Image Inpainting by Multiscale Spline Interpolation
Ghazale Ghorbanzade, Zahra Nabizadeh, Nader Karimi, Shadrokh Samavi · 2020
Recovering the missing region of images is a task that is called image inpainting. Due to the shape of missing areas, different methods are presented with different views. One of the challenges of this problem is extracting various features which lead to better results. Experimental results show that both global and local features are useful for this task. In this paper, a multi-scale image inpainting method that utilizes both local and global features is proposed. The first step of this method is selecting the scales of the image according to the width of the lines in the mask. Then adaptive image inpainting is applied to the damaged images, and the lost pixels are predicted, and at the end, before voting the results, the repaired image of each scale is up-sampled. The proposed method is tested on damaged images with scratches and creases. The metric that we use to evaluate our approach is PSNR. In this metric, 1.2 dB improvement exists compare with previous works.