Brazil’s banking industry association, FEBRABAN, caused a minor uproar earlier this year when it made the .PNG file its official legal format for scanned check images. While the move got little attention outside of its narrow industry niche, it had many banks complaining about operations nightmares and clamoring to get the decision reversed.
Why was it such a big deal? Well, on the surface it wasn’t – but digging deeper reveals an important milestone in the ongoing tug-of-war between image quality, file size, and bandwidth. FEBRABAN liked the higher resolution of the .PNG files; banks worried about the strain on their internal networks. Which side has the better point?
Well, to start with, it helps to know the answer to one really important question: How much file size and bandwidth are we talking about? And for a single check, it’s not much. The black-and-white images that are actually exchanged between banks, in Brazil or any other country, measure about 15-20 KB per side, or less than 40 KB in total.
With a high-resolution grayscale like the .PNGs that Brazil has made official, that jumps to 150-200 KB per side – a tenfold increase to be sure, but either way, we’re still talking about kilobytes, while Internet bandwidth is measured in megabits and storage in terabytes. For some perspective, the U.S. uses nearly as many checks as the rest of the world combined – yet you could comfortably fit all the black-and-white images scanned throughout the entire country each day onto a single 8-terabyte hard drive that retails for about $300.
But (File) Size Does Matter
Where problems start to come up, though, are with behind-the-scenes bandwidth. As we mentioned, all the images actually exchanged between Brazilian banks are of the sub-40K black-and-white variety. The higher-resolution .PNG versions (which have the official legal standing) are preserved offline.
But – and this is a big BUT – even though the .PNG files aren’t used for clearing, they still have to be moved. They can’t just sit at the individual branches where they were originally captured. So each bank needs to transmit the larger .PNGs to a central storage location and THAT is where they start causing trouble.
The first bottleneck that banks run into is that many of their internal networks were built under the assumption that they’d only be transmitting transaction data of, at most, a few kilobytes at a time. Second, many branches are in places where broadband is not prevalent, and therefore rely on dial-up or satellite uplinks.
Now the math gets much less appealing: If a branch with a low-speed connection took in 1,000 checks in a day, the front-and-back grayscales would weigh in at around 200MB. On a 56-kilobit connection, that would take 7.9 hours to transmit, or in other words, all day. So we can see why there’s a problem.
But wait, there’s more! You might be thinking that surely FEBRABAN didn’t just decide on the larger file format for no reason, and you’d be right. The rationale for keeping the bigger file was that Brazilian checks often have other documents that need to go along with them – most notably the boleto, a type of scannable invoice that’s used in bill pay applications there. So while black-and-white images might be acceptable for a scanned check, the higher resolution is necessary for more complicated full-page documents. There may be truth to that, but it adds to the misery of any office with a slow connection by piling on even more documents and increasing the network load even further.
A Black-and-White Issue
Now that we’ve spent all this time explaining why larger files cause problems for banks, let’s talk about why they’re actually better to have if you can keep up with the network demands.
We said earlier that banks clear checks by exchanging black-and-white images – but a more accurate description would be black-OR-white images. Each check is initially scanned as a high-resolution grayscale, then converted into an image where every pixel is literally 100% black or 100% white. These “bi-tonal” images, as they’re called, achieve their purpose of reducing file size – but that comes at the expense of a LOT of readability.
So in order to shrink the file for transmission, we have to “dumb down” the image by making a decision to turn each pixel “on” or “off,” while throwing away the underlying information. Any shades in between are simply forced to one extreme or the other.
That can create big problems when you encounter, for example, checks with faint printing, vivid backgrounds, or overlapping print. The cameras and software do the best they can to convert the image faithfully, but when you only have two choices for each pixel, it’s easy to get trapped with a bad image. Even though this happens only infrequently, it’s a problem that costs banks millions of dollars in time, effort, and fees each year to deal with rejected items.
The obvious solution, then, is to try and clean up the image while you still have the grayscale to work with – and that’s exactly what scanner manufacturers like ourselves try to do. We’ll try and even out the background interference by adjusting contrast and brightness levels, making several different images, and intelligently selecting the best. This all happens more or less immediately; the scanner operator doesn’t even see it.
But the catch is that you need the original grayscale image to work with – if you’ve already thrown out the shading information and converted it to black-and-white, there’s nothing there to interpret.
That can come into play two ways: One, if there’s a question about a check after the initial clearing phase, and you have to go back and pull the original image out of your archives. Most banks hang on to grayscales for a few days but quickly clear them out to save space. So once a check is archived, it’s very difficult to turn back the clock.
Second, in some situations, such as ATM deposits, the grayscale may be discarded immediately after the customer finishes the transaction – or in the case of remote deposit capture, the bank may not receive a grayscale at all. Since these situations represent a growing percentage of deposits, more and more transactions are happening with no fallback.
So, Should .PNG Be the Thing?
If we were considering accuracy alone, retaining high-resolution grayscales would be almost a no-brainer choice. As a scanner manufacturer, we’d certainly prefer it.
Would it actually work, though? Well, as Brazil’s struggles so far have shown, it’s great in theory, but tricky in practice. However, it stands to pay off big-time if they can last it out.
Back when check scanning started in the United States, there was a vigorous debate over whether the official standard should be grayscale (for the image quality) or bi-tonal (for the file size). In fact, many people thought that grayscale was so superior that it was going to win out even despite its larger file size – this was in 2003, but even then, word was the networks probably could’ve handled it. (Keep in mind, there were also twice as many checks written then as there are now).
Fast-forward to 2016, with broadband near-universal and Gigabit Ethernet everywhere, and there’s no doubt that the infrastructure in the U.S. could handle grayscales. Will we ever switch, though? Probably not, because that would require rewriting a lot of software and reorganizing the workflow at every bank in the country. Once a system is in place, going back and making fundamental changes to it is extremely difficult – so much so that they probably won’t be made unless the new way is better by leaps and bounds. (For the same reason, even though UV document security has been introduced in other parts of the world, it is a long shot to ever make it to the U.S.)
So for this reason, we think that Brazilian banks will be glad later if they stick with it, and ride out the challenges of grayscale today. Some day in the near future, networks will be able to handle the traffic with room to spare, and the end result will be a system that is superior in many respects to our own.
However, since we’re not the ones facing the issues ourselves, it’s easy to see why it may be a plan that looks good on paper but is much harder in practice.