DE19953610A1 - Font detection device for optical character recognition system selects system font from table whose width best corresponds to width of font in image - Google Patents

Abstract

The device includes a logic circuit (100) which receives an image including text characters. A logic circuit estimates the point size of the font using a number of font metrics (110,120) corresponding to a system font contained in a table (110,120) including font metrics and inverse font metrics. The logic circuit calculates a scaling factor based on the estimated point size. A logic circuit compares a width of the font in the image with the widths of each system font in the table. The system font that corresponds best to the width of the font in the image is selected. Independent claims are included for a method of determining the font of some text in an image, and an apparatus for detecting the font in an OCR system.

Description

The present invention relates generally to optical character recognition and especially on a Vorrich and a method for quick and efficient determination text font in an optical character recognition (OCR-; OCR = Optical Character Recognition) system.

Optical detection technology is used in scanning devices used to scan a printed document and the like be converted into an electronically editable format. Ty Typically, a user will scan a document feel that is attached to a computer. The rech ner includes software that interprets the scanned image animals, the image is converted into a text file and the text then to the user in an editable electronic form presents.

Specific OCR facility that is part of certain OCR systems include a limited degree of detection capability keiten. They typically only choose between a few Fonts based on whether the text has a fe most space between characters or is proportional, and whether the text has or has no serifs. A disadvantage of such a system is that reproduction of an original document in one way or another is possible that is evenly and exactly the original Can preserve the layout of the document. This is the case there different fonts different characteristics have different amounts of space on one side ingest even when using the same character size becomes. Different fonts consume varying amounts in horizontal space, which causes the lines of text to break different places. Consequently  Paragraphs of a scanned document without a text window skills are processed, not the same Number of lines as in the original document. This Situation prevents the exact reproduction of an abge scanned document by known optical character recognition packages.

Certain OCR devices have been used in the past character recognition to recognize characters in an image, however, they had limited success in making one Recognize and identify the target text font. Further prevents the reliability of a scanning system in the OCR character recognition capability seriously affects the operation of the Sy stems as a whole.

The object of the present invention is a Concept for quickly and accurately determining the font of a To create text in a scanned document so that the Document can be reproduced accurately.

This task is accomplished by a device for recognizing the Fonts of a text according to claim 1, a method for Determining the font of a text according to claim 5, and a device for recognizing the font of a text Claim 9 solved.

The present invention provides an apparatus and an Method for quickly determining the text font in one Text image.

Architecturally, the present invention can be used as a device to recognize the font of a text with the following note paint: a logic that is configured to receive an image in a font determination system where the image includes text characters that represent a font in the image represent, and logic configured to a Determine the point size for the font in the image by using a Majority of font metrics is used, the font  metrics also correspond to at least one system font, which is contained in a table, the table being the Contains font metrics and inverse font metrics, where the Font metrics a width for each system font in the table include. The invention also includes logic that is confi is gured to use a scaling factor to calculate estimated point size, taking the scaling factor corresponds to the point size, and being the scaling factor is averaged in a paragraph of the image, and wherein the logic is configured to match a width of the font in the Image with the widths of each system font in the table compare, with the logic configured to at least select a system font from the table that matches the Width of the font most closely matches.

The present invention can also be used as a method for Determine the font of a text displayed in an image that includes the steps of: Receiving a Image in a font determination system, the image being text contains characters representing a font in the image; Estimate a point size for the font in the image below Using a plurality of font metrics, the Font metrics also correspond to at least one system font, which is contained in a table, the table being the Contains font metrics and inverse font metrics, and where the font metrics a width for each system font in the Table included; Calculate a scaling factor under Use the estimated point size, being the scale factor corresponds to the point size and in one paragraph of the picture is averaged; Compare a width of the font in the image with the widths of each system font in the Table; and selecting at least one system font from the Table that most closely matches the width of the font in the image matches.

The present invention has many advantages, ei Nige are exemplified.  

An advantage of the present invention is that the quick determination of a text font in a text picture allows.

Another advantage of the invention is that it the determination of a text font using minimal information from an optical character recognition device tion allowed.

Another advantage of the present invention is in that they are independent of any optical device Character recognition can be implemented.

Another advantage of the invention is that it Design is simple and in large quantities for a com commercial production can be easily implemented can.

Preferred embodiments of the present invention are referred to below with reference to the attached drawing explained in detail. Show it:

Comprising 1 is a block diagram that illustrates a computer system is the Font determination logic of the present the invention.

FIG. 2 is a block diagram that further illustrates the font determination logic of FIG. 1;

Fig. 3 represents a schematic view, the metrics of two different fonts;

Figure 4 is a flow chart illustrating the operation of the present invention; and

Fig. 5A and 5B are schematic views that together represent the effect of the character bounding boxes and Wortbe grenzungskästen to the font determination logic of FIG. 2.

It should be noted that the components in the drawing are not necessarily to scale, but that emphasis was placed on the principles of the present to highlight the invention.

The font designation program of the present invention can in hardware, software, firmware or a combination of the same implemented. In the preferred embodiment example or the preferred embodiments is Implement the font determination program in software or firmware tiert, which is stored in a memory, and by an appropriate command execution system is executed.

Referring to the drawings, FIG. 1 is a block diagram showing a computer system that includes the font determination logic 100 of the present invention. A computer 15 may be any computer capable of executing the logic of the present invention. For example, computer 15 may be a personal computer system. A scanner 14 is connected to the computer system 15 . The scanner 14 , as is known, can be used to convert an image on a printed page to an electronically readable format. For example, the scanner 14 can scan a page of text, creating a scanned image, and then, after some further processing, recognize the text image and convert it to text characters in an electronically readable format. The computer 15 further includes a user interface 16 associated with the document capture system 36 , the font determination logic 100 of the present invention forming part of the same. Furthermore, the font determination logic 100 is shown as a discrete element for reasons of clarity. The same can also be integrated within the document recognition system.

The computer 15 further includes a word processor 38 and egg ne word processor user interface 33 which communicates with the microprocessor 18 via a logic interface 21 . The logic interface 21 is intended to represent the communication bus that binds components within the computer 15 . For the sake of simplicity, it is only shown in fragments. An optical character recognition system 19 is also included and in communication with the logic interfaces 21 . The system 19 for optical character recognition, although shown as an independent element, can be implemented modularly as part of the document recognition system 36 . A document storage device 39 is positioned between the document recognition system 36 and the word processor 38 . Word processor 38 is a typical application program that can use the output of document recognition system 36 .

A storage device 22 and the font determination logic 100 according to the present invention are also connected via the logic interface 21 . In one aspect of the present invention, storage device 22 further includes a conversion factor table 110 that contains a table of font metrics and a conversion factor table 120 that includes a table for inverse font metrics. The content of the conversion factor table 110 and the conversion factor table 120 contains text width and basic information relating to a plurality of fonts, which will be described in more detail below.

The scanner 14 communicates with the computer 15 via a connection 28, via which a scanned image is transmitted to the computer 17 Com 15 °. The scanned image 17 is then fed to the document recognition system 36 via a connection 27 . The document recognition system 36 supplies the data 17 of the scanned image to the OCR device 19 via a connection 41 . The OCR device 19 provides the input for the font determination logic 100 according to the present invention in the form of bounding boxes over a connection 24 and text characters over a connection 26 . The text characters that are supplied by the OCR device 19 are generated from the scanned image. The font determination logic 100 communicates with the memory 22 and in particular with the conversion factor table (font metrics) 110 and the conversion factor table (inverse font metrics) 120 via the connection 13 . The font determination logic 100 provides its output to the document recognition system 36 via a connection 34 in the form of a specific font, bounding boxes and text characters. As is known in the art, a bounding box is the smallest rectangle that can completely surround a particular graphic element.

The document recognition system 36 takes the output of the font determination logic 100 and integrates this information with a document to be presented (ie the scanned document to be displayed), which can be communicated via a connection 47 to the document memory 39 , where it is connected via a connection 48 to a word processor 38 Display via the word processor interface 33 is made available. The word processor interface 33 sends information to be displayed over a connection 44 to the display 12 . Alternatively, a document containing the scanned image and containing the output of the font determination logic 100 can be displayed by the document recognition system user interface 16 via a connection 46 to the display 12 . Alternatively, other word processing applications can be used to display the information to an end user.

FIG. 2 is a block diagram illustrating the font determination logic 100 of FIG. 1 in more detail. The font determination logic 100 includes a font size calculation logic 130 , a text size calculation logic 140 and a distance calculation logic 150 . Font size calculation logic 130 communicates with conversion factor table (inverse font metrics) 120 via connection 31 and text size calculation logic 140 communicates with conversion factor table (font metrics) 110 . Font determination logic 100 receives input from OCR device 19 in the form of bounding boxes over connection 24 and text characters over connection 26 . The bounding boxes on the connection 24 can take the form of character bounding boxes or word bounding boxes and are described in more detail below. The font determination logic 100 provides on the link 34 the determined font plus the bounding boxes and the text characters to the document recognition system 36 as described above.

The present invention takes advantage of the fact that different font metrics have different fonts. An example of different font metrics can be shown with reference to FIG. 3.

Figure 3 is a schematic view showing a portion of the font metrics of two different fonts. Font metrics can include font height and font width in addition to the metrics discussed below. To determine the point size of a particular text, the font determination logic 100 of the present invention divides a font character set according to the height of a character as follows:
Above: Under x, x-height, small top line, capital letter and full top line;
Below: Above the base, base, small underline and underline.

Using these classifications and predetermined measurements of characters from the fonts to be recognized, it is possible to determine the point size of a character for each font to be the measured size. From Fig. 3 it can be seen that using example word 40 and example word 50 the height and width of the same letters vary from font to font. As can be seen from word 40 , the capital letter "A", which is printed in the font "Arial", extends completely to the top line 41 , while the capital letter "A", which is printed in the font "Times New Roman" (Word 50 ) is printed, does not fully extend to line 41 . Similarly, the x-heights (the height of a lower case letter that has no descenders and no ascenders) of word 40 and word 50 are different. In particular, the letter "x" in the word has a greater x height than the letter "x" in word 50 . Similarly, descenders or portions of letters that extend below the x-height or below the baseline 42 , that is, that extend to the bottom line 44 , may vary from one font type to another.

In essence, the present invention determines the font of text in an image by analyzing the size of the font to determine the amount of space on a page that the text in one font or another would require. The invention analyzes a particular font (the font of interest in a scanned image), measures the size of the particular font, and then compares these measurements to the measurements (metrics) of known fonts in a table (such as the conversion factor table) (Font metrics) 110 and the conversion factor table (inverse font metrics) 120 ) and takes the most suitable font from the table to assign it to the particular font.

Fig. 4 is a flow chart illustrating the operation of the present invention. The flowchart of FIG. 4 shows the architecture, functionality and operation of a possible implementation of the font determination software of FIG. 2. In this regard, each block represents a module, segment or section of code that includes one or more executable instructions. to implement the specified logic function or functions. It should also be noted that in certain alternative implementations, the functions identified in the blocks may occur in a different order than in FIG. 4. For example, two blocks shown in succession in Fig. 4 may in fact be executed substantially simultaneously, or the blocks may sometimes be executed in the reverse order, depending on the functionality involved, as detailed below.

Furthermore, the font determination program, which is an ordered one List of executable commands to implement lo contains functions in any computer readable medium for use by or in connection with an instruction execution system direction or a command execution device the, such as B. a computer-based system, a Sy stem, which contains a processor, or another Sy stem that the commands from the command execution system, from the command execution device or from the command can call and execute the guidance device. In the context of this Document can be any "computer readable medium" Be a facility that contains the program, stores, communicates, carries on or transports so that it can pass through or in connection with the command execution system, the Be mis-execution device or the instruction execution device is used. The computer-readable medium can, for example wise an electronic, a magnetic, an optical, be an electromagnetic or an infrared medium, or but also a semiconductor system, a device, a device or a propagation medium. Specific examples (a non-exhaustive list) for possible computer-readable me these can include: an electrical connection (electronic) with one or more wires, a trag bare computer diskette (magnetic), a direct access memory (RAM) (magnetic), a read-only memory (ROM) (magnetic), an erasable, programmable read-only Memory (EPROM or flash memory) (magnetic), an op table fiber (optical) and a portable compact disc Read-only memory (CDROM) (optical). It should be noted that the computer-readable medium even paper or another  suitable medium on which the program is printed is because the program, for example, by means of an optical Scanning the paper or another medium electro niche can be captured, then compiled, then can be interpreted or otherwise can be suitably processed if necessary and then can be stored in computer memory.

Referring again to FIG. 4, the font size calculation logic 130 ( FIG. 2) in a box 101 determines the size and height category of the specified character or word. For each font, the point size for which the character height is closest to the character in the paragraph is determined. The dot sizes are then used to calculate a scaling factor relative to the dot size of the present text, which is determined by the OCR device ( FIG. 2) that provides the text characters on line 26 . The scaling factor is then averaged in the paragraph. This allows the point size in the paragraph to be constant, but it can vary as the height in the text changes significantly.

Either character limit boxes or word limit boxes can be supplied by the OCR device 19 of FIG. 2 via the connection 24 , while text characters can be supplied by the OCR device 19 via the connection 26 . Text characters are supplied by the OCR device 19 from the scanned image, as is known in the art.

Fig. 5A and 5B are schematic views that are in common the effect of the character bounding boxes and Wortbegren wetting boxes to the font determination logic 100 of FIG. 2 zei gene. Whether word bounding boxes or character bounding boxes from the OCR device 19 is supplied to the Font determination logic 100 determines the accuracy, created by the font determination logic 100 .

FIG. 5A shows word 60 and word 70 surrounded by word limit boxes 61 and 71 , respectively, while FIG. 5B shows word 60 and word 70 , each letter of each word being surrounded by character limit boxes 62 and 72 , respectively. Although only one character bounding box 62 is designated for word 60 and only one character bounding box 72 is designated for word 70 , it should be understood that each letter in each word is surrounded by a character bounding box.

If only the word bounding boxes are supplied to the font determination logic 100 , the difficulty in obtaining an accurate estimate of the font is greater than when character bounding boxes are available. If the word limit boxes 61 and 71 are supplied from the OCR device 19 , there is a smaller probability that words containing only x-height characters, such as e.g. B. the word 60 , are supplied. It is more likely that many more words, such as B. the word 70 , the at least a few ascenders, such as. B. the letter "t" in the word 70 , and a few descenders, such as. B. the book letter "y" included, which has been described with reference to FIG. 3 ben. For this reason, it is more difficult to estimate the size of the words when supplying word limit boxes because the probability of words containing only x-height characters is reduced. However, it is possible to generate the x-height if there is a sufficiently large sample of the text so that a few words with no descenders and ascenders, such as B. the word 60 can be found.

Referring again to FIG. 4 determined in block 101, the font determination logic 100, the largest and smallest book case letters in a word using the text characters from the OCR device is supplied 19 for each word, and using the character size metrics, as described above has been.

In block 102 , the height of the word from the bounding box 61 or 71 or the height of the character from the bounding box 62 or 72 is converted to a point size estimate by the font size calculation logic 130 using the conversion factor table (inverse font metrics) 120 of FIG. 2. Different point sizes are obtained for each of the fonts.

As described above, the point size estimate is converted into a scale factor with respect to the point size that is returned from the OCR device 19 . The scaling factors are averaged in a paragraph that contains the particular text that is being analyzed to ensure that a uniform dot size is applied to all words in a paragraph.

For each font in the table, the font size is predicted based on the category size of the image (font on the page) that is being analyzed using the font size calculation logic 130 in cooperation with the conversion factor table (inverse font metrics) 120 .

In a block 104 , the font determination logic 100, and in particular the text size calculation logic 140, removes the conversion factor table (font metrics) 110 in order to determine the width of the words for each font in the corresponding size. As used in this specification, the term "suitable" refers to the estimated point sizes predicted in block 102 for each font, in tables 110 and 120, respectively.

In block 105 , a distance metric between the system text size and the image text size at word level or at line level is calculated depending on the specific embodiment and the desired effect by the distance measuring logic 150 . In particular, the squared errors (between the actual size and the size returned by the text size calculation logic 140 ) are added up in word widths in a paragraph to obtain a total error margin for each font.

In block 106 , the font with the smallest distance metric is selected and output, this font being used to represent the paragraph.

In one embodiment of the present invention, the conversion factor table (inverse font metrics) 120 and the conversion factor table (font metrics) 110 hold five fonts. However, fewer or additional fonts can be kept in Tables 110 and 120 .

The following is an example of how to calculate the font size using heights and categories. The conversion factor table (inverse font metrics) 120 contains the pixel sizes s of a 10-point text in each font with every possible combination of size categories. The point size p becomes for a character (or a word) of height h by p = 10. h / s predicted before, where s is the size of the table for the height category of the sign (word). To represent an exception for a font with a fixed space between characters when the point size estimate is generated for the "Courier" font, the size is made (width + 6). 120 / (length. 300), where width is the width of the word in 300 dots per inch (DPI) pixels, and length is the number of characters in the word. This allows the "Courier" font to match any font with a fixed space between characters, even if it has a different vertical size. The total squared error of width will be large for any proportional font, but small for a fixed space font.

When the character bounding boxes are supplied from the OCR device 19 via the link 24 to the font determination logic 100 of FIG. 2, it is possible to estimate the dot size of almost any recognized character and the same when calculating the averaged scaling factor as described above was written to use. Characters for which the uppermost point or the lowest point do not match the "confirmed" measuring points are kept out of the point size estimation procedure. Confirmed measurement points include: x-height, capital letter, full ascender, and below: base and descender. These are the positions taken by characters, the size of which can be used reliably to predict the point size. The non-confirmed measuring points are at the top: below x and small ascender, and below: above the base and small descender. Characters such as B. "", "t", etc. are omitted because they fall into the measurement category "not confirmed". In addition to the fact that there are more characters to estimate the point size, there is another difference that occurs with character bounding boxes. The ratio of the x-height to the capital letter height and the ratio of the x-height to the lower length can also be used in the font selection process. By weighting the differences of these ratios against the width errors, it is possible to influence the font selection, either to get the widths correct or preferably to make the heights correct.

The ratio of the x-height to the capital letter height and that Ratio of the x-height to the descender can also be used after a given font for the paragraph is selected to capture individual words in which the font clearly differs from the given font det. These words can then be represented by another font be put.

It should be noted that the above described benen embodiments of the present invention and especially any "preferred" embodiments are only possible examples of implementations that have been carried out to illustrate the principles of the invention increase.

Claims (12)

1. Device ( 100 ) for recognizing the font of a text, having the following features:
logic ( 100 ) configured to receive an image ( 17 ) in a font determination system ( 100 ), the image ( 17 ) including text characters ( 26 ) representing a font in the image ( 17 );
logic ( 130 ) configured to estimate a point size for the font in the image ( 17 ) using a plurality of font metrics ( 110 , 120 ), the font metrics ( 110 , 120 ) also at least correspond to a system font contained in a table ( 110 , 120 ), the table comprising the font metrics ( 110 ) and inverse font metrics ( 120 ), the font metrics ( 110 , 120 ) being a width for each system font in the table ( 110 , 120 ) include;
logic ( 130 ) configured to calculate a scaling factor using the estimated point size, the scaling factor corresponding to the point size and averaged over a paragraph of the image ( 17 );
logic ( 150 ) configured to compare a width of the font in the image with the widths of each system font in the table ( 110 , 120 ); and
logic ( 150 ) configured to select at least one system font from the table ( 110 , 120 ) that best matches the width of the font in the image ( 17 ). 2. The device ( 100 ) according to claim 1, wherein the font metrics ( 110 , 120 ) comprise metrics selected from the following group: below x, x height, small upper length, capital letter, full upper length, above the base, Base, small descender, descender. 3. The apparatus of claim 1 or 2, further comprising a logic ( 19 ) configured to encode the text characters in word limit boxes ( 61 , 71 ). 4. The apparatus of claim 1 or 2, further comprising logic ( 19 ) configured to characterize the text characters in character delimitation boxes ( 62 , 72 ). 5. A method ( 100 ) for determining the font of a text in an image ( 17 ), comprising the following steps:
Receiving an image ( 17 ) in a font determination system ( 100 ), the image ( 17 ) including text characters ( 26 ) representing a font in the image ( 17 );
Estimating ( 102 ) a point size for the font in the image ( 17 ) using a plurality of font metrics ( 110 , 120 ), the font metrics ( 110 , 120 ) also corresponding to at least one system font that is in a table ( 110 , 120 ) is contained, wherein the Ta ble (120) contains the font metrics (110) and inverse font metrics (120) 110, wherein the font metrics for each system a width font in the table (110, 120) comprise;
Computing ( 102 ) a scaling factor using the estimated point size, the scaling factor corresponding to the point size and averaged over a paragraph of the image ( 17 );
Comparing ( 105 ) a width of the font in the image ( 17 ) with the widths of each system font in the table ( 110 , 120 ); and
Select ( 106 ) at least one system font that most closely matches the width of the font in the image ( 17 ) from the table ( 110 , 120 ). 6. The method ( 100 ) according to claim 5, wherein the font metrics ( 110 , 120 ) comprise metrics selected from the following group: under x, x-height, small upper length, capital letter, full upper length, above the base, Base, small descender, descender. 7. The method ( 100 ) according to claim 5 or 6, wherein the text characters ( 26 ) in word limit boxes ( 61 , 71 ) are summarized. 8. The method ( 100 ) according to any one of claims 5 or 6, wherein the text characters ( 26 ) are enclosed in character delimitation boxes ( 62 , 72 ). 9. Device ( 100 ) for recognizing the font of a text in a system for optical character recognition (OCR), having the following features:
means for receiving an image ( 17 ) in a font determination system ( 100 );
means for identifying a text region within the image, the text region comprising at least one text character representing a font in the image ( 17 );
means for estimating (102) (110, 120) further comprises at least correspond to a point size for the font in the image (17) using a plurality of font metrics (110, 120), the font metrics a system font, which in a table ( 110 , 120 ), the table ( 110 , 120 ) containing the font metrics ( 110 ) and inverse font metrics ( 120 ), the font metrics comprising a width for each system font in the table ( 110 , 120 );
means for calculating ( 102 ) a scaling factor using the estimated point size, the scaling factor corresponding to the point size and averaged over a paragraph of the image ( 17 );
means for comparing ( 105 ) a width of the font in the image ( 17 ) with the widths of each system font in the table ( 110 , 120 ); and
means for selecting ( 106 ) at least one system font that best matches the width of the font in the image ( 17 ) from the table ( 110 , 120 ). 10. The device ( 100 ) according to claim 9, wherein the font metrics ( 110 , 120 ) comprise metrics selected from the following group: under x, x height, small upper length, capital letter, full upper length, above the base, Base, small descender, descender. 11. The apparatus of claim 9 or 10, wherein the text characters in word limit boxes ( 61 , 71 ) are bordered. 12. The apparatus of claim 9 or 10, wherein the text characters are enclosed in character delimitation boxes ( 62 , 72 ).