Novel Adaptive Filtering for Salt-and-Pepper Noise Removal from Binary Document Images

摘要

Noise removal from binary document and graphic images plays a vital role in the success of various applications. These applications include optical character recognition, content-based image retrieval and hand-written recognition systems. In this paper, we present a novel adaptive scheme for noise removal from binary images. The proposed scheme is based on connected component analysis. Simulations over a set of binary images corrupted by 5%, 10% and 15% salt-and-pepper noise showed that this technique reduces the presence of this noise, while preserving fine thread lines that may be removed by other techniques (such as median and morphological filters).