Optical Character Recognition of Low Resolution Text Sequences From Hand-Held Device Supported by Super-Resolution

In this paper a super-resolution technique especially optimized for enhancing low-resolution text images from handheld devices is presented. Considered here, the most naturally, projective motion model has eliminated some drawbacks of previously presented SR schemes. Modification of initial guess estimation idea has resulted in accuracy improvement and convergence speed up. We also applied a quadratic unsharp masking filter which allowed for highlight high frequencies which are then combined with the warped and interpolated image sequence following projective motion estimation using "projective fit" and "projective flow" techniques. Finally, comparison of outputs in the form of OCR results is shown