改进的基于样例修补的目标移除方法

借助结构张量具有表达图像局部结构的性质，提出一种改进的区域填充图像修补方法．通过结构张量可以使图像的任意像素点的各个色彩分量获得一致的进化方向，其特征向量主导的目标块填充次序使得信息传播方式具有各向异性的特点．在真实自然图像上移除巨大目标物的比较实验表明：文中方法比Criminisi的基于样例修补目标移除方法具有更合理的、视觉上更真实的修补效果．

An Improved Exemplar-Based Inpainting Method for Object Removal

An improved region filling inpainting method is proposed by using structure tensor that can successfully capture the local variations of an image.A structure tensor which takes into account all color channels for a given point in the image can obtain the common evolution direction,and the filling order depending on structure tensor's eigenvectors can make the information propagate in an anisotropic manner.Experiments for large object removal on real images show that the improved method outperforms(Criminisi's) exemplar-based inpainting method by producing visually more plausible effect.