关键词: 图像修复, 图像放大, 距离变换, 分水岭, 像素填充

Abstract: To preserve sharp and smooth edges in magnified images, an algorithm based on contour fitting interpolation and watershed transform is proposed. The main idea of the algorithm is to magnificate the original image, and automatically locate blurry edges, and do the inpainting process. The edges of initially magnified image are extracted by piecewise cubic Hermite interpolation. A mark image is automatically obtained through edge expansion. A distance transform is applied on the initially magnified image to provide the relief image for watershed algorithm. With the help of the mark and relief image, pixels in blurry regions are gradually filled based on the watershed transform. Experiments demonstrate that：compared with some existing algorithms, the method reduces the sensitivity to the image gradient and makes the edges sharper and smoother. Besides, it also has high computational efficiency.

Key words: image inpainting, image magnification, distance transform, watershed, pixel filling