Barcode of Judgment

As I continue to sit in traffic on I-35, I grow bored with calculating the substrings of car manufacturers and notice an ad for a particularly intriguing restaurant on a nearby cab. Noticing that the ad has an embedded 2-D barcode, I take out my cell phone, snap a picture, and my phone gives me a web address where I find a map to the nearest franchise location.1

Unfortunately I'm still stuck in traffic. So instead of the delicious shrimp dish featured in the ad, I search my floorboards for a spare ketchup packet...

Given an "image" of a 2-D barcode, determine the text string that it encodes. 2-D barcodes have both fixed and variable parts. In the following diagram, the fixed parts are shown as "0" or "1", while variable parts are represented as "-":

   1010---01   1000---00   10-------   10-------   10-------   1000---00   101000001

Note that the variable areas contain the data, and are decoded left-to-right, top-to-bottom when the barcode is oriented as shown. For example, given the actual barcode:

   101000101   100011000   100100010   101101011   100111011   100011000   101000001

The data bits would be:

   001110010001011010110111011110

To translate them into characters, we break them up into 5 bit chunks.

   00111 00100 01011 01011 01110 11110

Then translate them into decimal.

   7 8 11 11 14 30

Finally, decode them according to the following table:

So the example barcode says:

   hello!

The input will begin with a line containing a single integer N (1 ≤ N ≤ 100) indicating the number of data sets. Each data set will begin with a line of the form "H W" where H (7 ≤ H ≤ 80) and W (9 ≤ H ≤ 80) are respectively the height and width of the image in this data set. This is followed by H lines of input, each of which contains W binary digits (i.e., "0" or "1"). These bits represent the black & white images taken by your camera. To simplify further, you can assume that any valid barcode photographed will be squarely photographed, though it may have been rotated any multiple of 90 degrees.

Each data set may contain a single, valid 2-D barcode, in which case you should decode it and display its value. It may contain no valid barcodes, in which case you should print "NOCODE". Finally, it may appear that there are multiple valid barcodes. In this case, other data in the image is obscuring your ability to identify the true barcode, so you should simply print "INTERFERENCE".

[1] Rekimoto, Jun and Ayatsuka, Yuji. 2000. CyberCode: Designing Augmented Reality Environments with Visual Tags. Sony Computer Science Laboratories, Inc.