For developing a high quality Optical Character Recognition (OCR) system removal of noise from the document image is an utmost important step. To make this possible, filtering plays a significant role. Although mean and median filters, the two well-known statistical filtering techniques, are used commonly but sometimes these filters may fail to produce noise-free images or sometimes may introduce distortions on the characters in the form of gulfs or capes. In the work reported here, we have developed a new filtering technique, called Middle of Modal Class (MMC), for smoothing the input images. This filtering technique is applicable for both the noisy and noise free text document image at the same time. We have also compared our results with mean and median filters, and have achieved better result.
Malakar, Samir; Mohanta, Dheeraj; Sarkar, Ram; and Nasipuri, Mita
"A Novel Noise-removal Technique for Document Images,"
International Journal of Computer and Communication Technology: Vol. 3
, Article 12.
Available at: https://www.interscience.in/ijcct/vol3/iss1/12