A Method Of Salt-pepper Denoising For Text Images

Salt-pepper noise is a kind of typical noise in images. It degrades the quality of images. There are many denoising methods, such as the mean filtering, the median filtering and the weighted mean filtering based on histogram, etc. These methods are simple but effective in image restoration.This thesis mainly discusses the removal problem of salt-pepper noise in text images where the character fracture or stroke loss appears due to the noise. Firstly, this thesis shows the differences of text images to generic images. Then it points out the existed problems in traditional salt-pepper denoising methods and addresses the key problems-restoring the character structure while reducing the stroke fuzzy. Finally, it presents a new salt-pepper denoising method for text images by first marking the noise points, then classfying signal points into character points and background points, and at last defusing the information to a noise point from its neighboring signal points. Experiment results show that the proposed method is simple and effective.