FR2955407A1 - Document i.e. text file, compression method, involves generating error of image reconstruction from calculated variation between image and image reconstruction, and compressing error - Google Patents

Abstract

The method involves performing recognition of a text (112) and recognition of a font (113) of an image (10) of a document, where the document is scanned using a scanner of a scanning device. A reconstruction (12) of the image is generated using a processor (20) from the recognized text and the recognized font. Variation between the image and the image reconstruction is computed. An error (13) of the reconstruction is generated from the calculated variation between the image and the image reconstruction. The error is compressed. Independent claims are also included for the following: (1) a processor cooperating with a scanning device for forming an image (2) a computer program product comprising a set of instructions for implementing a method for compressing a document (3) a method for decompressing an error (4) a system for decompressing an error (5) a computer program product comprising a set of instructions for implementing a method for decompressing an error.

Description

GENERAL TECHNICAL FIELD The present invention relates to a method of compressing a document comprising text, the document being scanned on a device comprising a scanner to form an image of the document and adapted to cooperate with a processor. The invention also relates to a processor for implementing such a method, as well as a computer program product which, when loaded onto a processor mentioned above, allows the implementation of such a method.

In addition, the invention relates to a method of decompressing an error generated by a compression method mentioned above. Finally, the invention relates to a system for implementing such a decompression method, and a computer program product which, when loaded on a system mentioned above, allows the implementation of such a method. STATE OF THE ART As shown in FIG. 1, the prior art envisages two solutions for compressing scanned images that represent text. As shown in Figure 1A, the first known solution is to treat the document as an image, regardless of whether it contains text. The first solution thus has the advantage that a view of the document, at the time of display or printing, is perfectly faithful to the original. It has the disadvantage that the generated files are of a large volume on a memory. As shown in FIG. 1B, the second known solution consists in applying a character recognition and a font recognition, using one of the several known and commercially available software, for example the OmniPage brand software. , and to store the document as a text file. The document is reconstructed as a view at the time of display or printing. The second solution thus has the advantage that the files generated have a high compression ratio and are of a low volume on a memory. It has the disadvantage that the view is not perfectly faithful to the original. In particular, and as shown in Figure 1 B, if the character recognition software is wrong (reading "ine" instead of "me" in the example of Figure 1 B), the error is directly passed on on display or when printing. In order to allow searching for text in the document, the first solution is sometimes supplemented by a character recognition. Two versions of the document are stored: a pure image view, intended for display and printing, and a representation in the form of a text file, invisible to a user, but used for searching. In this case, the file is obviously even larger. PRESENTATION OF THE INVENTION The invention proposes to overcome at least one of these disadvantages.

For this purpose, it is proposed according to the invention a method of compressing a document comprising text, the document being scanned on a device comprising a scanner to form an image of the document and able to cooperate with a processor, said method being characterized in that the processor performs text recognition and font recognition on the image; generates a reconstruction of the image, from the recognized text and the recognized font; calculates a difference between the image and the reconstruction; generates an error of the reconstitution, starting from the calculated difference; and compresses the error. The invention is advantageously completed by the following features, taken alone or in any of their technically possible combination: the processor calculates the difference between the image and the pixel-by-pixel difference reconstruction between the image and reconstitution; the processor calculates the difference between the image and the reconstitution by transparency, comparing the difference between the image and the reconstruction, for at least one pixel, at a threshold, and marking said at least one pixel as opaque or at least partially transparent depending on the result of the threshold comparison; the processor marks said at least one pixel as opaque in the error if the difference is greater than the threshold, and then copies in the error the corresponding color of the image; or mark said at least one pixel as opaque or at least partially transparent in the error if the difference is less than the threshold, and then copy in the error respectively the corresponding color of the image or a color promoting the compression step and storage; and the processor places a flag in the error or in a file to indicate a mode of generating the error to a decompression system. The invention also relates to a processor for implementing such a method, as well as a computer program product which, when loaded onto a processor mentioned above, allows the implementation of such a method. Further, according to the invention there is provided a method of decompressing an error generated from an image of a document having text by a method, said decompression method being implemented on a system having a microprocessor, and characterized in that the microprocessor: generates a reproduction of the image, from a text and a font recognized in the document; access the error and decompress it; and generates a view of the document by combining the error thus decompressed and the reproduction thus generated. The invention finally relates to a system for implementing such a decompression method, as well as a computer program product which, when loaded onto a system mentioned above, allows the implementation of such a method. The invention has many advantages. The invention makes it possible in particular to obtain a better compression ratio than the first known solution, while allowing faithful reproduction, even in the event of an error when recognizing 30 characters, unlike the second known solution.

PRESENTATION OF THE FIGURES Other features, objects and advantages of the invention will emerge from the description which follows, which is purely illustrative and nonlimiting, and which should be read with reference to the appended drawings in which: FIGS. , schematically represent views generated by known solutions according to the prior art; - Figures 2 schematically show the main steps of a compression process according to the invention, implemented in particular on a processor according to the invention; - Figure 3 schematically shows an error generated during the implementation of the invention; FIG. 4 diagrammatically represents the main steps of generating an error according to FIG. 3; - Figures 5 schematically represent the main steps of a decompression process according to the invention, implemented in particular on a system according to the invention; and FIG. 6 schematically represents the main steps of generating a view according to the invention. In all of the figures, similar elements bear identical reference numerals. DETAILED DESCRIPTION As shown in FIGS. 2B and 5B, the invention concerns both the compression of a document 1 comprising text 100, and the decompression of such a compressed document. Before compression, the document 1 is scanned on a device 2 comprising a scanner 3 to form an image 10 of the document 1. The image 10 can be stored on a memory of the device 2 or remote device 2, waiting for its treatment, or treated immediately. The device 2 includes but is not limited to a scanner, a printer, a photocopier, a fax machine, or any multifunction machine that can incorporate a scanner 3 for scanning a document with text 100.

As indicated in FIG. 2B, the device 2 is able to cooperate with a processor 20, comprising all the conventional storage (memory) and processing means for implementing a method according to the invention.

The processor 20 can thus be incorporated in the device 2 (solid lines), or remote (dashed lines), for example incorporated in a computer or any other assembly comprising a memory and processing means. In both cases, the processor 20 can thus access in particular the image 10 of the document 1.

The invention is first described with regard to compression. Figures 2, in combination with Figure 4, schematically show the main steps of a compression process according to the invention (Figure 2A), implemented in particular on a processor 20 according to the invention (Figure 2B). According to the compression method according to the invention, the processor 20 performs, during a first step E1, a text recognition 112 and a font recognition 113 on the image 10. In the example of FIG. Fig. 10 has the text "me" with two slight imperfections: one above the first bar of the "m" and another at the first leg of the "m". Character recognition and font recognition are known in themselves to those skilled in the art and are not described in more detail later in this description. They may be performed by the processor 20 of one of several commercially available software, such as the OmniPage brand software. In an optional second step E2, the processor 20 stores the recognized text 112 and font 113, for example on a memory 21. As shown in FIG. 2B, the memory 21 may be incorporated into the device 2, or processor 20, or to a remote set of the device 2 and the processor 20, for example on a remote server or another set comprising a storage memory.

The storage step E2 is optional, but in all cases, the processor 20 must have access to the recognized text 112 and font 113 for a step E3 described below. In the example of Figure 4, the recognized text 112 is "ine" and the recognized font 113 is Times New Roman. In a third step E3, the processor 20 generates a reconstruction 12 of the image 10, from the recognized text 112 and the recognized font 113. The reconstitution 12 is then in the form "ine". In a step E4, the processor 20 calculates a gap A between the image 10 and the reconstruction 12. The calculation of the difference A is described in more detail later in this description. During a step E5, the processor 20 generates an error 13 of the reconstruction 12, starting from the calculated distance A. The error 13 is represented at the end of the chain in FIG. 4, and more specifically in FIG. 3, and corresponds to the dark portions in these figures. In a step E6, the processor 20 compresses the error 13. In the step E6, the processor 20 can either store the error 13 on the memory 21 or transfer the compressed error 13 to a system decompression, for example via a telecommunications network. Very advantageously, error 13 is associated with recognized text 112 and font 113 for subsequent decompression described later. The error 13 contains little information and therefore compresses better than the image 10, with a smaller volume on the memory 21. Advantageously, the reconstruction 12 is erased to further gain 30 memory space. The recognized text 112 and font 113 can also be compressed. The compression, in particular of the error 13 performed by the processor 20, is known per se by the person skilled in the art, and may for example, but not exclusively, be of the jpeg ("Joint Photographic Experts Group") type according to the English terminology. generally used), png ("Portable Network Graphics" according to the English terminology generally used), jbig (Joint Bi-level Image Experts Group), etc.

Very advantageously, during the step E6 for example, the processor 20 also creates a file 110 comprising the recognized text 112, the recognized font 113, and the compressed error 13. The file 110 is stored preferably on the memory 21, or may be sent to a decompression system via a telecommunications network.

Two possible examples of error generation modes 13 are described here, starting from the gap A. According to a first mode, the processor 20 calculates, during the step E4, the difference A between the image 10 and the pixel-by-pixel difference reconstruction 12 between the image 10 and the reconstruction 12. The gap A pixel by pixel then allows the generation of the error 13 for the pixels for which the difference is non-zero. According to a second mode, the processor 20 calculates, during the step E4, the gap A between the image 10 and the reconstitution 12 by transparency. According to the second mode, the error 13 contains, for each pixel or group of pixels, transparency information (opacity, partial transparency or transparency) and color information. In the second mode, the processor 20 compares the difference between the image 10 and the reconstruction 12, for at least one pixel, with a threshold S. In other words, the processor can compare the difference pixel by pixel, or for a group of pixels of the image 10 and the reconstruction 12. The value of the threshold S is set by design and depends on the desired level of precision. The processor 20 then marks said at least one pixel as opaque or at least partially transparent as a function of the result of the comparison with the threshold S. If the difference is greater than the threshold S, the processor 20 marks said at least one pixel as opaque in the error 13, and advantageously then copies in error 13 the corresponding color of the image 10. If the difference is smaller than the threshold S, the processor 20 can mark said at least one pixel as opaque in the error 13 and can then copy respectively the corresponding color of the image 10 if the latter favors the step E6 of compression and storage. If the corresponding color of the image 10 does not favor the compression and storage step E6, the processor 20 can mark said at least one pixel as at least partially transparent in the error 13 and choose an opacity and a color that optimize the compression, while allowing to reconstitute the original image. The processor 20 may also mark the pixel or group of pixels as transparent, and in this case may choose any color favoring the compression and storage step E6 depending on the neighboring pixels.

Such management of the transparency as a function of the threshold S is known per se by those skilled in the art and is not described in more detail in the present description. The processor 20 advantageously places a flag 111 (or "flag" according to the terminology generally used by those skilled in the art) in the error 13 or the file 110 to indicate the generation mode of the error 13 (first mode or second aforementioned mode) to a decompression system, described below.

In a second step, the invention is described with regard to decompression. It is recalled that the compressed error 13 is stored on a memory 21, or compressed and sent to a system 2000 described below. The error 13 is for example contained in a file 110. FIGS. 5, in combination with FIG. 6, schematically represent the main steps of a decompression method according to the invention (FIG. 5A), implemented in particular on a system 2000 according to the invention (Figure 5B).

The system 2000 comprises in particular a microprocessor 200 having all the conventional means of storage (memory) and processing for the implementation of a method according to the invention. The system 2000 may also comprise a display screen 201 and / or a The system 2000 includes, but is not limited to, a microcomputer, a scanner, a printer, a photocopier, a fax machine, or any multifunction machine for displaying on the screen 201 and / or printing on the screen. printer 202 of a view 1000 of the document 1. It is understood that the system 2000 and the device 2 may form a single entity, or be separate and physically separated from each other, connected or not between them by a telecommunication network (for example LANs of the LAN type ("local area network" in the English terminology) or global networks of the Internet type.) As shown in FIG. 5B, the system 2000 is able to access in particularthe compressed error 13 and at least the recognized text 112 and the recognized font 113. As mentioned above, the access can be done by cooperation with a memory 21, or through a telecommunications network linked to the processor 20. According to the decompression method according to the invention, the microprocessor 200 generates, when a step E10, a reproduction 120 of the image 10, from the text 112 and the font 113 recognized in the document 1. In the example of Figure 6, the reproduction 120 has the form "ine". During a step E11, the microprocessor 200 accesses the error 13 and decompresses it. The decompression of the error 13, carried out by the microprocessor 200, is known per se to those skilled in the art, and depends on the compression, for example, but not exclusively, of the type jpeg, png, jbig, etc.

It is understood that for the steps E10 or E11, the microprocessor 200 can also access, where appropriate, the file 110 including including the text 112 recognized, the font 113 recognized, and the error 13. The microprocessor 200 generates, when a step E12, a view 1000 of the document 10 by combining the error 13 thus decompressed and the reproduction 120 thus generated, as shown in FIG. 6. Advantageously, the microprocessor 200 of the system 2000 uses the flag 111 of the file 110 or the error 13 indicating the generation mode of the error 13, to effect the combination.

The combination steps are described in more detail later in this description. The microprocessor 200 generates a faithful view 1000 of the document 1 for its display on the display screen 201 or on the printer 202, as shown in FIG. 6, even in the event of an error in the recognition of characters (" ine "in reproduction 120). Two possible examples of generation modes of the view 1000 are described here, as a function of the generation mode of the error 13, indicated to the microprocessor 200 by the flag 111. According to a first generation mode, the error 13 has been generated. as a difference, and the microprocessor 200 generates the view by adding the decompressed error to the reproduction 120. According to a second generation mode, the error 13 has been generated by transparency. If a pixel of the error 13 is marked as transparent, the microprocessor 200 copies the color of the corresponding pixel of the reproduction 120 into the view 1000. If a pixel of the error 13 is marked as opaque, the microprocessor 200 copies the color of said error pixel 13 in the view 1000. If a pixel is marked as partially transparent, the microprocessor 200 calculates an average between the color of the reproduction 120 and the color of the error 13, weighted by the opacity 30 said pixel of the error 13. Such a management of the transparency is known per se by the person skilled in the art and is not described in more detail in the present description.

The invention also relates to a computer program product which, when loaded onto a processor mentioned above, allows the implementation of a compression method mentioned above, as well as a computer program product which, when loaded onto a aforementioned system, allows the implementation of a decompression method above.

Claims (10)

REVENDICATIONS1. A method of compressing a document (1) comprising text (100), the document (1) being scanned on a device (2) comprising a scanner (3) to form an image (10) of the document (1) and suitable cooperating with a processor (20), said method being characterized in that the processor (20) performs (E1) text recognition and font recognition on the image (10); generates (E3) a reconstruction (12) of the image (10), from the recognized text (112) and the recognized font (113); calculates (E4) a difference (A) between the image (10) and the reconstruction (12); generates (E5) an error (13) of the reconstitution (12), starting from the calculated difference (A); and - compresses (E6) the error (13). 2. Method according to claim 1, wherein the processor (20) calculates (E4) the difference (A) between the image (10) and the reconstruction (12) by difference, pixel by pixel, between the image ( 10) and reconstitution (12). The method according to claim 1, wherein the processor (20) calculates (E4) the gap (A) between the image (10) and the reconstruction (12) by transparency, comparing the difference between the image ( 10) and the reconstruction (12), for at least one pixel, at a threshold (S), and marking said at least one pixel as opaque or at least partially transparent depending on the result of the threshold comparison. The method according to claim 3, wherein the processor (20) - marks said at least one pixel as opaque in the error (13) if the difference is greater than the threshold (S), and then copies in the error ( 13) the corresponding color of the image (10); or - marks said at least one pixel as opaque or at least partially transparent in the error (13) if the difference is less than the threshold (S), and then copies in the error (13) respectively the corresponding color of the image (10) or a color promoting the step (E6) of compression and storage. 5. Method according to one of claims 1 to 4, wherein the processor (20) places a flag (111) in the error (13) or in a file (110) to indicate a mode of generating the error (13) to a system (2000) for decompression. Processor (20) adapted to cooperate with a device (2) comprising a scanner (3) for forming an image (10) of a document (1) comprising text (100), said processor (20) being characterized in it is suitable for the implementation of a method according to one of claims 1 to 5. 7. Computer program product which, when loaded on a processor (20) according to claim 6, allows the implementation of a method according to one of claims 1 to 5. A method of decompressing an error (13) generated from an image (10) of a document (1) having text (100) by a method according to one of claims 1 to 5, said decompression method being implemented on a system (2000) comprising a microprocessor (200), and characterized in that the microprocessor (200): generates (E10) a reproduction (120) of the image (10); ), from a text (112) and a font (113) recognized in the document (1); - accesses (E11) the error (13) and the decompress (E11); and - generates (E12) a view (1000) of the document (10) by combining the error (13) thus decompressed and the reproduction (120) thus generated. 9. System (2000) comprising a microprocessor (200), characterized in that the microprocessor (200) is adapted for carrying out a method according to claim 8. 10. Computer program product which, when loaded on a system (2000) according to claim 9, allows the implementation of a method according to claim 8.