Camera captured text embedded image processing on handheld mobile devices is an emerging research area. The present work deals with the associated problems of this domain. In this context, images are acquired with built-in digital cameras of various handheld devices such as cell-phones, iPhones, iPods, smart phones, PDAs, etc. As a result, acquired images are often geometrically distorted, unevenly illuminated and blurred. Moreover, acquired images are often heterogeneous and complex in nature, and image information is lost due to image compression while saving the image. This work attempts to develop computationally efficient and light-weight yet effective algorithms for text localization, handling geometric distortions, reverse text correction, binarization, segmentation, character recognition, and indexing as well as retrieval. Along with the algorithms, benchmark camera captured image databases are also prepared. Although, the specific applications targeted are an automatic business card reader and a general text image recognizer followed by indexing and retrieval, the developed techniques may be implemented to serve any applicable purpose.
