{"id":1096,"date":"2015-01-23T15:31:29","date_gmt":"2015-01-23T20:31:29","guid":{"rendered":"https:\/\/blogs.mathworks.com\/loren\/?p=1096"},"modified":"2015-01-23T15:31:29","modified_gmt":"2015-01-23T20:31:29","slug":"what-was-sheraton-trying-to-tell-us","status":"publish","type":"post","link":"https:\/\/blogs.mathworks.com\/loren\/2015\/01\/23\/what-was-sheraton-trying-to-tell-us\/","title":{"rendered":"What was Sheraton Trying to Tell Us?"},"content":{"rendered":"\r\n

What was Sheraton Trying to Tell Us?<\/h3>

Today's post is by Sean de Wolski<\/a> and me. We spent a few days this week at training for the start of the year, held in a Sheraton hotel.<\/p>

At the completion of our stay, we each got a note in our rooms. Here is a sample.<\/p>

\"\" <\/p>

Actually we read this in with the App from the Image Acquisition Toolbox<\/a> and cropped the image (imcrop<\/tt><\/a> from Image Processing Toolbox<\/a>) to just display the mysterious binary text.<\/p>

\"\" <\/p>

Using functionality from the Computer Vision System Toolbox<\/a>, we then converted our binary text using the ocr<\/tt><\/a> (Optical Character Recognition) function.<\/p>

Icrop = imread('Sher.png'<\/span>);\r\nbintext = ocr(Icrop)\r\nstr = bintext.Words;\r\nstr = [str{:}];\r\n<\/pre>
bintext = \r\n  ocrText with properties:\r\n\r\n                      Text: '010101000110100001100001011011100110101100100...'\r\n    CharacterBoundingBoxes: [879x4 double]\r\n      CharacterConfidences: [879x1 single]\r\n                     Words: {14x1 cell}\r\n         WordBoundingBoxes: [14x4 double]\r\n           WordConfidences: [14x1 single]\r\n<\/pre>
Next guess was how the text was encoded - 4 bits, 8 bits?  We figured 8 was most likely.<\/p>
newstr = reshape(str',8,[])';\r\n<\/pre>
Let's look at the first few lines.<\/p>
snip = newstr(1:3,:)\r\n<\/pre>
snip =\r\n01010100\r\n01101000\r\n01100001\r\n<\/pre>
Could these be text?<\/p>
bin2dec(snip)\r\n<\/pre>
ans =\r\n    84\r\n   104\r\n    97\r\n<\/pre>
These could be ASCII characters!<\/p>
Have you ever misread an O for a 0?  Or an l for a 1?  We have!  So we decided to program defensively for those and any other weird swaps that might confuse the OCR.<\/p>
newstr(newstr == 'O'<\/span>) = '0'<\/span>;\r\nnewstr(newstr ~='0'<\/span>) = '1'<\/span>;\r\n<\/pre>
Finally let's write out the result.<\/p>
message = char(bin2dec(newstr))';\r\n\r\ndisp(message(1:74))\r\ndisp(message(75:end))\r\n<\/pre>
Thank you for your continued loyalty to Sheraton Boston. It is a pleasure \r\nto host Mathworks year after year.\r\n<\/pre>