Research On Denoise Method For Ancient Inscriptions

Image denoising is the key point of image analysis. In the field of digital image processing,ancient inscription denoising is key technology for restoring inscription. The research on inscription denoising technology has an important practical significance in cultural relics protection and art appreciation.This paper focuses on ancient inscription denoising. Two new methods are proposed, and experiments verified the validity of the proposed method. Main contributions of the paper are summarized up as follows:Firstly, image denoising method based on guided image filter and run-length probability statistics is proposed. Guided image filter is compressed in the stroke smoothing, and using multi-scale Retinex for image enhancement, then, Otsu method is applied to reduce the effect of irrelevant pixels in the subsequent image processing, finally the method of run-length probability statistics is used to remove random noise and isolated noise. Experimental results show that this method can not only reduce the noise effectively, but also can protect the structure of Chinese characters.Secondly, a denoising method for ancient inscriptions based on butterworth low-pass filter and KSVD is developed. This method takes advantage of butterworth low-pass filter to extract low-frequency components,and KSVD method was employed so as to reconstruct high-frequency components and reduce noise, then Otsu method is applied for fusion results of high frequency and low frequency components, finally the visual effect of the image is improved, including the removal of the isolated point noise points by the connected region,the filling of the holes on the stroke and the removal of the burr on the stroke by median filter. Experimental results suggest that the proposed method can remove noise effectively,improve images quality, and improve the visual effect of the image.Thirdly, on the basis of above work, ancient inscription denoising system is designed and implemented.This system is incorporates our proposed two methods,and achieves parts of the classical denoising algorithm.