|Publication number||WO1989007804 A1|
|Publication date||24 Aug 1989|
|Filing date||9 Feb 1989|
|Priority date||9 Feb 1988|
|Also published as||CA1332791C, EP0398982A1, EP0398982A4|
|Publication number||PCT/1989/515, PCT/US/1989/000515, PCT/US/1989/00515, PCT/US/89/000515, PCT/US/89/00515, PCT/US1989/000515, PCT/US1989/00515, PCT/US1989000515, PCT/US198900515, PCT/US89/000515, PCT/US89/00515, PCT/US89000515, PCT/US8900515, WO 1989/007804 A1, WO 1989007804 A1, WO 1989007804A1, WO 8907804 A1, WO 8907804A1, WO-A1-1989007804, WO-A1-8907804, WO1989/007804A1, WO1989007804 A1, WO1989007804A1, WO8907804 A1, WO8907804A1|
|Inventors||Robert B. Nally|
|Export Citation||BiBTeX, EndNote, RefMan|
|Patent Citations (10), Referenced by (12), Classifications (15), Legal Events (5)|
|External Links: Patentscope, Espacenet|
QUALIFICATION SYSTEM FOR PRINTED IMAGES
- 1 - BACKGROUND OF THE INVENTION FIELD OF THE INVENTION
This invention relates to a system for automatically determining the quality of various types printed images. In particular, the system electronical compares printed images to a set of standard images stored in an electronic memory. DESCRIPTION OF THE PRIOR ART
Machines are presently available that will analyze printed images on a background material for the purpose of determining how closely the input printed image conforms to a specified printed image stored in a electronic memory. However, previous machines are not satisfactory in that they are not sufficiently accurate or they are not fully automatic or they require constan operator supervision or they are not sufficiently versatile or they require manual input.
SUMMARY OF THE INVENTION It is an object of the present invention to provide a system for automatically determining the quality of printed images where the system utilizes either a single element optical sensor or a multi-eleme magnetic and/or optical sensor or a combination thereof A system for automatically determining the quality of printed images on a background material by comparing said images to a set of standard images store on an electronic memory has at least one multi-element sensor that is selected from the group of magnetic or optical. The sensor is arranged to scan the printed images. The sensor produces a series of analog signals in the form of a two-dimensional electronic image for said printed images. The sensor is connected to amplifiers and filters and to an analog digital convert to convert said analog signals to digital signals. The are means for electronically ccr. --ring said electronic images with appropriate images of said set of standard images. 'There are means for producing a result of sai comparison.
In a variation of the present invention, the system has at least one sensor having an optical eleme said sensor producing a series of analog signals in th form of a two-dimensional electronic image for said printed images.
BRIEF DESCRIPTION OF THE DRAWINGS In drawings which illustrate a preferred embodiment of the invention:
Figure 1A is a block diagram of one embodime of the invention having two transducers;
Figure IB is a block diagram showing a breakdown of a printed image recognition logic portion the block diagram of Figure 1A;
Figure 1C is a block diagram showing a breakdown of a printed image analysis logic portion of Figure 1A; Figure 2 is a block diagram of a further embodiment of the present invention; and
Figure 3 is a block diagram of still a furth embodiment of the present invention.
DESCRIPTION OF A PREFERRED EMBODIMENT Referring-to the drawings in greater detail,
Figures 1A, IB and 1C, there is shown a block diagram a preferred embodiment of the qualification system of present invention. It can be seen that the system has one or more transducers 1, lx. The transducers can be various types, for example, magnetic, optical, acousti or another type. The transducers are sensors. The sensors are multi-element sensors in one embodiment of the invention v ": one of the elements, in all cases, being either magnetic or optical. In another embodime of the invention, the sensor can be single element or multi-element as long as the sensor always has at leas one optical element. The transducers can be single element transducers, or a two-dimensional array of multiple element transducers. From Figure 1A, it can be seen that the transducers 1, lx are connected to amplifiers in filte 2, 2x respectively. The transducers produce a plurali of analog signals for the printed images to be analyze and these signals are amplified and filtered to minimi distortion and to remove interference from the detecte signal representing the printed images. The transduce
1* lχ are connected to analog-digital converters 3, 3x respectively. The converters convert the analog signa from the transducers to digital signals having a numbe of grey levels. The digital signals are then fed into computer interfaces 4, from the converters 3, 3X respectively. The digital signals are temporarily sto in an electronic digital buffer memory before being in into a special purpose firmwear logic controller and/o computer 9, 10. The controller and/or computer 9, 10 perform the automatic printed image recognition functi 5 and the printed image quality analysis, measurement readability function 6. A set of standard images is stored in an electronic memory of the computer 10, the standards being laid down by an appropriate standards body, for example, ANSI (a trade mark), ABA (a trade mark), ECMA (a trade mark), CBA (a trade mark) or BSI ( trade mark) . Print quality standards are also included in the electronic memory of the computer 10. For example, OCR (a trade mark), MICR (a trade mark) and BA CODES (a trade mark).
The printed image recognition logic and the printed image analysis logic in the controller 9 togeth with the computer 10 electronically compare the electronic images with appropriate images of the set of standard images and produce a result of said comparison The result is communicated graphically to an operator o the system through a peripheral 11, for example, a printer or display, to allow the operator to interact with and query the results of the printed image quality analysis.
The system shown in Figure 1A has an optional diagnostic logic system 7 connected thereto. The purpo of the diagnostic logic 7 is to allow an operator to determine the cause of a system fault and to correct th fault. The fault would include any major out of tolerance conditions detected during a system calibrati process.
As a further option, the system also contains calibration logic subsystem 8 which allows an operator routinely check to ensure that the system itself is properly calibrated. The calibration of the system wil be checked by using a proprietary set of very accurate, high quality, printed images on an appropriate backgrou carrier for example, paper, for the particular type of useage for the system. The calibration images on the proprietary set may be printed in the magnetic and/or other types of inks depending on the desired use of the system and will consist of various shapes such as bars, squares, lines of various orientations and curvatures of varying widths, lengths, angles and spacing. When i the calibration mode, the system will analyze the inpu image of the calibration documents with respect to the ideal prototype requirements stored in memory and use this information to modify the key image quality analys parameters stored in the printed image recognition log 5 and the printed image analysis logic 6.
For example, if image motion is used, the calibration documents will allow the machine to determ whether or not the transport speed is within specifications and the machine will incorporate a correction factor into the analysis if the speed is no within specifications. As another example, if the expected amplification factor is not within the specification based on the reference levels of the calibration documents, a correction factor will be sto in the machine to eliminate this variation from the actual image quality analysis. The calibration logic could also be used to eliminate consistent noise. In this situation, a blank document could be passed in fr of the transducer or transducers and the noise level measured and stored. The signal could then be subtrac from the input image signal to minimize the consistent background noise component. in Figure IB, a block diagram is shown setti out a further breakdown of the Printed Image Recogniti Logic (henceforth PIRL) 5 of Figure 1A. The embodimen of Figure IB is based on the multi-element sensor of Figure 1A being a single slot magnetic transducer and linear optical array sensor.
In the embodiment shown in Figure IB, the PI operates on a digitized magnetic image stored in a digital buffer memory. The image data is pre-processe in an image filtering and thresholding processor 21 to remove electrical noise and unwanted frequency components. The optical image is also processed here enhance and sharpen the image and remove background information.
Next, the image data is scanned in a segmentation module 22 to detect and locate character data within the continuous stream of digitized data. appearance of a positive going peak may signal the starting location for individual character data. The process■begins by scanning the digitized data until a sample Si is located whose value is above a primary threshold Tl. Several checks are then performed to test for the presence of valid character data, namely: 1) contiguous data values must remain above the primary threshold level Tl for a minimum number of sample Nl; 2) contiguous data values must remain above a secondary threshold level T2 for a minimum number of samples N2, where T2 > Tl and N2 < Nl; 3) the sum of the data values for all samples whose values were continuously greater than Tl must be greater than a minimum energy threshold El; and 4) a negative going peak of value less than T3 must be found within N3 samples of the location of SI. hen a set of samples is found which satisfies the above criteria, a character is said to be located at SI. The actual values selected for Tl, T2, T3, Nl, N2, N3 and El are dependent upon the choice of sampling density and the resolution of the A/D converters 4, 4x.
When a character has been located at SI, key feature extraction 23 is performed. A fixed numer N4 of data samples immediately following SI is scanned to determine the locations and values of major peaks or features. The value of N4 is dependent upon the font being analyzed. The peak values of the features are combined to produce a normalization factor which is applied to the data for subsequent processing. Its purpose is to remove the effects of magnetic signal strength from the data so that data for all characters is roughly the same amplitude.
The key feature data, after application of the normalization factor, is next compared to a set of standard images P(i) in module 24 stored in electronic memory. There is one such predefined pattern for each character in the font being analyzed. As the key feature data for the character data being analyzed is compared to a standard image, a "Quality of Match" figure, Q(i) is generated for each pattern. The set of all Q(i) is forwarded to the next and final module 25 where the final classification decision is made.
The image classification decision logic 25 is presented with the set of Q(i) to make the final recognition decision. The algorithm is a simple minimum distance classifier, well described in the literature. For the present embodiment, the Q(i) represent "distances" of the data from the pattern. The lowest value of Q(i) is compared to a threshold T4. If the value of Q(i) is less than T4, the data is deemed to be correspond to the standard image P(i) and the character is recognized. If the value of Q(i) is greater than T4, the data is deemed unrecognizable.
The entire procedure 22 through 25 is repeated, working through all data in the image buffer until the end of data is encountered. At this point, all characters will have been located and an attempt will have been made to recognize them. If one or more characters were deemed unreadable, operator intervention rnay be required to identify the unknown characters.
It should be noted that the PIRL function is not an essential feature of the present invention. The purpose of the PIRL function is to enhance the operation of the image quality determination system by eliminating the need, in most instances, for the operator to manually input the printed image information to be analyzed.
When all characters have been identified or recognized, whether automatically or through operator intervention, the image data from all sensors are passed to the Printed Image Analysis Logic (henceforth PIAL) shown in Figure lc. Considering the present embodiment, the digital buffer memory containing the digitized image data acquired from each sensor 4, 4x is connected to a series of modules 31 through 39 for print characteristics measurement by control logic. The data for each sensor may or may not be operated on by a particular module, depending on the nature of the print characteristic bei measured. For example, for characters printed with magnetic ink, image data derived from an optical sensor would not be passed to a module intended to examine the magnetic signal strength. Similarly, image data derive from a magnetic sensor would not be passed to a module intended to measure the optical print contrast ratio.
The print characteristic measurement modules to 39 inclusive each examine a specific characteristic the print. The characteristics that are examined are a function of the sensor(s) and the print font specification in use and may include but are not limite to the following: 31 Image location and spacing measurement: determinin the exact placement of characters with respect to the edges of the document and with respect to each other, both horizontally and vertically;
32 Image print contrast/intensity measurement: determine the absolute levels, the variation from one character to the next and the variation within a character;
33 Image skew angle measurement: determine the angle the baseline of individual charactaers with respect to the baseline of the document;
34 Image stroke width measurement: determine the widt of horizontal and vertical strokes in each character;
35 Image edge variation measurement: examine the edge regularity of horizontal and vertical strokes in each character;
36 Image void measurement: examine all strokes of a character for voiding (absence of ink);
37 Image size and dimension measurement: determine th overall width and height of each character; 38 Image extraneous ink measurement: examine the sp around all strokes of each character for evidence of unwanted ink caused by ink splattering, ribbon flakin etc. ; 39 Image curvature measurement: examine individual strokes of each character to determine the curvature each stroke and the radii of stroke intersections and terminations where appropriate.
When all print characteristics have been examined and measured for each character, the measurements for an individual character are compared with a predefined set of measurements M(i) for the standard images, one such set existing for each characater defined in the font being analyzed. Individual measurements within each set M(i,j) are a function of the character font and document format requirements.
The results of the comparisons for each character R(i) are accumulated for the document and a judgment is made as to the acceptability of the print.
The criteria of the judgment are also functions of th character font and document format requirements.
For many of the print measurements, it will possible and desirable to cross correlate the observe measurements for each character between the data sets derived from each sensor. For example, in the present embodiment, vertical stroke widths can be measured bot optically and magnetically. When there is disparity between the two measurements, special criteria will be used to priorize the result, such criteria also being based on the character font and document format specifications. As another example, consider a print sample consisting of a magnetic ink font which contain non-magnetic extraneous ink splatters. The optical im will show the presence of the extraneous ink while the magnetic image will not. If the print specification allows non-magnetic extraneous ink, no fault will be issued.
The final result is then presented to an operator, for example, by means of computer video interface. The result may be condensed to a simple "Ye or "No" or may be accompanied with a detailed descripti of print faults.
While Figures IB and 1C are based on the mult element sensor of Figure 1A being a single slot magneti transducer and a linear optical array sensor, the proce of analysis is similar for other types of multi-element sensors as well. For example, there could be two multi element sensors a linear magnetic array sensor and a linear optical array sensor.
One type of printed character measurement device described for this invention is based on a syste using optical sensing. Other methods of shape detectio and measurement, including magnetic, thermal and acoust are equally feasible and will be readily apparent to those skilled in the art using the methods and devices described.
The optical measurement device consists of several functional blocks of equipment. These include: (a) a source of illumination directed at the area of the document sensed by the scanner. The wavelength of this illumination is chosen so as to maximize the probability of reliably detecting the printed information; (b) a lens capable of focussing the image of the area of the document to be scanned onto the active area of the sensor element; (c) a sensor controller which produces all the control signals required to operate the sensor in an optimum synchronized manner, with respect to both the measurement system and the illumination requirements; (d) a document transport and control system, which moves the document into the field of 5 view of the sensor, if necessary moves it past the sensor at a controlled rate while the sensor scans the document and then removes it from the field of view when completed;
10 (e) an analog-to-digital converter which changes the analog, time-varying signal from the sensor into a series of equivalent numerical values, representing the instantaneous value of the reflected light level at appropriate
15 measurement times. The brightness measurement resolution required by this device will be dependent upon the type of information to be detected. Information which is represented as a series of very
20 gradual shares will require very fine brightness measurement intervals. Information which is represented as stark black/white steps will be detectable with very coarse brightness intervals. The
25 output of this converter is equivalent to a digital representation of the source document brightness; (f) a conversion controller, which generates control parameters for the A/D converter,
30 causing it to operate in a manner which will ensure that all pertinent information on the document is properly digitized and available for subsequent analysis, with a minimum of extraneous, unwanted background information;
35 (g) a digital image format converter, which stores and reformats the image information from the A/D converter into a form suitable for retention in the digital image memory; (h) a digital image memory capable of storing 5 enough information to represent at least one full document image. This memory receives reflected brightness information from the scanning device using the scanner's time reference controller. It then produces the
10 stored information for the analysis devices in their time scale, upon command. It is in effect both a time compressor and expander; (i) a memory controller device, which operates the image memory under control of signals
15 from both the scanning device and the image analysis devices. This controller ensures that information from the scanner is stored at the appropriate place in memory at the correct time and that the information is
20 made available to the analysis devices as it is required by them. In addition, by signalling the document transport controller, it ensures that no document information will be lost, in effect
25 controlling the flow of source documents through the scanning device; (j) a device for detecting patterns of information in the image stored in memory, separating these patterns from the
30 background image and passing the pattern information and pattern location information to an information converter; (k) an information converter which accepts information patterns and locations from the
35 pattern detector and converts them, if necessary, to a form more suitable for use by the recognition logic only. For example, the shape of a character may be altered from an exact image form to one in which the 5 character is described as a set of predetermined features; (1) an information recognizer, which identifies the converted information patterns presented to it, using sets of pre-stored parameters
10 and recognition logic suited to the complexity of the patterns themselves. Patterns which cannot be recognized exactly are marked as unknown, and the recognizer may attempt a "best-guess" whenever the need
15 is present;
(m) a comparator device, which compares the recognized patterns against stored ideal measurements, producing a quality judgment factor which is a measurement of the
20 deviation of the original pattern detected on the doucment from the stored ideal pattern; (n) a set of decision logic which accepts the quality factor result from the comparator,
25 and dependent upon the invention of the measurement system, produces a measurement result ranging from "Yes/No" to a "percentage of acceptability"; (o) an optional computer, which provides a means
30 of automatically operating the measurement system unattended, storing measurement results for statistical purposes and presenting these results to an operator in a meaningful form. In Figures 2 and 3, there are shown two embodiments respectively of optical printed shape measurement devices. These embodiments are self- explanatory in view of the description already set out for the block components.
The present invention is a general purpose print quality analysis/measurement system which actually senses or detects the printed image through sensors having one or more transducers. The background material on which the image is printed can be either transported by the transducer or the transducer itself can be moved over the image to be sensed. Various types of backgroun material for the printed images are suitable, for example, paper, plastic and textiles. Various types of print can be analyzed, for example, bar codes, logos and alpha numeric using a variety of printing inks, for example, magnetic, optical, flourescent and embossed. The image can also be sensed using a transducer which ca scan the complete image with no mechanical movement of the image or transducer.
As previously stated, the transducer(s) can be of various types, including optical, magnetic and/or acoustic. A transducer may be used in the system alone if the transducer is an optical transducer or in combination with other transducers. The type of transducer to be used depends on the physical image attributes to be detected and analyzed. For example, t acoustic transducer may be used to sense and measure th raised print on an embossed plastic credit card. An optical transducer may be used to detect the print of t OCR and magnetic fonts. A magnetic transducer may be used to detect the magnetic image information in magnetically printed fonts such as E13B (a trade mark) and CMC-7 (a trade mark). A combination of optical and magnetic transducers for example could be used to detec the optical and magnetic print image and characteristi for automatic analysis by the print quality analysis a measurement system.
Single element transducers require that the printed image to be sensed and transducer be moved wit3 respect to each other such that the whole image is scanned by the transducer. The transducer thus output an analog waveform which represents the original Image. This analog waveform is typically amplified and filter and have conditioned by these various and other signal processing means to minimize distortion and remove interference from the detected signal representing the image. After conversion from analog waveform to digit the waveform is then sent to the PIRL and/or PIAL (Printed Image Analysis Logic) to analyze the specific quality characteristics of the waveform with respect t the ideal specified waveform stored for that particula image. The waveform thus represents a transformed version of the original image with some information lo but which is a lower cost version of the multiple elem transducer approach.
A linear or one-dimensional multiple element transducer produces a number of analog signals representing various points on the image when the transducer and image are moved with respect to each other. This process in effect produces a two-dimensio electronic "image" of the original printed image. The electronic "image" thus produced will have a certain spatial resolution depending upon the sampling density the transducer. The multiple transducer element analog signal samples are then amplified via multiple ampli ie and the signals are conditioned by various multiples of filters and minimum signal level thresholders before being converted to digital signals by the A/D converter The digital image representation is then analyzed and measured and compared to the image specification requirements stored in memory. The output from this process gives a measure of the quality and automated machine readability of the original printed image. Two-dimensional array transducers also produ a two-dimensional electronic image of the original printed image except that once the image to be scanned placed in front of the two-dimensional transducer arra and the scanning action is initiated, either manually automatically, no motion of the image to be scanned or the transducer is required since the two-dimensional transducer scans electronically.
This invention is primarily aimed at the analysis and measurement of a variety of types of prin images sensed by multiple element transducers to determine their quality and machine readability to a specific standard. The inputs from single element transducers are an aid to the multiple element transducers in this process. However, where the transducers are optical transducers, they can be singl or multiple element transducers and they can be used along or together with other transducers, within the scope of the attached claims.
One of the more common implementations of th invention will involve a combination of at least two different types of transducers. The two most likely combinations are; 1) a single element magnetic transdu and a multiple element one-dimensional linear array optical transducer, or 2) a multiple element magnetic transducer and a multiple element one-dimensional opti linear array.
The most common single transducer implementations of the invention will be 1) multiple element magnetic 2) multiple element one-dimensional optical 3) multiple element two-dimensional optical 4) single element magnetic, when accompanied by another multiple element transducer 5) single element optical.
The image qualification system of the present invention can be implemented using a combination of hardware, firmware and software. Essentially, referrin to Figure 1A, all the functions 1 to 4, inclusive, coul be implemented in hardware and the functions 5 to 8, inclusive, could be implemented in firmware in a programmable logic controller or in software in a personal computer. In the case where the quality analysis logic is implemented in software, the program algorithms and logic will be protected by copyright.
The system of the present invention has the ability to collect, accumulate, analyze and communicate the printed image quality statistics over a range of sample and/or time set by the system operator. A specific software module will perform this function in the personal computer.
Some specific applications of the present invention will be in the area of alpha-numeric and special symbols as used in the item processing industry. The qualification system of the present invention provides improved results whether multi-element sensors are used or a single element optical sensor is used.
|Cited Patent||Filing date||Publication date||Applicant||Title|
|US3764978 *||2 May 1972||9 Oct 1973||Optical Recognition Systems||Combined magnetic optical character reader|
|US3863219 *||9 Oct 1973||28 Jan 1975||Ibm||Data preprocessing system for character recognition systems|
|US3949363 *||28 Jun 1974||6 Apr 1976||Recognition Equipment, Incorporated||Bar-Code/MICR/OCR merge|
|US4205780 *||3 Jul 1978||3 Jun 1980||Teknekron, Inc.||Document processing system and method|
|US4315246 *||11 Jul 1979||9 Feb 1982||Magnetic Pheripherals, Inc.||Document character recognition system for identifying magnetic ink characters on bank checks and the like|
|US4356473 *||23 Sep 1980||26 Oct 1982||Gte Laboratories Incorporated||Monetary document profile location and predetermined selected path apparatus|
|US4490850 *||17 Dec 1981||25 Dec 1984||Ncr Corporation||Matrix character recognition|
|US4523330 *||23 Dec 1982||11 Jun 1985||Ncr Canada Ltd - Ncr Canada Ltee||Banking system and method|
|US4555617 *||23 Dec 1982||26 Nov 1985||Ncr Canada Ltd.-Ncr Canada Ltee||Concurrent, image-based, reject-re-entry repair system and method|
|US4685141 *||19 Dec 1983||4 Aug 1987||Ncr Canada Ltd - Ncr Canada Ltee||Method and system for finding image data associated with the monetary amount on financial documents|
|Citing Patent||Filing date||Publication date||Applicant||Title|
|WO1997039428A1 *||14 Apr 1997||23 Oct 1997||De La Rue International Limited||Document of value|
|DE4321177A1 *||25 Jun 1993||5 Jan 1995||Heidelberger Druckmasch Ag||Vorrichtung zur parallelen Bildinspektion und Farbregelung an einem Druckprodukt|
|EP0432616A2 *||4 Dec 1990||19 Jun 1991||Bull HN Information Systems Inc.||Page printer with machine-readable-character-based controls|
|EP0432616A3 *||4 Dec 1990||2 Sep 1992||Bull Hn Information Systems Inc.||Page printer with machine-readable-character-based controls|
|EP0474002A2 *||17 Aug 1991||11 Mar 1992||Wea Manufacturing Inc.||Print scanner|
|EP0474002A3 *||17 Aug 1991||14 Oct 1992||Wea Manufacturing Inc.||Print scanner|
|EP0590522A1 *||23 Sep 1993||6 Apr 1994||Sharp Kabushiki Kaisha||Thermal printer|
|EP1156459A3 *||14 Apr 1997||6 Feb 2002||De La Rue International Limited||Sorting documents of value|
|EP2220620A1 *||30 Nov 2007||25 Aug 2010||Hewlett-Packard Company||Method and system for measuring text-rendering quality|
|EP2220620A4 *||30 Nov 2007||20 Mar 2013||Hewlett Packard Development Co||Method and system for measuring text-rendering quality|
|US6155605 *||14 Apr 1997||5 Dec 2000||De La Rue International Limited||Document of value|
|US8594384||30 Nov 2007||26 Nov 2013||Hewlett-Packard Development Company, L.P.||Method and system for measuring text-rendering quality|
|International Classification||G07D7/12, G06K5/00, G07D7/04, G06K9/00, G07D7/20|
|Cooperative Classification||G07D7/12, G07D7/04, G06K5/00, G07D7/20, G06K9/036|
|European Classification||G06K9/03Q, G07D7/12, G07D7/04, G07D7/20, G06K5/00|
|24 Aug 1989||AL||Designated countries for regional patents|
Kind code of ref document: A1
Designated state(s): AT BE CH DE FR GB IT LU NL SE
|6 Aug 1990||WWE||Wipo information: entry into national phase|
Ref document number: 1989902599
Country of ref document: EP
|28 Nov 1990||WWP||Wipo information: published in national office|
Ref document number: 1989902599
Country of ref document: EP
|14 Apr 1996||WWR||Wipo information: refused in national office|
Ref document number: 1989902599
Country of ref document: EP
|9 Jul 1996||WWW||Wipo information: withdrawn in national office|
Ref document number: 1989902599
Country of ref document: EP