1. INTRODUCTION
With pervasive use of consumer-domain hand-held devices
such as Smartphones, camera-captured optical character
recognition (OCR) is a key capability of interest to both
commercial and defense communities. However, images
such as document images, road signs, etc. acquired using
cameras on these hand-held devices exhibit several artifacts
that significantly impact OCR accuracy. Degradations such
as out-of-focus blur, uneven and insufficient illumination, etc.
simply lead to failure of OCR on these images.
Most document retrieval and recognition systems based
on OCR rely on acquisition of high-quality document images.
Therefore, automated image quality assessment methods that
automatically select images of high quality (as shown in Fig.
1(b)) and filter out images that do not satisfy the minimum
quality requirement for the system (as shown in Fig. 1(c))
can significantly improve the usability of such systems. Additionally,
the image quality assessment can be integrated into