Text Extraction From images for speech synthesis

We usually come across images that have some text written on them. These may include images of certain signboards that have the name of the destination, or text written on certain billboards. Also, some T-shirts may have text written on them. Other examples include menu cards, business cards, resumes, book covers, etc. The text may have a single font or multiple fonts, the font size may vary character wise or word wise. Also, the text can be written in multiple colors and the background colors can also vary. There may be certain objects in the image other than the text. The aim is to handle all these factors and extract only the text from the image. The text then needs to be converted into speech, which will be the output of the implementation. We currently have certain applications and software’s that can read out text from a PDF document. Most such documents have a plain white background and black-colored text. Most images do not come under the category of such plain images. The proposed system deals with those images that have the above mentioned factors and complexity. The proposed system can be useful in multiple applications, most important of which being robot eyes. It can also be used in software’s that scan resumes and business cards to shortlist people based on certain keywords found in these documents. Coupled with an image acquisition module, the proposed system can be useful to blind people as well.