Digital Image Inpainting using Speeded Up Robust Feature

This paper focuses on the inpainting of damaged digital images. It uses relevant image and speeded-up robust features (SURF) for this purpose. A concept wherein, the missing information is restored using relevant image is presented. The relevant image may be a snapshot of the same location with different viewpoint or geometrical transformation. The proposed algorithm is divided into three main stages: Initially, key feature points of damaged and relevant image are found out. In second stage, the relation between the damaged and relevant image is found out in terms of affine transforms (i.e. scale, rotation and translation). Finally, the inverse transformation is applied to reconstruct the damaged area. PSNR is used to compare proposed method with the existing exemplar based method [4] and Hay's scene completion method [9]. The experimental results demonstrate that the proposed inpainting method is efficient in terms of quality and speed.