Ravi,

Which text are you talking about? Are you talking about embedded image tags and comments that contain text or a text representation (OCR) of the image. For instance, JPG's can contain author and copyright information embedded in the binary image seperate from the image data.

I use LeadTools to deal with both types, in version 14 the OCR engine is great, the levels of OCR accuracy are dictated by factors other than the engine ie. image quality (DPI), cleanup. I abstract these settings into "OCR Profiles" that the end-user can configure to match a particular type of source media.

If you want to extract embedded "metadata" from images then again I use LeadTools as this provides an abstract interface for general embedded comments/information and a specialised interface for TIFF tags.

The downside to using Lead is that its not cheap for the OCR functionality.

HTH
Neil