`
`
`
` Exhibit 2
`Exhibit 2
`
`USAA Ex 2024-p.1
`Wells Fargo v. USAA
`CBM2019-00004
`
`USAA Ex 2024-p.1
`Wells Fargo v. USAA
`CBM2019-00004
`
`

`

`White Paper
`
`Check 21
`
`Controlling Image Quality
`at the Point of Capture
`by
`Harvey Spencer
`
`Digital Check Corporation
`466 Central Avenue
`Northfield, IL.
`
`© Digital Check Corporation & HSA 2004
`
`USAA Ex 2024-p.2
`Wells Fargo v. USAA
`CBM2019-00004
`
`

`

`Check 21 -- Controlling Image Quality at the Point of Capture: Page 1
`
`Why is it important to ensure that an acceptable image of a check is captured at the first
`point of entry to the check clearing process under Check 21 legislation?
` The answer is simple -- if the image quality is not acceptable, the truncating bank negates the
`benefits of image conversion. Either the check must be re-imaged and reprocessed or the physical
`paper check must be submitted. If it is impossible to create a quality image and the paper check is
`not available, the truncating bank will have to pick up the liability.
`Check 21 Image Requirements
`The Check Processing for the 21st Century Act (aka Check 21) was passed by Congress in
`October 2003. On October 28, 2004, its provisions become active. This means that legally a bank
`may convert physical checks to digital image representations and pass those images through the
`clearing system. As a result, it is likely that the bank’s customers will not be able to get their original
`checks back any more.
`Check 21 law does not mandate use of images. All it does is permit the legal use of ‘substitute
`checks’ as legal drafts if they are created according some simple rules from an image of the original.
`This has the effect of increasing interest in converting a paper check to a digital image representation
`and then using that check image for clearing and movement instead of the paper. The benefits are
`major improvements in the process, that can be achieved at lower cost for all parties, than are cur-
`rently possible. In the case of the collecting bank, some of the areas of saving include reduction or
`elimination of traditional proof/encode operations, later cut-off times, elimination or reduction of
`courier costs and no sort by routing numbers. In the case of clearing operations, the costs of clearing
`electronic images is likely to be a great deal less than processing paper.
`
`Paper Image prop’d Fed 10pm fees
`2004 fees
`image fwd
`sub check fwd
`0.5 - 8 cents
`2 - 7 cents
`4-13 cents
`City
`0.5 - 34 cents
`2 - 7 cents
`4-13 cents
`RCPC (country)
`$2.00 - $37
`18 - 42 cents
`30 - 80 cents
`Cash letter
`*Source: Federal Reserve
`
`In the case of the paying bank, there is no need to fine sort the items or return the physical
`paper.
`As a result it is expected that over the next few years, all 40 billion odd checks that are written
`in the United States will be converted to image.
` The Federal Reserve has dictated one key piece of image management -- that they will only
`accept bitonal (black and white) images at 200 or 240 dots-per-inch resolution compressed using a
`group/4 modified huffman algorithm. It makes sense as current back office check processing equip-
`ment creates these sorts of images, they are commonly used in the document imaging business and
`they use a small amount of bandwidth (typically 15K bytes for each side of the check). But it also
`means that challenging backgrounds on personal checks have the ability to wreak havoc with the
`image quality if the conversion (thresholding) from grayscale or color originals is not carried out
`sufficiently well.
`
`© Digital Check Corporation & HSA 2004
`
`USAA Ex 2024-p.3
`Wells Fargo v. USAA
`CBM2019-00004
`
`

`

`Defining Image Quality
`
`Check 21 -- Controlling Image Quality at the Point of Capture: Page 2
`
`Figure 1: Good quality image
`
`We all know quality when we see it, but what does it
`mean? Often we define quality in terms of negatives -- if you
`cannot read it, then clearly it is not a quality image -- but what if
`you can read most of it, but it is not usable? Figures 1 and 2 are
`clear examples of a good quality and bad quality image. What
`is an acceptable image to one person may be unacceptable to
`another. The FSTC committee on image quality defined quality
`as “The totality of characteristics of an image that bear on its
`ability to satisfy stated or implied image needs” -- figure 3
`shows a hierarchy of image quality needs.
`Checks consist of a defined format regulated through
`standards developed many years ago. The format consists of a
`limited number of fields. On the front these include the account
`and bank number on which the check is drawn, date, payee,
`amount to be paid and approval signature. These fields are
`located in predefined areas so that the checks can be quickly
`and effectively processed -- the account number, routing code
`and check number are printed on the bottom of the check using
`an E13B font designed to be read magnetically at high speed. The back of the check often contains
`endorsement information showing where the liability occurs.
`Quality is a subjective matter dealing with legibility and acceptable legibility may depend on
`how easily automated ‘reco(gnition)’ sytems can read the data. In order to have sufficient quality to
`process the check, it must be readable at minimum as well as the physical paper checks that are
`currently passed through the clearing process to the paying bank. But under Check 21 law, it is likely
`that only images will be passed through the clearing system -- so the physical paper may not be
`available. In this case the image must then be of sufficient readability to be used to pay the funds or
`the paying bank may refuse payment (see Figure 3).
`
`Figure 2: Poor quality image
`
`No Data
`
`Figure 3
`
`Insufficient
` Data to be
`Valid
`
`Unacceptably
`Low
`Confidence
`by Reco
`
`Minimally
`Decipherable
`by Reco
`
`Decipherable
`by Reco with
`Acceptable
`Confidence
`
`Decipherable
`by Reco with
`Highest
`Confidence
`
`Minimally
`Decipherable
`by a person
`
`Easily
`Decipherable
`by a person
`
`Easily
`Readable
`Text
`
`Source: Frank Jaffe, FSTC
`
`Who has the liability for not delivering a legible / usable
`check image?
`The liability for image quality of the check image lies with the truncating organization and
`remains there even though other liabilities shift once the correspondent or paying bank has accepted
`the check.
`
`© Digital Check Corporation & HSA 2004
`
`USAA Ex 2024-p.4
`Wells Fargo v. USAA
`CBM2019-00004
`
`

`

`Check 21 -- Controlling Image Quality at the Point of Capture: Page 3
`
`So the truncating bank has to ensure that it creates and sends a quality image on to their core-
`spondent or clearing house since if the paying bank cannot read the check, it will not pay it. At this
`stage, the truncating bank will either have to go back to the original paper (if it has it) and pass that
`through the system or recreate the bitonal image from a stored higher quality image (color or
`grayscale) and resubmit. Either way it will be expensive to manage a check image returned for poor
`quality reasons. To keep this cost to a manageable amount, each organization accepting a check
`image preferably needs to ensure that they are receiving a quality image.
`The ANSI Standard
`Check image interchange formats and supporting file formats have been defined by ANSI X9.
`The image formats are a part of X9.90 while the supporting electronic cash letter is X9.37. Con-
`tained within the X9.37 standard are a set of fields designed to tell the receiving organization
`whether each item is a quality image or not. These test flags essentially say you may have a problem
`because the image is damaged, it is too light, too dark, too noisy etc.. The problem then became
`what does ‘too’ mean.
`From the bank’s standpoint it mostly means unpayable. From the bank’s customer’s standpoint
`it may mean something else depending on the information they require. From other people’s stand-
`point (e.g. check guarantee companies) it may mean something very different.
`FSTC Input
`A sub committee associated with FSTC (Financial Services Technology Corporation) which
`consists of vendors, banks and supporting organizations was set up to define measurement criteria
`for quality. It has just issued its Phase 1 report.
`What are the elements involved?
`FSTC has identified 16 metrics to ensure overall image quality. These are designed to enable
`‘rapid low cost measurement of image characteristics to determine the probability that a check image
`will be usable’. They are not designed to perform a more subjective analysis of the data elements
`within the check and that may effect its usability as a payment instrument. Examples are whether the
`MICR line could be read magnetically or whether the amount could be automatically recognized.
`For this type of field specific analysis, users will have to employ additional software tools beyond the
`FSTC recognized ones.
`The 16 different defect metrics identified are as follows:-
`1. Undersized Image
`2.
`Folded or Torn Document Corners
`3.
`Folded or Torn Document Edges
`4. Document Framing Error
`5.
`Excessive Document Skew
`6. Oversized Image
`7.
`Piggyback Document
`8.
`Image Too Light
`9.
`Image Too Dark
`10. Horizontal Streaks Present in Image
`
`© Digital Check Corporation & HSA 2004
`
`USAA Ex 2024-p.5
`Wells Fargo v. USAA
`CBM2019-00004
`
`

`

`Check 21 -- Controlling Image Quality at the Point of Capture: Page 4
`
`11. Below Minimum Compressed Image Size
`12. Above Maximum Compressed Image Size
`13. Excessive Spot Noise in Image
`14. Front-Rear Image Dimension Mismatch
`15. Carbon Strip Detected
`16.
`Image Out-of-Focus
`These are the basic measurement metrics identified to ensure that the check image may be
`successfully cleared, but each vendor has the ability to add more measurements.
`Each metric has a definition, a measurement and units used, criteria to identify whether the
`defect is present or not and possible sources (the complete document may be downloaded from the
`FSTC site -- www.fstc.org). Pass/fail criteria are set as standard measurements to be incorporated in
`the ANSI X9.37 standard and FSTC is asking ANSI to expand X9.37 to allow passing of the quality
`values. The intent is that the accepting organization can quickly evaluate an image for acceptability,
`probably in conjunction with the amount of the check and other factors.
`In order to assist with this, a second phase FSTC sponsored project is being planned to assess a
`value for each of the metrics, transforming them in combination using empirical analysis to create a
`single minimum usage quality flag. The intent is that this flag can then be quickly tested by each
`user of the images for acceptability.
`Create Quality Images as Early as Possible
`“To realize the full promise of Check 21, paper checks should be truncated as early as possible
`in the payment cycle,” says Fred Herr, Senior Vice President, Retail Payments Office, Federal Re-
`serve Bank of Atlanta. ”Everyone will benefit as organizations expand their use of image technology
`and more institutions are able to receive electronic image files rather than paper checks.”
`Capturing at the Point of First Presentation.
`The earliest stage in the payment cycle is at the point of first
`presentation and the key to reducing liability and realizing substan-
`tial cost savings is to assure a high quality image is sent on to the
`next stage.
`This means that the best place to create a quality image is while
`you still have access to the paper (i.e. the teller station if practicable,
`the corporate AR department or the retail store). This way, each
`check image (item by item quality assurance) can be analyzed and
`improved while the paper is available -- even while a customer is
`depositing a payment or paying a bill. Any problems involving the
`physical paper can be identified and fixed immediately. The key is
`to maintain levels of customer service and not to keep the customer
`
`“To realize the full promise
`of Check 21, paper checks
`should be truncated as
`early as possible in the
`payment cycle,” says Fred
`Herr, Senior Vice President,
`Retail Payments Office,
`Federal Reserve Bank of
`Atlanta.
`
`waiting any longer than normal.
`An alternative is to scan batches of items at the back counter or in the back office. Batches are
`most effective when used for capturing large numbers of check items in a deposit or where multiple
`deposits are grouped together. Traditionally batch scanning has been carried out in the back office on
`high speed scanners. A high speed back office sorter has approximately 20 milliseconds to make a
`quality decision and, if the check image is poor, will outsort it into a reject pocket for exception
`
`© Digital Check Corporation & HSA 2004
`
`USAA Ex 2024-p.6
`Wells Fargo v. USAA
`CBM2019-00004
`
`

`

`Check 21 -- Controlling Image Quality at the Point of Capture: Page 5
`
`Figure 4: Original Color Image
`
`Figure 5: Bitonal thresholded image
`
`processing. This only allows for basic global quality
`checks to be performed. The outsorted exception items,
`which usually represent a small minority of the total,
`are then handled manually and scanned separately.
`Although the time scale and the batch volumes may
`vary, this is the way that any batch processing operation
`must work -- whether at the counter, at back counter or
`in the back office.
`In small volume environments such as at the teller,
`a batch operation is not an efficient way to work.
`Batch processing involves analyzing each check within
`the batch and attempting to fix it or out sorting the
`offending items and re-imaging. If the image has
`already been converted into black and white
`(thresholding), then the possibilities of fixing it elec-
`tronically are very limited. Even if the user has access
`to the grayscale or color, it will probably have been
`compressed using a lossy compression such as JPEG
`which can make fixing it difficult. Using higher level
`intelligence such as CAR/LAR automated handwritten amount recognition, or segmented analysis
`and intelligent thresholding becomes difficult because the of times involved.
`Front counter scanning with an item by item QC process solves the problem. But it works best
`when the total number of deposited items is low -- usually 5 or less items. Commercial deposits with
`a much higher volume of items are better handled at the back counter or the back office.
`Analyzing the Image at the Point of Capture
`At the front counter the teller can accept the items and convert them
`without significant time impact. The teller’s scanner can output a raw
`grayscale or color image (see figure 4) to the PC and work on it from there.
`Using today’s PC’s connected directly to the teller scanners, allows over 500
`milliseconds to be available for work on each check item without impacting
`on performance or causing the bank’s customer to wait longer in line -- over
`20 times more time than a back office sorter has. The processor power
`available today on a PC at the counter or in a commercial environment
`allows for reiterative multiple thresholding (conversion from color or
`grayscale into black and white -- see figure 5) with reference to recognition
`engine results in order to produce the best quality image and minimize teller
`keying amounts. It also has reference to files such as positive pay in order to detect fraud while the
`customer is in front of the teller. And if the truncating bank wants it, the original grayscale or color
`image can be maintained for reference -- available in case of queries as well as the black and white
`image.
`Exception items that cannot be truncated, such as checks in carriers can be caught immediately.
`In the case of corporate truncation, the receiving bank can be assured of a quality image that will
`clear without problems.
`
`Truncating at the
`teller allows 20x more
`time for analysis,
`optimal image quality
`adjustment, and fraud
`detection.
`
`© Digital Check Corporation & HSA 2004
`
`USAA Ex 2024-p.7
`Wells Fargo v. USAA
`CBM2019-00004
`
`

`

`Check 21 -- Controlling Image Quality at the Point of Capture: Page 6
`
`Conclusions
`The truncating organization has responsibility for image quality and:-
`• The liability for creating a poor quality image lies with the truncating bank.
`• Poor quality images will result in slower payment and higher costs and the depositor or depositing
`bank may have to find and submit the original paper check.
`• Truncating the checks at the point of first presentation in many cases keeps the size of each trans-
`action low. Low item transaction volumes allows time for the creation of a guaranteed good
`quality image at the time of truncation. This eliminates the need for outsorting and exception
`handling as a separate process.
`• Higher volumes need a batch type operation, which will require a separate exception management
`activity.
`• Under Check 21, successful scanning systems will be able to handle both image-by-image QC for
`front office and low volume operations in conjunction with batch type processes for higher vol-
`ume operations.
`About Digital Check Corporation
`Digital Check Corporation is the leading manufacturer of countertop check scanners in the
`world today and has the largest range of scanners for this market. Digital Check’s ‘TellerScan’ brand
`check scanners, which are manufactured in the US and Europe, are used in the world’s largest teller
`scanner automation systems and have been used by many of the world’s top banks. Digital Check
`Corporation is a privately owned corporation located in Northfield, IL (tel: 847-446-2285).
`Digital Check’s scanners are interfaced through a common API which supports item by item
`quality assurance as well as high speed batch scanning using the same software. For more informa-
`tion please visit www.digitalcheck.com.
`
`© Digital Check Corporation & HSA 2004
`
`USAA Ex 2024-p.8
`Wells Fargo v. USAA
`CBM2019-00004
`
`