Machine translation has reached a high level of maturity and accuracy over
the last decade. However, a problem arises with machine translation of text
within images, as this requires manual transcription of the text from the
image, which becomes especially problematic with longer texts or a larger
number of images. The problem becomes even more pronounced when dealing with languages that use characters unfamiliar to the user and not present
on the user’s keyboard. The solution to this problem is the use of optical
character recognition technology and its integration with machine translation
technology in a single application. An additional important functionality is
the automatic replacement of the original text on the image itself with its
translation. This thesis presents the development of such an application and
the broader fields of optical character recognition and machine translation.
|
