基于超像素分割的扫描图像去网算法

目的研究一种扫描图像的去网算法,以去除影响扫描图像质量的半色调网纹或者加网网纹。方法首先基于SLIC超像素算法对扫描图像进行分割并对边界进行矢量化,然后采用线性平滑滤波器和非线性平滑滤波器对图像进行逆半调处理,得到平滑的背景图像,最后把矢量边界与平滑背景整合在一起得到去网后的图像。结果在去除网纹的同时有效保留了边界信息,实验图像的峰值信噪比和结构相似度优于其他算法。结论文中方法可以有效去除网纹,与其他算法相比具有较好的去网效果。

Descreening Algorithm of Scanned Images Based on Superpixel Segmentation

The work aims to study a descreening algorithm of scanned image to remove the halftone pattern or the screen pattern which influences the quality of scanned images. Firstly, the scanned image was segmented based on SLIC superpixel algorithm and the boundary region of image was vectored. Then, the smooth background image was got by inverse halftone processing with linear smoothing filter and nonlinear smoothing filter. Finally, the vectored boundary and smooth background were combined together to get the descreened image. The boundary information was effectively retained at the time of descreening. The PSNR and the SSIM of experimental images were superior to other algorithms. The proposed method can effectively remove the screen patterns, and has better descreening performance compared with other algorithms.