Automatic restoration of corrupted regions in images using object width transform
Toan Nguyen Dinh, Jeong-Hwan Kim, Guee-Sang Lee · 2012
Natural scene images are usually corrupted by dust, streaks, shadows, or small unwanted objects such as tree branches. In this paper, a new method based on object width transform is proposed to automatically detect and restore corrupted regions in images. First, the input image is converted into a width map that contains the width information of objects in the image. Based on this width map, corrupted regions are detected. Finally, the corrupted regions are restored using adaptive median filter or image completion algorithm. The experimental results attained in many real images show the effectiveness of the proposed method.