Image analysis – Image transformation or preprocessing – Image storage or retrieval
Reexamination Certificate
1999-08-23
2004-04-27
Mehta, Bhavesh M. (Department: 2625)
Image analysis
Image transformation or preprocessing
Image storage or retrieval
C382S232000, C358S003010
Reexamination Certificate
active
06728426
ABSTRACT:
FIELD OF THE INVENTION
The present invention relates generally to image processing, and specifically to efficient compression of varying image data appearing in a fixed template, such as characters written on a form.
BACKGROUND OF THE INVENTION
Businesses and organizations are increasingly turning to computerized storage of information filled in on documents such as paper forms, in addition to or in place of storing the forms themselves. After the form has been filled out, it is optically scanned and transferred to an electronic storage medium, such as a magnetic or optical disk. In order to ensure that the document is accurately reproduced when it is recalled from storage, it is necessary to scan the document at high resolution, typically on the order of 100 pixels/cm. To store an entire A4 page in this manner at eight bits/pixel, however, requires about 6 Mbytes of memory. Methods of image compression, as are known in the art, are applied to reduce the volume of data that must be recorded in order to reproduce a document.
A significant portion of the data in an image of a filled-in form corresponds to the form template, i.e., the lines, boxes, instructions, etc., that are preprinted on every form of a given type. Since this element of the form is predetermined and fixed, a useful way to reduce the volume of data that must be stored in recording the image of a filled-in form is to first remove the fixed template. Then the “fill image”—what remains of the form after removal of the template—is compressed and stored. Only a single image of the template needs to be stored, regardless of how many filled-in instances of a given form are to be stored and reproduced. To view the form as originally filled in, the template and fill images are recalled from memory and recombined. Another advantage of separating the fill image from the template is that it improves the accuracy of optical character recognition (OCR) programs used to interpret the characters in the fill image, which otherwise tends to be compromised by the presence of portions of the template near the characters.
U.S. Pat. Nos. 5,182,656, 5,204,756 and 5,793,887, which are incorporated herein by reference, describe methods for compressing and decompressing form images based on separation of the form template from the variable fill portion of the form. These patents are concerned mainly with improving the accuracy of registration of a pre-stored template image with the scanned document image that is to be compressed, so that the template can be more precisely removed from the image before compression and restored upon decompression. As is common in the art of document imaging, the methods of these patents address only binary images and do not relate to the distribution of gray levels in the actual document or in images captured thereof. Furthermore, these patents do not address problems that can arise when portions of the characters in the fill portion of the form overlap elements of the template. As a result of such overlap, when the template is removed from the document image, gaps may be left in some of the characters. Thus, improved image compression may be achieved, but at the expense of reduced readability in the stored images.
SUMMARY OF THE INVENTION
It is an object of the present invention to provide improved methods for compressing and decompressing images of filled-in form documents.
It is a further object of some aspects of the present invention to provide methods for improving readability of form document images following decompression.
In preferred embodiments of the present invention, a document scanner captures a gray-level image of a filled-in form document, based on a form template. An image of the template is removed from the captured image, as is known in the art. Groups of pixels in the image that remains having gray-level values below a given threshold are identified as belonging to a bi-level kernel of the image. The kernel thus defines the shapes of the characters that were filled into the form. A gray-level value is recorded for each of the pixels in the kernel, corresponding to the original value of the pixel in the captured image. A halo is defined surrounding the kernel, preferably by morphologically expanding the kernel by a predetermined amount in all directions, typically by a single pixel. For each of the pixels in the halo, the gray-level value from the captured image is recorded. Preferably, adjoining gray-level values are grouped together, and the gray-level resolution is reduced, most preferably to a user-selectable degree, so as to reduce the volume of gray-level data that must be stored. The gray-level data for both the kernel and halo pixels are then compressed, using any suitable algorithm known in the art. Both the kernel and the compressed gray-level data are stored in a memory.
Preferably, gray-level information in the image is used to close any gaps that may be left in characters in the kernel due to removal of the template, most preferably using methods described in European Patent Application 98480087.0, which is assigned to the assignee of the present patent application and incorporated herein by reference. Alternatively, the gaps are closed using methods based on distance transforms, as described in a U.S. Patent Application, filed on even date, entitled “Separating Text from Form Data in Document Images,” which is assigned to the assignee of the present patent application and incorporated herein by reference. It will be understood, however, that while such gap-closing is desirable, and may be carried out using any suitable method known in the art, it is not essential to the present invention.
In order to reconstruct the image, an image of the template and the bi-level kernel are recalled from memory and combined in proper registration. The stored gray-level data are then decompressed and assigned to the appropriate pixels in the kernel and the halo. Preferably, the image of the template includes gray-level data, as well. At points of overlap between the kernel and features of the template, such as lines and text, substitute gray-level values are preferably computed so as to blend with the adjacent points of the kernel and of the template features. Similarly, at points at which the halo overlaps a dark or gray portion of the template, the gray-level values assigned to the halo may be replaced by darker values in the template.
In this manner, an image is reconstructed of the filled-in form that is more faithful to the actual appearance of the original, and is therefore easier and more pleasing to read than binary images of the form as are known in the art. While the volume of data storage required for a typical compressed gray-level image, in accordance with preferred embodiments of the present invention, is greater than for the binary image alone, it is still far less than could be achieved by conventional methods of gray-scale image compression.
There is therefore provided, in accordance with a preferred embodiment of the present invention, a method for compressing an electronic image of an object including gray-level information, including:
processing the gray-level information to identify a kernel of pixels in the image having respective gray-level values darker than a given kernel threshold;
storing a locus of the pixels in the kernel in a memory; and
storing in the memory the gray-level values of the pixels in the kernel, in order to reconstruct a gray-level image of the object by applying the stored gray-level values to the pixels in the locus.
Preferably, the object includes a document, and processing the gray-level information includes identifying a kernel corresponding to information in the document. Most preferably, the document includes a form having a template printed thereon, and processing the gray-level information includes separating an image of the template from characters filled into the template using gray-level values of pixels corresponding to the template and to the characters. In a preferred embodiment, separating the image of the template from the char
Darby & Darby
Mehta Bhavesh M.
Patel Kanji
LandOfFree
Compression of form images in gray-level does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Compression of form images in gray-level, we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Compression of form images in gray-level will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-3210768