A new Thresholding Method
In our rough draft, as was pointed out by some members of the classmates and Professor Vacarri, our method for finding the ideal threshold for the OCR was pretty inefficient. Wanting to optimize our process, we added a pair of if-statements to our algorithm to check if the image is underexposed or overexposed by comparing the first and second halves of the threshold's cumulative summary. If the first half is larger, we assume its overexposed. If the second half is larger or equal, we assume its underexposed. We kept our original example because many images may still need to run through multiple thresholds.
​
If the image is underexposed, like the image we use for our example, the thresholds begin at 0.10 and run up to 1.0 (We don't include 0.0 and 0.05 because they almost always produce bad results). In our example, this would mean that only one threshold would need to be tried before the OCR would properly pull the name.
If the image is overexposed, the thresholds begin at 0.90 and run down to 0.0
​
This allows our algorithm to run much faster.
