Image de-fencing framework with hybrid inpainting algorithm

Detection and removal of fences from digital images become essential when an important part of the scene turns to be occluded by such unwanted structures. Image de-fencing is challenging because manually marking fence boundaries is tedious and time-consuming. In this paper, a novel image de-fencing algorithm that effectively detects and removes fences with minimal user input is presented. The user is only requested to mark few fence pixels; then, color models are estimated and used to train Bayes classifier to segment the fence and the background. Finally, the fence mask is refined exploiting connected component analysis and morphological operators. To restore the occluded region, a hybrid inpainting algorithm is proposed that integrates exemplar-based technique with a pyramid-based interpolation approach. In contrast to previous solutions which work only for regular pattern fences, the proposed technique is able to remove both regular and irregular fences. A large number of experiments are carried out on a wide variety of images containing different types of fences demonstrating the effectiveness of the proposed approach. The proposed approach is also compared with state-of-the-art image de-fencing and inpainting techniques and showed convincing results.