ABSTRACT This paper presents the super-resolution algorithm for text images, capable to resize small text image to a bigger one with minimal loose of quality. The algorithm basic flow is divided in 3 main tasks. First task is to generate the dictionary patterns from a big resolution sample text image. The second task is to obtain the nearest similar dictionaries from the low-resolution image and replace them. The last task optimizes the new resized big-resolution image by reducing the noise and improve quality of the characters presuming that the single characters are represented continuity.
↧