【摘要】 数字图像的修复,一般都需要人们用手工来标记待修复的区域。提出了一种新的方法,能够自动标记图像中的待修复区域——网状障碍物。首先将网状障碍物看作网格,利用网格是由空间中有规律两组线交叉形成的这一显著特点,然后通过计算来提取覆盖在图像上的网状障碍物,这样就不需要人们用手工来标记。为了保证得到的网格线能够完全覆盖图像上的网状障碍物,把提取的网格线进行适当的膨胀,实验证明该方法是有效的,且比较容易实现。更多还原

【Abstract】 Image inpainting requires manual operations to specify the regions to be completed.This paper proposes a new method of signing the obstacles on a still image.The obstacles are regarded as the fence,and this method uses the obvious character that the fence consists of two sets of regular lines crossing each other in a space.Then the obstacles can be extracted by computation,and the method doesn’t need manual operations.In order to ensure that the fence obtained can cover the obstacles on the image,we may dilate the lines of the fence.The experimental results show the method is effective.更多还原