We need a BIB recognition application. We first thought in using Octave, but I think that Java & think OpenCV is the best option.
Basically what we need is that from N images named <code>_<filename>.jpg located in a specific folder, they get classified in a folder for each <code>, and inside it a folder with the identifiers (bib's) appearing in the images. A single image can contain more than one identifier, so this image must be replicated in each folder.
i.e. Image containing B3667K and H6647J, must be in folders <code>/B3667K and <code>/H6647J.
I can attach you all kinds of sample images, and the extra requirements are maximum recognition rate, accuracy and efficiency.
Could you please make an estimative budget for this?