| The MIM Algorithm is a mathematical function which can be used for signal restoration by minimizing the mutual information of the signal. Because of these enhancement properties, the MIM can be very useful in the area of image enhancement and more importantly text restoration. Unlike with other text restoration methods, the MIM algorithm can iteratively enhance the image without prior information such as the signal power of the image and the noise. In previous work the MIM has been shown to enhance blurry and noisy text images of normal text. This was accomplished by converting the image into the Radon domain. By changing to the Radon domain via the Radon Transform, the lines in the image were converted to points or peaks in the Radon domain. The MIM algorithm was able to exploit the peaky nature of this domain, enhance those peaks, and convert back to the image domain with a de-blurred image estimate. The use of the radon domain also allows for considerable computation speed and storage enhancements due to the fact that data is converted from line segments to points. Since the Radon Transform maps lines to points, this algorithm has limitations when it comes to round text. In this paper we present an expansion on this previous approach using overcompleteness coupled with the generalized radon transform. This allows for enhancement of both regular text as well as that of round text. |
