Optical Character Recognition (OCR) software is typically optimized for flatbed scanners. The image output of a digital camera can be much different from that of a scanner. Since DocScanner is mostly used in mobile devices that come with an integrated digital camera, this has been the major challenge for DocScanner’s OCR in the past.
One part of OCR in DocScanner is based on noise reduction and edge detection. Noise reduction in mobile OCR is of special importance since all digital photographs include a certain degree of noise. The more noise the image has, the more challenging it becomes to recognize characters in it, as the noise shows up as random black dots among the characters once the image is binarized. Therefore noise reduction is the key for better OCR results.
A part of the DocScanner development team is currently improving the existing OCR feature even further. The noise reduction is being fine tuned into a whole new level to make nearly errorless character recognition possible, even if the original photo has a relatively high degree of noise in it. In this quest for the best OCR results, latest hardware and software techniques, such as neural network calculation are being used. Currently, the enhanced OCR is getting faster and more accurate every day. Once complete, these improvements will be supplied to all DocScanner users in an update. The new OCR will also be the backbone of our soon-to-come contact management tool ContactStash.com.