基于自适应模板和置信度更新的图像修复算法

去除图像背景中的部分目标,是图像修复技术应用之一。为了修补去除目标后留下的大面积空白区域,文中在基于样本图像修复方法的基础上,引入了自适应思想。首先利用Canny算子检测图像的边缘,再根据图像边缘特征自适应地改变模板块的大小;为了避免误差的累积传递,提出了新的置信度更新方法,使更新后的置信度与累积误差成反比。实验结果证明,文中方法比原方法能更好地修复图像边缘、复杂纹理,并且保留边角,使效果更自然。

Image Inpainting Algorithm Based on the Adaptive Template and Update of the Confidence Term

Removing parts of objects from digital images is one of the applications of image inpainting techniques.To restore blank areas caused by removal of objects better,an adaptive method grounded on exemplar-based inpainting technique is introduced.The algorithm uses Canny operator to detect edges,and then changes block size according to local character of the edges.In order to avoid accumulation of the error,a new method to update the confidence term is proposed.The experimental results demonstrate that the alg...