Return the id and shift of the code with the minimum bit error

[meyerj]

…from Decoder::decoder().

We did observe some spurious detections of the wrong code, if multiple candidates are sufficiently close, depending on the value of errorCorrection passed to the STag constructor. Even for high values it may be desirable to return the candidate with the least amount of bit errors.

We are using this patch for a while now in our fork (Intermodalics#1), and the number of false detections has been reduced. What do you think?

[MikeK4y]

If we can get the best available result with no significant penalty in performance I think we should merge. Have you noticed any significant delays? I guess that the size of codewords is not big enough to cause a noticeable delay.

[meyerj]

We did not evaluate the impact on run-time explicitly. But indeed, for a library with less than 1000 codewords or so it should not matter much compared to the computational cost of finding STag markers in the image in the first place. Maybe for HD13 or HD11 with more codewords the patch has a more significant impact, but then the importance of not just returning the first, but the best codeword, with the minimal error, is also higher because they tend to be closer to each other, if I understand correctly.