提出一种新的图像修补方法,该方法基于样本的图像修补思想,通过曲率驱动扩散(CDD)模型来计算合理的信心度和数据条件,改进了样本图像修补的有效性并增强图像中线性结构扩散.因此,用本文方法进行图像修补时,能有效地避免其他算法共同存在的"垃圾物"的生成问题.实验结果表明,与其他类似方法相比,本文方法能够得到更令人满意的视觉效果.
In this paper,we present a new image inpainting method based on exemplar-based image inpainting idea by Curvature-Driven Diffusion(CDD) model.This method improves effectiveness and the linear structure propagation by rational confidence and data computing method.Therefore,our method can effectively prevent the "garbage" from producing during the process of inpainting,which is a common problem faced in other methods.With this method,one can obtain more pleasurable vision results than those obtained by other similar methods.