摘要： 提出一种自适应平滑滤波数字图像修复新算法，采用变动中心的非线性平滑滤波算子，克服了固定中心点算子产生的边缘模糊问题。为了兼顾处理效果和计算效率，根据待修复的像素是位于图像平滑区还是边缘突变区，自适应地切换处理方法。实验表明，该算法对2种类型的像素均能取得较好的修复效果。提出用“分块校验和”方法在需要保护的图像中嵌入脆弱数字水印，一旦图像受损，即可精确得到损坏像素的定位模板，在修复中不再需要人工干预。

关键词: 图像修复, 平滑滤波, 边缘, 脆弱数字水印, 损伤定位

Abstract: An image inpainting technique with adaptive filter is proposed. The filter is a nonlinear smoothing operator with a window whose center may deviate from the current missing pixel to be estimated in order to avoid edge blurring. The size and position of the window are adaptively chosen so that the gray-level variance of the interior pixels is below a given threshold and the number of undamaged pixels is sufficiently large. For a tradeoff between image quality and computation efficiency, the algorithm switches between the eccentric window approach for pixels near edges, and the conventional centered window for pixels in a smooth area. Experimental results show that the proposed method provides good results for both types of pixels. In addition, a fragile watermark is embedded into the image to be protected by using a modified checksum method. In this way, subsequently damaged regions can be easily located.

Key words: image inpainting, smoothing filter, edge, fragile digital watermark, damage localization

中图分类号: 

• TP391.41