by Glenn Spitz

It seems that those more radical changes will not be proposed as an international standard, or if they are at some later point they will be introduced in a new standard, perhaps to be called iQR, but I have no new information about that.

I think that the most radical of the changes that will be proposed include some more efficient ways to include common preambles to urls, such as http:// and that alone is enough to render existing QR Code reading apps incompatible with new symbols. So it will be interesting to see if this proposal does indeed go forward and if it will be approved and adopted.

A less radical change to the document (which is ISO/IEC 18004 by the way) is the change its title back to QR Code Symbology Specification, from QR Code 2005 Symbology Specification.  That change was originally made in 2005 because the new “symbology” contained incompatible additions, such as Micro QR Code and some other things that would render existing readers not fully compatible with the new “QR Code”. Changing the name of the symbology was an efficient way of making it clear existing readers were compatible with QR Code (the original) and not the new formats included as part of QR Code 2005.  This same conundrum will appear now and it may be too clever by half to change the symbology from the old QR Code 2005 to the “new” QR Code. (Never mind any possible confusion between QR Code before 2005 and QR Code now).  Anyway, all ISO documents, including 18004 are dated with the publication date so it is possible to unambiguously specify which version of the ISO specification you are compliant with. It is just not so easy with the title of the document, or the symbology name.

These kinds of issues with compatibility come up often when considering revisions to ISO documents.  In general, we attempt to minimize incompatibility and disruption to existing technology, which at the same time minimizes the scope of changes that we are comfortable with.  This can indeed prevent some improvements from being made.  If such changes are compelling and needed they can be specified separately in a new standard.  It becomes a balancing act to choose the best way forward with such improvements.  I’d appreciate your comments and feedback on this issue, specificaly with regard to improvements to QR Code.  Does the “soft” nature of QR Reader apps suggest a different approach to incompatibility and redefine “disruption” out of the equation?

by Glenn Spitz

QR Code Grading Parameters

QR Code Introduction for UPC Printers

The process of grading a 2D matrix symbol according to ISO/IEC 15415 begins with the process of decoding it according to the symbology specification (reference decode algorithm). Actually, it begins by blurring and thresholding the image, and then running the reference decode algorithm, but essentially, it is dependent upon the reference decode algorithm for creating a grid of module centers before proceeding to the grading itself).

The rest of ISO/IEC 15415 consists of evaluating parameters based on the reflectance values throughout the symbol, most of the time at the module center locations.  For example, MODULATION (and RM) is graded based on the reflectance levels of each of the data modules within the blurred (reference) image at the module centers.

So, the refrain is: “if only the reference decode algorithm would succeed on dot printed symbols (the term I will use to refer to matrix symbols that are printed with a dot for each module that is not necessarily connected to neighboring dots), then the symbol would pass the grading criteria of ISO/IEC 15415.”  In this way, it is not ISO/IEC 15415 that is biased against dot printed symbols.  However, the symbology specifications clearly are biased against dot printed symbols.

We introduced the stick algorithm into AIM DPM which addresses this problem directly.  We also introduced many other grading based modifications into AIM DPM. The stick algorithm would address the issue of non-connected dots for symbols whether those symbols are dot-peened or otherwise “called” DPM symbols or not.  For example, an ink jet process may create dot printed symbols and could be decoded using the stick algorithm and then graded according to the “normal” ISO/IEC 15415 rules and it could pass.  It does not “need” the other DPM related CM and CC related grading processes in order to pass.

There is no facility in ISO/IEC 15415 to use a stick algorithm to grade a symbol.  There is no facility within any symbology specification to use a stick algorithm as part of its reference decode algorithm either, but it would be easy to introduce this.  This change might be very disruptive to existing equipment and therefore, such a change would surely “create” a dot printed version of a symbology since symbols that would be undecodable before, may now obtain a passing grade once the stick algorithm were used in the reference decode algorithm.  However, there is nothing preventing a version of a symbology being created (such as Dot QR or Dot Data Matrix, just like there is Micro QR and Micro PDF 417) which could utilize this technique.

I believe this should be considered – to introduce dot versions of the symbology within the symbology specification that uses a reference decode algorithm that can succeed on dot printed symbols.  (The stick algorithm is only one way to do that, you could introduce other reference decode algorithms to decode such symbols but no one was willing to contribute what amounted to a proprietary technique for decoding dot-peened symbols to the standardization effort and that is why I contributed the stick algorithm in the first place).  For reasons of compatibility, I don’t know if such a change to a symbology specification would be well advised, but it seems to me that an application specification should have the power to do so.  I think that currently, it is well that this technique is well differentiated from “standard” symbols because of DPM grading according to AIM DPM being prefaced with “DPM” so that only applications which specifically call for it will allow it.  By the way, I do not think the “additional grading criteria” clause could be construed to allow an application specification to add the stick algorithm to the reference decode.  Perhaps this power for application specifications can be added, with a corresponding notation such as “DOT” before the grade, in future revisions to 2D matrix symbology specifications.

First let me reiterate my earlier advice to print QR Codes with an X dimension of at least .025 inch and achieve a print quality grade of 2.0/10/660 or better.

Now, I want to discuss a more subtle point.

The QR Code to the right, is typical of codes printed on packages or advertisements and intended to be scanned by cell phones.  The size of the symbol can be thought of in several different ways — the overall size in inches or millimeters, or in terms of how many square units are used to build up the symbol, and how big each of those units are.

Obviously both are related to the overall size of the QR Code.  But the size of the of the unit square (also called a “cell” or a “module”) affects the ability of a camera to resolve and thus successfully read the code.  Of course, the overall size of the code affects the readability too, in that the whole code has to fit inside the camera frame.

If you have a large QR Code, then you have to back away with your camera to fit it into the frame.  As you do so, the resolution of the camera on the image is decreasing.  So it is desirable to have the largest ratio of module size to overall symbol size possible to enhance readability, especially by cameras with lower resolution, and this can be done by minimizing the amount of data encoded in the QR Code.

QR Codes vary in the number of modules that are used to construct it.  In general, the more information that is encoded in the code, the more modules will be required.  The code on the left is typical of most that I have seen in real life.  Note the small “eye” pattern in the lower right area of the symbol.  If you count modules across the the whole symbol (or up and down since QR Codes are square) you will find that there are 25 of them.

The way that a QR symbol encodes data depends upon the data itself.  It is possible for QR Codes to more efficiently encode some types of data than others.  For example, long strings of number digits are more efficient than mixed sets of letters and numbers.  The reason is that the software that constructs the QR Code notices that there are only digits and goes into an efficient encoding mode that only needs to encode one of 10 characters at a time, rather than encode a choice from a larger set for the alphabet, digits, dashes etc.  (Other bar codes have taken advantage of this for a long time, it is not an innovation of QR Code).

Notably QR Code has an encoding mode for Kanji characters which are used for Japanese, even though there are many Kanji characters.  Also, there is a somewhat efficient mode for encoding letters and numbers, but only for Capital Letters.  This means that whenever lower case letters are encoded, the QR Code will be bigger than it would be for the same letters in Upper Case.  This QR Code encodes the same URL, with the same functionality as the one at the top of this page, but it is only 21 (not 25) modules in width and height.  This means for the same overall size, this code can have modules which are roughtly 20% larger.

At the last SC31/WG1 meeting, the Japanese delegation introduced some proposed enhancement to QR Code (ISO/IEC 18004) which would define some hitherto unused modes to encode url’s (internet addresses) in a shorthand format. For example, one mode would prepend the message with the typical preamble to an internet address, such as http://www.

Then, the QR symbol would only have to encode the meaningful part of the url. This shorthand would allow the QR symbol to be “smaller”, at least in terms of the number of modules encoded, so that the X dimension could potentially be larger for symbol that could fit in the same space as an “equivalent” symbol that encoded the whole url.

Could this be the type of QR code that is being referred to as iQR?

Well, actually no.  While they definitely are considering the changes described above, they are also considering a more radical change, possibly keeping is totally separate from the existing QR specification, called iQR.  This would include additional formats, such as some that are rectangular.  They would all resemble Micro QR, in that there is only one finder pattern, not three, and this is clearly to increase the data capacity of a symbol.  So far, it is not clear how or whether this will be developed and introduced or whether it is merely being considered for the future.

- Glenn Spitz

1) Displaying bar codes on displays and having them readable (example: for airline check in)

2) The reading capability of the cameras build into cell phones

3) The size and other requirements for printed symbols (such as on advertisements, posters, etc.) that are intended to be read by the cameras within cell phones.

Obviously 2 and 3 above are intricately related.  The first topic is more disjoint from the second two.

There is a movement toward having the under development ISO document limited in scope to  just the first issue.  That is, how to display readable codes on displays.  It may be that it makes sense to limit the ISO document to this topic only.  I will try to keep posting updates on this as the situation develops.

- Glenn Spitz

First of all there is an ISO document in the works to address these issues and that will be the “final word” on this topic and finally provide a single point of reference for cell phone designers and printers of advertisements. It will also give guidance on how cell phone (or any LCD screen for that matter) should display, or render, a QR symbol (or other bar code) for it to be readable. These dynamic bar codes are already in use at airport check-in counters, but nobody has any idea of how to produce them and ensure that they can be read at the check in counter. However, this particular article will focus only on guidance for printing QR symbols (generally in advertisements) so they can be read by cell phones.

There are generally two sizes of QR symbols to be considered: those on large billboards or posters, and those on advertisements (in newspapers for example), or in stores such as Best Buy. The basic difference in the distance from which you are expected to see the symbol. Cell phone cameras are good at taking pictures of people, or people’s faces, and not so good at taking close up pictures of small objects. Those on posters are very large, similar to the size of a persons face. Those on advertisements or in store (where they actually have a little more room for a larger symbol) are an inch or two across.

Any digital camera has a limited resolution, or number of pixels composing the image. The original iPhone has a 2 Megapixel camera, which actually means that it has 1600×1200 pixels. To get an image in focus you have to hold the iPhone about 24 inches away from the symbol. (You can get a slightly blurry image if you hold it closer). At 24 inches, you have about 1 pixel for every one hundredth of an inch. In bar code land, that is called 10 mils. Reading a bar code typically requires about 3 pixels per module (but some advanced readers can do better) so lets say that the X dimension should be at least 30 mils. That means that a small QR symbol, which has 21 modules across would be less than three quarters of an inch. This is in fact the smallest QR symbol used in advertisements that I’ve seen.

The 3G iPhone has a 3 Megapixel camera, and it also can focus closer to an object.  I believe that the lack of this “macro” capability was one of the major complaints that people had about the camera in the original iPhone.  Holding this iPhone about 3 inches from a bar code gives about 1 pixel per mil, which is actually very good resolution.  With this iPhone, you can read U.P.C. symbols and also smaller QR Codes.

The fact that different cell phone cameras have such vastly different resolutions and reading capabilities, leaves people catering to this market with a big problem.  No one can be sure what size codes they need to print.  Larger X dimensions will make the symbol much easier to read. Therefore, you will see symbols printed very large, in order to be more certain that a larger segment of the cell phone population will be able to read them.  The value of the upcoming ISO document will be that everyone will have a single reference point to target in their designs.

The QR symbols in Best Buy have 29 modules across (not including the quiet zone) and they are slightly less than 1 inch square.  I believe that the X dimension is 33 mils, which is produced with a 300 dpi printer with 10 dots per module.  Speaking of the quiet zone,  QR Code specifies a four module quiet zone around each of the three finder patterns but I’ve noticed that people are not respecting this quiet zone requirement.  Nonetheless, I think that readers do not require this quiet zone in order to find the QR Code symbol.

If a symbol is meant to be read from much further away, the X dimension has to be correspondingly larger. I’ve seen huge symbols will X dimensions approximating an inch (although I’m sure they were denominated in metric units since I was in Japan at the time!)

Now, as for symbol quality: it should be no challenge to produce QR symbols with huge X dimensions with nearly perfect quality. There should be no problem with “modulation” brought on by bar width growth, such as plagues bar codes of all types of smaller (typical) sizes. Furthermore, even “pixel round off” which occurs when the X dimension is not an integer multiple of the dot pitch of the printer will not be a large percentage of a module width.

However, I predict that the color and texture of the material that the symbol is printed on will play a large role in symbol readability.  This is why verification of symbol quality is important at the production stage of advertisements, and the people involved in the design and production of these (many of whom have not been involved in bar code production before) need to become educated on these issues.  And of course, there is not yet any public standard, such as the ISO one I mentioned above which is in the works, to specify quality requirements which can close the loop on the print quality/reader performance relationship.

In the meantime, I recommend that QR Code symbols that are printed on advertisements, and intended to be read by cell phones, should be printed with a minimum X dimension of 30 mil (.75 mm) and have a minimum print quality grade of 3.0/10/660/45.

In future posts, I hope to address the issues of how to render bar codes on LCD screens and how to read these codes with relatively low resolution cameras typical of cell phones.  Now a trivia question: What does “QR” stand for?

- Glenn Spitz

Due to the overwhelming number of Spam comments, I have disabled comments on the Bar Code Blog.  Sorry.

by Glenn Spitz