Into electronic management of documents
usually the attention is given to issues related to the themes of
storage, of information
retrieval, of security, of
sharing,
of workflow, ecc...undervalue, instead,
an important aspect of the image processing.
Because the key element of electronic management of documents,
is the document, should give the specific attenction so that
hers rappresentation as digital image
is most natural and more similar to the original paper,
so as not to regret the fact that they do not get your hands on a physical document,
but only to have a virtual document on a monitor.
The evolution technology, introduced from
continuing research in the field of images processing and the continuing
increase of the capacity of
elaboration of the modern personal
computer now make it possible to operate
on digital images so as extremely effective,
impossibile only a few years ago.
We see in practice how is
possible to ensure that your computer can
be of help in this task,
analyzing nearly the technologies
most useful in document field of image processing.
Black Border Removal and Dynamic Thresholding applied to a page of a book acquired with a planetary scanner: on the left the original image on the right that processed
Deskew
For the acquisition of the documents often
use professional scanners with a automatic document feeder (ADF).
For the inevitable mechanical tolerances
and the low stiffness of the paper it is normal
that the images of document are slightly tilted to the right or left instead
of perfectly straight and aligned with the horizontal and vertical axes.
The technical term commonly used to refer to this inclination is called skew, which literally means slope, which usually does not exceed + / - 5 °.
With the deskew can automatically straighten a inclined picture: on the left the original image on the right that processed
The correction operation of the slope, which is the straightening of the image, is known, then, deskew. This correction can now be executed in total autonomy by software without human intervention: the image processing subsystem responsible for this activity analyzes the image for evaluating the entity of the angle of inclination so that it can appropriately rotate it to make it perfectly straight, correcting the defect introduced by the scanning. The most effective tool available today are able to contain the processing time in fractions of a second and to get really high accuracy,
reaching even to 1/100 of a degree, simultaneously being able to work not only on monochrome images but also images in grayscale and color.
Black Border Removal
Some types of scanners with high productivity using a black background instead of white, for two main reasons: to avoid problems of transparency acquiring very thin sheets and permit the simultaneous acquisition of sheets of different sizes. The image of the scanned document can then to present thick borders made up of bands blackened. The process of identification and elimination of these boards is called a black border removal, and can be performed automatically by a software process that, after analyzing the image, it can cut off the black part additional four sides of the sheet or can simply
turn white, based on the settings you choose. Also in this case it is not necessary intervention of the operator,
the processing times are almost instantaneous and can act indifferently on images in black and white, gray-scale, and color.
Black Border Removal, Deskew and Dynamic Thresholding applied to a page scanned from microfilm: on the left the original image, on the right the one on trial
Despeckle
Sometimes it happens that the acquired images present dirt consists of isolated points blacks,
ue to the dust deposited on the sensors capture or even to electrostatic interference due to overheating of electronic components.
The english term used to identify this type of dirt is the word speckle, and the cleaning operation for the identification and removal of these points is defined
despeckle. The software tools that do this type of processing do not require human intervention, are extremely fast and can work on monochrome, gray-scale or color.
The most effective ones can employ advanced algorithms such that, although eliminating elements of very large dirt, prevent that punctuation marks or dots on the "i" can be removed accidentally.
Auto-orientation
Not all paper documents are written in "portrait", or thinking of a sheet of A4 paper, with the natural orientation for reading such that the narrow part is the basis of the paper. Sometimes it happens that graphs, tables, cards and elaborated similar to be printed in "landscape", that is, always referring to a sheet of A4 paper, with the wide part of the paper as a base. When you scan documents, it is always in the same orientation for which we will be left with rotated images correctly for reading and others do not.
To prevent that, during the next consultation document, the user must repeatedly use rotation functions to be able to correctly read the full document, have been specifically developed auto-orientation that allow you to automatically perform this operation in batch, on a group of documents, without human intervention. These tools allow then to identify the correct orientation of reading and decide for yourself if the document is straight or whether to automatically rotate images 90, 270 or 180 degrees.
Dynamic Thresholding
Because the electronic management of documents is essentially addressed to management of business documents, usually the color is not important, so it is oriented to the acquisition of the documents in monochrome mode. The scanners, even those monochromatic, acquiring a document making in fact a measurement of the amount of light reflected from the surface of the sheet, obtaining by electronic circuits usually a value between 0 and 255. Having to return an image in grayscale, is returned precisely this value, but having to return a monochrome image, it must be set a threshold value, below which the point is considered black, above is considered white.
The selection of this threshold is very delicate operation because according to it you can get readable documents or too light or too dark. If an image is too light or too dark, it could be difficult to read the data contained in it, which was quite serious if you do not notice the problem immediately, but only when it is too late, that is when you need the document and the copy original is no longer available.
The major problem, however, is given by the fact that, more often than not just imagine, there is no possibility that a threshold value, whatever it is, makes the image readable in all its parts.
It is to solve these problems that were born the tools of dynamic thresholding. Practically instead use a threshold value valid for the whole image, is used a threshold value specific for each image point, calculated on the basis of different algorithms, more or less advanced. With this system it is possible to avoid losing written extremely bright light background, or to turn into black spots lettering on dark background very dark.
Probabily this technology is a more complex and the most problematic is identify both the type of dynamic thresholding
more appropriate at your documents, and the any required parameters.
However, the latest in this
field allow you to not have to set
any parameter, so that both
the same system to analyze the image
and decide which parameters are
more appropriate to process it in
the best way.
Color dropping
Not infrequently it happens that you need to scan a document in color to have a version similar to the
original for the long term preservation, but at the same time can serve as a monochrome copy for data capture, or data extraction.
An example is the medical prescriptions, or the forms optimized for data capture, which consists of a preprinted part of a single color
such as red or green, chosen to be removed directly during the monochrome scanning.
In these cases, a technology that can be useful is the color dropping, namely the elimination of a specific range of colors
from the image.
Basically what happens is emulated using a scanner equipped with a colored lamp for filtering of color,
starting from the image in color and then coming to monochrome.
In this way, also acquiring color, you may later obtain monochrome images sorted ready to be subjected to data capture.
Color Dropping executed on two tone (red and purple): the original image on the left, on the right that processed
Compression
Even in the field of compression you can take advantage of recent developments in image processing.
For example, you can make sure that the system recognizes automatically if a worksheet contains photographs or graphics colored that it is preferable to store it in color,
perhaps with JPEG compression, or if there is an appropriate representation in monochrome and his rescue CCITT G4 compression.
Not only that you can also be able to get a highly compressed file splitting, again thanks to image processing tool,
the layer of black text on a white background from color, and save files in the compound, such as PDF,
with each layer compression and sampling more appropriate, that is perceptible without any loss of quality to the user.
Resampling
The resampling is to transform an image from one resolution to another.
This may be required, for example, when scanning documents at a high resolution for long term storage, but you also want to get a more "light"
for the current usage, for distribution on the web, etc ...Using appropriate tool is therefore possible, for example,
resample images acquired at 200 DPI to 300 dPI or 400, without the need of having to reacquire.
You can even make sure that, by sampling the low resolution images to create thumbnails or previews,
are transformed into gray-scale so as to make them more readable (antialiasing).
The user interface of a popular application for batch image processing (Recogniform Image Processor) with a wizard system, you can define all the operations to be performed on the batch of images, without the need to develop specific software
Conclusion
This technologies are enabled both SDK
(Software Development
Kit) for developer that wish
integrate directly in own software of electronic management of documents,
and with application ready to use,
suitable for batch processing
documents, for end user.
Concettualy the image processing software can be seen as
"balck box" able to process
images as shown in
given, however, without the need
to learn and develop
the complex algorithms that are to
the basis of its functionality.
Therefore the developer can consider the integration for
enhance the functionality
of software, as well as
directly to end users.
The service center, for example, can use this SDK software whithout any
investment in software development.
It is therefore right to say that, considering
the undoubted efficacy, ease of use
and low cost of these technologies,
it would be a shame not to use them.