*
AMS Adviser *
Volume 3 Issue 5 - September / October 1998
Welcome to a new issue of the AMS Adviser.
AMS adds new Microfilm
equipment.
We continue with the fourth part of our series on the Dynamics of Cost in
Image Capture.
Canons MS400/500
Digital Reader Printers.
Also, take special note of our new services.
Plus we have all the usual bits (Funny
Bit? and AMS Services).
AMS
AMS adds new Microfilm equipment
 |
We have taken delivery of two High Speed Microimagers
to further supplement our expanding Bureau Services.
They are the Kodak Imagelink 30 and 70.
These machines are capable of producing high quality and rapid
throughput Microfilm.
Reduction ratios of 24, 40 and 50:1 are available. With both sides of
the document being able to be filmed at the higher resolutions.
Both Cameras offer film Blipping to aid in the automated retrieval of
images. The Imagelink 70 also offers tri-level blipping with document imprinting for
guaranteed results.
Contact AMS for a demonstration or for more information. |
The Dynamics of Cost in Image Capture (Part 4)
Strategies for Reducing the Cost of
Document Capture
For the balance of this white paper we will be presenting concrete
strategies for reducing the cost of document capture, piece by piece and second by second.
Not all of these strategies will apply to every installation, but by carefully choosing
the ones that can be implemented at your site you will be able to shave precious seconds
off your document capture process and thousands of dollars from your back room labour
expenses.
Use batch processing to speed up scanning
This is central to getting the maximum throughput from a document capture system. If
pages are fed manually and indexed on the spot, the actual throughput of even a fast
scanner can be as little as 5-10 ppm. In a batch operation, by contrast, entire batches
are fed into the scanner, then OCRed, indexed, and finally released. This "assembly
line" operation is far more efficient than manual feeding. Some arithmetic shows why:
If documents are fed manually, a reasonable estimate for the time to scan
and index one page with three index fields is:
- 5 seconds to scan
- 2 seconds to switch between scanning and indexing
- 12 seconds to index
The total time per document is 19 seconds and the total time for a
100-page batch is 1900 seconds (about 31 minutes). A single operator could finish 15
batches in an 8-hour day.
In a batch operation, a good estimate for feeding the same 100-page batch is:
- 30 seconds to load the scanner
- 150 seconds to scan (assuming a 40 ppm scanner)
- 30 seconds to load the batch at the index station
- 12 seconds to index each document, for a total of 1200 seconds for the batch
The total time for the batch is 1410 seconds (about 23 minutes), a savings of 8
minutes. Using a batch scanning model, an operator could finish 20 batches in an 8-hour
day.
At a rate of .28¢ per second, you’ve saved $1.37 for just this one batch. If you
process 100 batches per day, 260 days per year, this amounts to $35,620 per year.
The savings from OCR varies dramatically depending on how accurate the OCR is. To
decide if OCR is appropriate for a particular index field on a particular document, you
need to figure out whether it takes less time to type the index manually or to OCR it and
correct the OCR errors. Here’s how to do the analysis:
- First, figure out how much time it takes to manually index the field. For example, if an
index field averages 10 characters in length and your index operators can type 10,000
characters per hour (the most frequently cited industry average), it takes 3.6 seconds to
index the field. In real life, you should add about a half second per field, so figure the
total time would be about 4.1 seconds. For 100 documents, the total is 410 seconds.
- Next, figure out how accurate the OCR is on the specific documents you will be using.
The only way to do this is to perform tests on real pages, since OCR accuracy varies
widely depending on how clean the original documents are.
In this example, assume that we have good quality documents and the per-character
accuracy of the OCR engine is 97%. The next step is to figure out the per-field
accuracy. If each character has a 97% chance of being accurate, the chance of every single
character being accurate is .97 multiplied by itself 10 times, or 74%. Therefore, since
the chance of the entire field being correct is 74%, the chance of error is 26%.
- Finally, figure out how long it takes to check each OCR field and how long it takes to
correct OCR errors. The time to check is usually about 2 seconds, although testing might
provide a more reliable figure for your particular site. The correction time is 4.1
seconds (the same as in step 1), but this is only done for 26% of the fields. Therefore,
for 100 documents, the total time is (2 * 100) + (4.1 * 26), or 306 seconds.
In this scenario, OCR pays off, saving 104 seconds on a
100-document batch. However, OCR is extremely sensitive to both the character accuracy and
the length of the field. As a rule of thumb, if OCR accuracy is less than 95%, or if the
field is longer than about 20 characters, you are probably better off keying the field by
hand.
This series has been reproduced from a Kofax
white paper and will be continued over the next few issues.
As usual, the whole article is available for those impatient types. Just
contact AMS and it can be mailed, faxed or
e-mailed in full.
Go to Top.
Canon MS400/500 Digital Reader Printers
Click here for more information on the Canon
MS400/500.
Go to Top.
Funny Bit?
A Duck walks into a bar and says, "Can I have some Duckmeal?"
The Barman replies, "We don’t sell any Duckmeal."
So the Duck turns around and leaves.
The next day, the Duck walks back into the same bar and says to the Barman, "Can I
have some Duckmeal?"
The Barman says, "I told you yesterday, we don’t sell Duckmeal, we never have
and never will."
So the Duck leaves the bar.
The next day, the Duck walks into the same bar again and says to the Barman, "Can
I have some Duckmeal?"
The Barman says, "Look here Duck, if you ask me once more for Duckmeal I am going
to nail your Bill to the Bar!"
The Duck says, "Got any Nails."
The Barman Replies "No?"
The Duck says, "Can I have some Duckmeal?"
Go to Top.
AMS Services.
For the complete run down on what AMS can do for you, click on
the following link.
AMS Services
Go to Top.
Next Month
The next issue for Volume 3 will continue with part five of the article on the Dynamics
of Cost of Image Capture.
In the next few issues we will have some new articles which will include the following:
- An article covering some concerns on the long term storage of information on digital
media.
- A discussion on the merits of film or digital for storage of endangered books.
- The benefits of Aperture Cards.
- A look at Alchemy’s CAD2CD.
- Plus other topics of interest.
Plus all the usual bits & pieces.
Should you want a topic covered or need an article in full, please
feel free to contact AMS.
Go to Top.
Go to AMS HomePage.
Go to Top |
