Image analysis – Image segmentation
Reexamination Certificate
1999-12-09
2003-01-14
Wu, Jingge (Department: 2623)
Image analysis
Image segmentation
C382S209000, C382S228000, C382S172000, C358S464000, C707S793000
Reexamination Certificate
active
06507671
ABSTRACT:
FIELD OF THE INVENTION
This invention relates generally to the processing of images such as images resulting from reading documents by an optical scanner and more particularly to a method for separating a fixed part of the image i.e. the template from filled in information.
BACKGROUND ART
The deployment of computers in all aspects of everyday life, along with the dramatic increase of their direct and associated storage capacity combined with their ability to be interconnected, thus forming networks on which information easily circulates, is considerably influencing the way administrative tasks are carried out. Instead of handling tons of paper forms and documents, information is digitized and stored in the form of files in nonvolatile computer memories and, often, in huge dedicated storage units. Thereafter, the information is quickly retrieved when needed and is available in any place, having access to the network and the proper equipment to display it or print it, regardless of the location where information is actually stored. A typical example is a nationwide insurance company electronically storing all customer contracts such that every branch and agent can have access to any of them as needed.
Gathering information from individuals wherever they are at a particular moment, for example, the customer of an insurance company or the patient in a hospital, is generally considered to be a cumbersome and error prone type of work. If entered directly from a computer keyboard, the input of requested information is time consuming, generally requires a third party, such as an agent in a branch office or generally speaking an attendant, acting as a typist while the individual provides sometimes private and essential pieces of information without immediate feedback. Moreover, this way of interfacing computers does not permit convenient entry of anything but textlike information and does not permit entry of a signature that could somehow authenticate the input record.
Thus, a preferred mode of interfacing, made possible since the storage and processing capabilities of computers have dramatically improved, consists in having a form filled in directly by the individual providing the information who is assumed to be able to self check its content. The form or the document is then fed into a reading device, such as an optical scanner, which transforms the acquired information into machine readable code permanently stored for later and possibly repeated processing so as to extract and only retain the informative part of the form excluding the template. The template comprises all the fixed forms and text whose only purpose was to guide and instruct the individual filling in the form with the required information. The filled in information or variable part, which generally differs from one document to the other, has only to be associated with a particular form, stored only once, so as to be able to reconstruct the complete filled in form if that ever becomes necessary.
The above described mode of interfacing computers is also frequently used whenever archives are digitized so as to be stored in electronic storage means rather than being kept in conventional storage units, thus saving a considerable amount of space and granting to the archives all the advantages of an electronic document that becomes available simultaneously at many places and is easily retrievable. Whenever archives are filled in forms such as the documents resulting from a census, all of what is discussed here applies.
Because the scanning of forms and documents must be performed with a sufficiently high resolution so that no entered information is lost or is significantly altered, becoming unrecognizable, the reading devices usually produce a high volume of scan data. Although the size of computer memories increase each year, the amount of stored information has to be limited to a reasonable size to permit the permanent storing of documents, the total size of which may have to be expressed in units of millions of pages. This limitation becomes obvious for economic reasons and the practical limitations on the maximum size of the storing devices. Another important reason for which the amount of stored information must be drastically limited deals with the fact that, as mentioned earlier, the documents are generally made available over a network and must be transferred, upon request, to the end user, sometimes through communication links or virtual connections that would have an insufficient bandwidth to permit the transfer of excessively large documents to complete within an acceptable response time. As a typical example, a page of A4 size (297×210 mm), scanned at 100 pixels/cm, requires about 700 Kbytes of storage space. If it is transferred over a standard 64 kbits/sec communication channel this would take about 90 seconds. A time that is about two orders of magnitude higher than what is tolerable. Thus, algorithms and methods, known of the art, to compress data are generally applied which reduce the amount of raw data coming out from the scanner typically by at least one order of magnitude. The A4 document is thereby reduced to below 70 Kbytes. However, even before applying compression techniques, a very significant step towards the reduction of stored data is accomplished by removing the fixed part of the form and retaining only the unique entered information. The entered variable information accounts, typically, for only 10% of the scanned data, thus providing another order of magnitude reduction.
On top of being a significant contributor to the reduction of the amount of data to be stored, and to be transferred to the user, the removal of the template has a second very important objective. It is key to permitting subsequent flawless running of optical character recognition (OCR) software, aimed at interpreting the variable part of the form, so that the variable entered information contents can effectively be processed according to the purpose for which the forms have been designed.
Preventing the fixed part of a form from being stored can be achieved during the scanning process itself. One method for the elimination of the fixed template has been reported by D. E. Nielsen and al., in “Evaluation of Scanner Spectral Response for Insurance Industry Documents”, 16/A44 NCI Program, Working Paper No. 2. May 1973. This method, also known as “dropout ink” technique, is based on the idea to use a special color ink when printing a form that is transparent to conventional scanners. If a completed form of this type is scanned, the fixed pattern is invisible to the scanner, and only the variable part is captured. On top of being more expensive an obvious disadvantage of using special ink sensitive scanners is that it prevents the application of this approach to existing archives.
Thus, another approach for the separation of the form template background from the filled-in information has been disclosed in U.S. Pat. No.5,182,656 entitled “Method for Compressing and Decompressing Forms by Means of very large Symbol Matching”. According to this approach, empty forms, the fixed parts, are prescanned and the data obtained is digitized and stored in a computer memory to create a library of forms. The original filled in form is then scanned, the data obtained is digitized and the retrieved representation of the empty form is eventually subtracted, the difference being the digital representation of the variable part i.e. the filled in information. In order to perform such form elimination, it is necessary to precisely align the input form image with an image of the empty template. Even when the input form image is globally aligned with the template and there are no offset, skew or scale differences, there usually are local distortions that must be straightened out before the template can be dropped out. Such local distortions are frequently the result of inconsistent scanner performances or distortions from photocopying.
A method to compute fine registration in order to align the fixed part with respect to the variable part of an im
Kagan Oren
Zlotnick Aviad
Hesse Karl O.
Wu Jingge
LandOfFree
Method and system for dropping template from a filled in image does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Method and system for dropping template from a filled in image, we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Method and system for dropping template from a filled in image will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-3031528