control unit 80 controls and manages the overall operation of the image search device.

note that the various units that have just been described are made assembling electronic elements and software (respectively designated by the terms hardware and software in English).

The image search method implemented in the image search device shown in FIG. 1 will now be described with reference to FIG.

image search method starts with step E1 in which a human user uses the example image input unit 30 associated with the input device 31, to define an example image which will serve as a reference for searching images in the database.

The user has the choice to define the example image, to designate an image of the database, or to provide an external image to the database.

In order to be able to choose an image of the database as an example image, the user has the possibility of asking the display unit 90 to display on the screen thumbnails corresponding to the images of the database. The user can then select, using for example a mouse, the thumbnail whose content seems to be characteristic of the one of or images he seeks. Alternatively, the user can also use keyboard to enter the reference of the stored image that he chooses as an example image. Finally, if the user chooses to provide an external database example image, he may for example provide the unit 30 with the access path to this image, which will be accessible via a data acquisition device such as a diskette or compact disc (CDROM) reader integrated with the input device 31.

return to FIG. 2, once an example image has been entered by the user (step E1), in step E3, the index extraction unit 50 of the example image retrieves or generates the index of the example image and extract information characteristic of its visual content.

Similarly, in the following step E5, the stored image index extraction unit 40 extracts, from the index of each of the images stored in the database, information characteristic of the visual content of the image.

The characteristic information of the visual content of the example image and each of the stored images is temporarily saved in the image data storage unit.

In step E7, the similarity calculating unit 60 retrieves the indexes of the stored images and the index of the example image previously stored in the unit 20, and performs a similarity calculation between the image of the image. example each of the stored images whose index was extracted from the database.

In the following step E9, the image selection unit 70 sorts and selects the stored images that have been compared with the example image, according to their degree of similarity therewith. For example, only stored images whose calculated degree of similarity is greater than a predefined threshold will be retained. Of these only a predefined number of images (eg ten) will be kept as having the highest degree of similarity with the example image. in step E11, the image generation corresponding to the images selected in the previous step is carried out according to the invention by the image generation unit 95.

Finally, in step E13, the thumbnails generated in the preceding step and corresponding to the selected images are then displayed on the screen by the display unit 90, in a similar order (increasing or decreasing) with the example image. The user can then choose one or more images that he deems consistent with his request.

Referring to Fig. 3, a method of generating thumbnails in accordance with the present invention will now be described. According to a preferred embodiment of the invention, this method is implemented at (step E11, 2) by the image generation unit 95 of the image search device shown in FIG.

method of generating imagers shown in Figure 3, begins with step E111 in which the images to be displayed are recovered. For example, these are images selected as a result of the search (step E9, Fig. 2). These images are recovered in the database 11 by the image input / output unit 10, in their compressed form (i.e. in the form of bit streams). As mentioned above in relation with FIG. 1, the images are compressed according to a coding method or compression algorithm, of a type that allows the extraction, from each stored image, of a plurality of sub-images of different sizes less than or equal to that of the original image.

The binary data corresponding to the images to be displayed are then temporarily saved in the image data storage unit 20.

The next step, E113, is a test step in which it is determined whether optimal thumbnail size has already been determined or, if so, goes to step E117 in which the bitstream saved in the unit is selected. 20 and corresponding to an image to be displayed. This image is called current image. If not, proceed to step E115 in which the optimum size for a thumbnail is determined.

The process of determining the optimal size for a thumbnail, according to the present invention, will be described later in connection with the figure. Back to step E117, after having selected a first bit stream corresponding to an image to be displayed, goes to step E119 in which extracted from this bit stream, the various sub-images that can be extracted.

According to a preferred embodiment, the images stored in the database are compressed according to a coding method of the type that uses a spatio-frequency transformation by which image is represented by a set of representative coefficients, coefficients allowing to decompose the image. in subbands with several levels of resolution.

More specifically, in this embodiment of the invention, the images are compressed according to a method of subband coding using a so-called dyadic decomposition.

The sub-images of a given stored image are then obtained by extraction of the representative coefficients corresponding to the different subbands of this image, and inverse transformation of these coefficients.

In the aforementioned embodiment, the spatio-frequential transformation used by the coding method is a discrete wavelet transformation, designated in English by the expression digital wavelet transform (DWT).

Thus, at the end of such a decomposition, a stored image is represented by a set of spatio-frequency coefficients. Each subband of the image is then represented by a particular subset of the spatio-frequency coefficients. reference to <B> figure 4, we will recall below the principle of such a decomposition that can be described as spatio-frequency multi-resolution decomposition.

multi-resolution spectral decomposition means generally consist of a subband decomposition circuit or analysis circuit formed of a set of analysis filters, respectively associated with decimators by two. This decomposition circuit filters the image signal in two directions, in sub-bands of low frequencies and high spatial frequencies. The circuit comprises several successive analysis blocks for breaking down the image into sub-bands according to several resolution levels.

A first analysis block receives the image signal and the filter through two digital filters respectively low-pass and high-pass, in a first direction, for example horizontal. After passage in decimators by two, the resulting filtered signals are in turn filtered two filters low-pass and high-pass, in a second direction, for example vertical. Each signal is again passed in decimator two.

As an example, in Figure 4, the image is decomposed into subbands at a maximum decomposition level of 3 (# ,, naX = 3). Each sub-band is characterized by its decomposition level (@,) and its orientation (0).

then obtained at the output of the first analysis block, four sub-bands LLI = 1, 0 = 0), LH, (a, = 1, 0 = 1), HL, (1 = 1, 0 = 2) and HH , (a, = 1, 0 = highest resolution in decomposition.

sub-band LL1 comprises the low frequency components according to the two directions of the image signal. The subband comprises the low frequency components in a first high frequency direction in a second direction of the image signal. Subband HL comprises the high frequency components in the first direction and the low frequency components in the second direction. Finally, the sub-band HH comprises the high frequency components in both directions.

A second analysis block in turn filters the low frequency sub-band LL, to provide in the same way four sub-bands (X = 2, 0 = 0), LH2 (X = 2, 0 = 1), HL2 (1 = 2, 0 = 2) and HH2 (@, = 2, 0 = 3) intermediate resolution level in the decomposition. Finally, in this example, subband LL2 is in turn analyzed by a third analysis block to provide four sub-bands LL3 (1 = 3, 0 = 0), LH3 (1 = 3, 0 = 1), HL3 (a, = 3, 0 = 2) (a, = 3, 0 = 3) of the lowest resolution in this decomposition.

This gives 10 subbands and three levels of resolution. lower-frequency sub-band LL3 is called an approximation sub-band and the other sub-bands are called sub-bands of detail.

Of course, the number of resolution levels, and therefore of subbands, can be chosen differently, and for example be equal to four levels of resolution with 13 subbands.

Ainsi, en considérant une image originale de taille 512x512, décomposée selon un niveau de décomposition maximal a,max égal à 3, on pourra extraire les 4 sous-images suivantes. In general, starting from an image of any size (M and N natural numbers), decomposed according to a maximum decomposition level a, maX, we can extract (a, max + 1) sub-images. The size of these sub-images can be expressed by the formula

Thus, considering an original image of size 512x512, decomposed according to a maximum decomposition level a, max equal to 3, we can extract the following 4 sub-images.

A first subimage, of size 64x64, can be reconstructed from the coefficients of the sub-band LL3.

A second subimage, of size 128x128, can be reconstructed from the coefficients of the sub-bands LL3, LH3, HL3, HH3.

A third subimage, of size 256x256, can be reconstructed from the coefficients of the previous sub-bands and those of the sub-bands LH2, HL2, HH2. Finally, a fourth subimage can be reconstructed from the coefficients of the subbands of the decomposition levels 2 and and the coefficients of the sub-bands LH1, HL ,, HH1. This last image is size 512x512.

Therefore, from each image stored in compressed form according to such an encoding method, it is possible to extract a plurality of sub-images (in the preferred embodiment, four sub-images), smaller than or equal to the size of the original image.

In a preferred embodiment of the invention, the indexing of any image stored in the base consists in associating bit stream corresponding to this image, a color histogram calculated from sub-image, called low frequency, of maximum decomposition level extracted from said bit stream. In the case of the aforementioned wavelet decomposition, this is the LL3 sub-band.

return to FIG. 3, once the sub-images extracted for the current image (step E119), proceed to the next step, E121, in which the selection of one of the sub-images extracted previously is carried out according to a size proximity criterion with the optimal thumbnail size. The size proximity criterion used in a preferred embodiment of the invention will be detailed below.

At the following step E123, the size of the subimage selected in the previous step is modified so as to reduce the size of this subimage to the predetermined optimal size.

The method used to modify the size of the sub-image considered will also be detailed later in the description.

In step E125, the subimage whose size has been changed (optimum thumbnail size) is saved in the image data storage unit 20. This modified subimage represents the thumbnail associated with the current image to be displayed.

The final step E127 is a test step in which it is determined whether all the images selected for display (Fig. 2, E9) have been processed. If not, we go back to step E117 to select another image that becomes the new current image, and the process starts again with this image. On the contrary, if all the images have been processed, then all corresponding thumbnails have been generated and are stored in the previously mentioned unit. In this case, the image generation process is over. The thumbnails are then available to be displayed on the screen by the display unit 90.

We will now detail the size proximity criterion with the optimal size of thumbnail used to select (E121) a sub-image, as well as the method used to modify (E123) the size of the sub-image chosen, to make it equal to the optimum size mentioned above.

According to a preferred embodiment of the invention, the size proximity criterion consists in selecting the subimage whose size is greater than or equal to and the closest to the predetermined optimal image size, that is to say the one whose size is the closest in excess of the said optimum size.

For example, suppose that the original stored image considered has a size of 512x512, the maximum decomposition level (@max) is 3, and the predetermined optimal thumbnail size is 128x1. extract four sub-images the stored image considered, respective size 64x64, 128x128, 256x256 512x512. In this example, the selected subpicture will be the size 128x128 since its size is equal to the optimal size.

avec p E telles que

l'on applique la relation (2) à l'exemple précédent, on commence extraire les sous-images de taille 128x128, 256x256 et 512x512. In general, if we designate by Topt = LoxCo the optimal size of a predetermined image and if we consider a stored image of original size MxN compressed according to a decomposition process of maximum level a, maX, we will then start with extract from the image considered sub-images r of size:

with p E such that

we apply the relation (2) to the previous example, we start to extract the sub-images of size 128x128, 256x256 and 512x512.

Then, among the sub-images extracted by application of the relation (2), the aforementioned proximity criterion consists in selecting from this subset of sub-images, the one whose size is closest to the optimal size Topt. That is, the one whose size Ts satisfies TS = min dist (T, Tops <I>) (3) </ I> where dist (..) denotes a Euclidean distance and <I> min </ I> the minimum function.

According to a preferred embodiment of the invention, when the criterion of proximity used is that which has just been defined, the step of modifying (FIG 3, E123) the size of the sub-image selected in order to to obtain the optimal size of a thumbnail, is achieved by the use of a sampling technique of said selected sub-image.

Dans ce cas, l'étape de modification de la taille de la sous-image sélectionnée est réalisée par l'utilisation d'une technique d'interpolation appliquée à cette sous-image. Of course, as a variant, it is possible to envisage using a proximity criterion of selecting the sub-image whose size is closer by default to the optimal size, ie the sub-image size less than or equal to and the closest to the optimal size. We then use the following relation (2 ') instead of the previous relation (2)

In this case, the step of modifying the size of the selected sub-image is performed by the use of an interpolation technique applied to this sub-image.

The sampling and interpolation techniques are known from the state of the art, they use in particular an averaging technique.

In a second, more general variant, applicable in particular when, for stored images, the number of rows is not always equal to the number of columns, or when it is not certain that it can extract a sub-image satisfying the relation ( 2) above, we will use the following size proximity criterion.

The proximity-size criterion consists in selecting the sub-image whose number of lines is closer by excess or lack of the number of lines corresponding to the optimum size of the image, and whose number of columns is the nearest one by excess or defect. the number of columns corresponding to the optimal thumbnail size.

In this case, the modification step consists in interpolating or sampling the number of sub-image lines selected, depending on whether this number of lines is strictly inferior or strictly greater than that corresponding to the predetermined optimal size of a thumbnail, and interpolating or sample the number of columns of the selected sub-image, depending on whether this number of columns is strictly less than or strictly greater than that corresponding to the optimal thumbnail size.

Referring now to Figure 5 comprised of Figures 5a and 5b, a method for determining an optimum thumbnail size in accordance with the present invention will be described. This method is implemented in step E115 of Figure 3 showing a method of generating a thumbnail according to the invention.

As will be explained in connection with FIG. 5, according to this method, this optimal image size is predetermined so that the computation time required to obtain an image of this size from any image stored in the image. database is minimized.

Referring to Fig. 5a, the method for determining an optimum thumbnail size starts at step E501 in which a plurality of appropriate thumbnail sizes, denoted Sk (where k is between 1 and K), are predefined.

The optimal display size is for example preset according to the dimensions of the on-screen display window, the number of thumbnails to display simultaneously the desired resolution for the thumbnails. The sizes of thumbnails thus predefined are said appropriate criteria above. In an exemplary embodiment of the invention, the predefined image sizes are: 64x64, 64x128, 128x64, 128x128. The following step E503 is an initialization step in which a number of variables are initialized. The variables n and k are counters initialized to 1. The variables C1 to CK are loaded variables temporarily storing the costs calculated later in connection with each K predefined sizes of thumbnail.

In step E505, a first predefined size, called the current predefined size, is selected. In step E507, a first image of the database, called the current image, is selected. In reality, the image data input / output unit 10 retrieves from the database the bit stream, Bn, corresponding to this image.

In step E509, all the sub-images that can be extracted from the bitstream corresponding to the current image are extracted, and the size of these sub-images is calculated. Each of the sizes obtained is denoted T (r) with r between 1 and R, where R is the total number of different sizes of sub-images. In the preferred embodiment, as previously described in connection with FIG. 4, R is equal to 4.

In step E511, the subimage size, denoted T (ro), which is the closest to the current predefined size Sk according to the proximity criterion previously described in relation with FIG. 3, is selected. It will be recalled here that, according to a preferred embodiment, the proximity criterion consists in selecting the nearest sub-image size by excess of the current predefined size Sk.

In the next step E513, a modification cost is calculated, note O (T (ro), Sk), indicative of the computation complexity to reduce the selected subpicture size T (ro) to the current predefined size Sk Therefore, this cost is indicative of the calculation time. According to the preferred embodiment mentioned above, the calculation of the above-mentioned modification cost consists in evaluating an averaging sampling cost of the subimage whose size is that (T (ro)) selected as being the closest of the current preset size (Sk).

It is known that the complexity of a process of changing the size of an image is proportional to the ratio of the size of the start image to the size of the arrival image. Accordingly, such a modification cost, O (T (ro), Sk), can be expressed as a function of the ratio between the size of the subimage considered ro and the current predefined thumbnail size Sk.

In particular, it will be possible to identify the modification cost O (T (ro), Sk) at the ratio of the aforementioned image sizes. Thus, for example, if the size of the considered subimage is 512x512 and the current thumbnail size is 64x64, the calculated cost is 64.

Returning to FIG. 5a, once the modification cost O (T (ro), Sk) has been calculated, proceed to step E515 in which the value of the cost that has just been calculated is added to the content of the variable Ck associated with the current image size Sk, and initialized to zero in step E503.

In the following step E517, the variable n (initialized '1 to E503) used to count the images is incremented and its value is compared with the number N of images of the database chosen to determine the optimal size of the image.

In general, the images of the database chosen constitute a large sample of the images stored in the database. According to a preferred embodiment of the invention, consider all the images of the database. In this case, N is the total number of images stored in the database.

the value of the counter n is strictly greater than N, it means that steps E507 to E515 have been applied to all the images the sample of images. In this case, we go to step E519 in which the variable Ck associated with the current thumbnail size Sk is saved. Otherwise, it returns to step E507 to select another image (ie say binary train). Thus, the variable Ck saved in step E519, contains the sum of the modification costs calculated for all the images of the sample in question, in relation to the current image size Sk.

Steps E505 to E51 are repeated for all predefined thumbnail sizes (step E501). this effect, we test in step E521 the variable k (initialized at 1 in E501) used to count the selected thumbnail sizes. According to the same principle as for the counter n (E517), the counter k is first incremented and then its value is compared with the number K of predefined image sizes.

If the value of k is strictly greater than K, then all predefined thumbnail sizes have been processed, in which case step E523 of Figure 5b is taken. At this step, the minimum value of the variables CI,..., CK obtained previously respectively associated with the predefined sizes of the thumbnail C1,..., CK is calculated. The minimum value of the variable, denoted by Cko, therefore contains a cumulative modification cost for all the images of the sample considered, and associated with one of the predefined sizes of a thumbnail, denoted Sko.

Finally, in step E525, the predefined size of the Sko thumbnail associated with the Cko cost is saved as the optimal thumbnail size sought. This optimum image size is then used in the novel image generation method according to the invention, which has been previously described with reference to FIG.

The optimal thumbnail size determined according to the invention is optimal because it makes it possible to statistically minimize the complexity of the calculation necessary to generate an image of this size, from any image of the database. The minimization of the computation complexity necessarily induces the minimization of the calculation time, and therefore of the response time for the display of the images on the screen.

Referring now to FIG. 6, a computer adapted to implement a thumbnail generation method according to the present invention and previously described in connection with FIGS. 2 to 5 will be described. It should be noted that the computer illustrated in FIG. 6 constitutes a particular embodiment of the general device described above in relation with FIG.

In this embodiment, the image generation method according to the invention is implemented by a computer program. This program comprises one or more sequences of instructions whose execution said computer allows the implementation of the steps of the method.

In FIG. 6, the computer 3, which may typically be a workstation microcomputer, is connected to different peripherals, for example a digital camera 310 or any other image acquisition or storage device, such as a computer. scanner, providing information (images, video) to the computer 3. These images can be stored in the storage means available to the computer such as a hard disk 304.

The computer 3 also comprises a communication interface 308 connected to a communication network 309, for example a well-known Internet network, capable of transmitting digital information to the computer.

The computer 3 also comprises data storage means such as a hard disk 304, a floppy disk drive 305 for writing data to a floppy disk 305 and reading this data. The computer may also include a compact disc player (CDROM) (not shown) on which images can be stored, thereby constituting a database, as well as a computer card reader (PC-CARD) (not shown). .

According to a preferred embodiment of the invention, the executable code of the program for implementing a thumbnail generation method according to the invention is stored in the hard disk 304.

According to an alternative embodiment, the program executable code is stored in a read-only memory ROM (ROM) of the computer. According to a second embodiment, the executable code of the program can be downloaded from the communication network via the communication interface 308 to be stored on the hard disk Computer 3 further comprises a screen 302 for viewing the thumbnails and to act as a graphical interface between the program and the user, who can formulate queries using, for example, a pointing device (not shown) such as a mouse or an optical pencil, and a keyboard 303.

The computer comprises a central processing unit (CPU) 301, for example a microprocessor, which controls and directs the execution of the instructions of the program according to the invention stored in the read-only memory or in the hard disk 304. The central processing unit 301 then exerts a function of the control unit 80 described above in relation with FIG.

The computer also comprises a random access memory (RAM) 307 (RAM) having registers for storing the variables created and modified during the execution of the program, in particular the variables mentioned previously, with reference to the description of FIG. 5, as can be seen in the enlarged representation of the RAM in FIG.

Finally, the computer comprises a communication bus 1 to allow communication and interoperability between the various units mentioned above making up the computer 3.

Of course, many modifications can be made to the embodiments of the invention described above without departing from the scope of the invention.

Claims (1)

<B> <U> CLAIMS </ U> </ B> A method of generating a thumbnail from an image stored in a database (11), said database containing a plurality of compressed images according to a coding method which allows the extraction, from each stored image, of a plurality of sub-images of different sizes, less than or equal to that of the stored image considered, characterized in that it comprises the following steps: - predetermining (E115) an optimum image size; extracting (E119) the extractable sub-images of said stored image; - selecting (E121) a sub-image among sub-images extracted according to a size proximity criterion with said optimum image size; modifying (E123) the size of the selected sub-image so as to obtain said optimum image size, the image thus obtained then constituting the image associated with said stored image. A method of generating a thumbnail according to claim 1, characterized in that the optimum thumbnail size is predetermined such that the computation time necessary to obtain a thumbnail of that size from any image stored in the base of data is minimized. A method of generating a thumbnail according to claim 1 or 2, characterized in that said size proximity criterion is to select the sub-picture whose size is closest to the optimum thumbnail size; and in that the modification of the selected sub-image size to obtain said optimal thumbnail size performed by sampling said selected sub-image. A method of generating a thumbnail according to claim 1 or 2, characterized in that said size proximity criterion is to select a subimage whose size is closest by default to the optimum thumbnail size; and in that the size modification of the selected sub-image to obtain said optimum image size performed by interpolation of said selected subimage. A method of generating a thumbnail according to claim 1 or 2, characterized in that said size proximity criterion consists in selecting a sub-image whose number of lines is the closest by excess or lack of the number of lines corresponding to said an optimal thumbnail size, the number of columns of which is the closest by excess or default of the number of columns corresponding to said optimal thumbnail size, and in that the modification of the size of the sub-image consists in interpolating or sampling the number of lines of the selected sub-image, depending on whether this number of lines is strictly less than or strictly greater than that corresponding to said optimal thumbnail size, and interpolating or sampling the number of columns of the selected sub-image , depending on whether this number of columns is strictly less than or strictly greater than that corresponding to said optimum image size. 6. A method of generating a thumbnail according to any one of claims 1 to 5, characterized in that the step of presetting an optimum size of thumbnail comprises the following steps (B) Predefine (E501) a plurality thumbnail sizes appropriate for display; selecting (E505) one of said predefined thumbnail sizes, said current preset size, and for each of at least one of a plurality of images stored in the database, (b1) calculating (E509) the size of the different sub-images. extractable images and select (E511) that which is closest, according to said size proximity criterion, of the current predefined size; (b2) calculating (E513) a modification cost, associated with the stored image considered, indicative of a computation time for returning the selected subpicture size to the current predefined size; (C) summing (E515) the modification costs calculated for each stored images considered in relation to the current predefined size save (E519) said sum of costs associated with the current predefined size; (D) restarting (E521, E505) steps (B) and (C) for all predefined thumbnail sizes; (E) choosing (E525) as the optimal thumbnail size, the predefined thumbnail size with the associated sum of costs (E523) minimal. 7. A method of generating a thumbnail according to claim 6, characterized in that the calculation of said modification cost is to evaluate an averaging sampling cost of the sub-image considered to reduce size to the current preset size. A method of generating a thumbnail according to claim 7, characterized in that the sub-image averaging sampling cost is a function of the ratio of the sub-image size to the current preset thumbnail size. 9. A method of generating a thumbnail according to claims 6 to 8, characterized in that the plurality of thumbnail sizes suitable for display is predefined according to the screen size of the screen, the number of thumbnails to display simultaneously and the desired resolution for the thumbnails. 10. A method of generating a thumbnail according to any one of the preceding claims, characterized in that the coding method used for the compression of the images stored in the database is of the type using a spatio-frequency transformation by which an image is represented by a set of representative coefficients, said coefficients making it possible to decompose the image into sub-bands at several resolution levels. 11. A method of generating a thumbnail according to claim 10, characterized in that the sub-images obtained from a given stored image are obtained by extraction of the representative coefficients corresponding to the different subbands of this image, and then inverse transformation. coefficients corresponding to each of these subbands. 12. A method of generating a thumbnail according to claim 10 or 11, characterized in that the spatio-frequency transformation used by said coding method, is a discrete wavelet transformation (DWT). 13. A device for generating a thumbnail from an image stored in a database (11), said database containing a plurality of compressed images according to a coding method that allows the extraction, from each stored image, of a plurality of sub-images of different sizes, less than or equal to that of the stored image considered, characterized in that it comprises - means for predetermining (95; 300, 304, 307) a size optimal thumbnail; means for extracting (95; 300, 304, 307) the extractable sub-images from said stored image; means for selecting (95; 300, 304, 307) a subimage among sub-images extracted according to a size proximity criterion with said optimum thumbnail size; means for modifying (95; 300, 304, 307) the size of the selected sub-image so as to obtain said optimum image size, the image thus obtained thus constituting the image associated with said stored image. 14. Device for generating a picture according to claim 13, characterized in that it comprises means adapted to implement a method of generating a picture according to any one of claims 2 to 12. 15. Device (1) image search in a database (11), characterized in that it comprises means adapted to implement a method of generating a thumbnail according to any one of claims 1 to 12. A device (1) for searching images in a database (11), characterized in that it comprises a device for generating a thumbnail image according to claim 14 or 14. 17. Computer (3), characterized in that it comprises a device for generating a picture according to claim 13 or 14.