Image De-fencing Using Wavelet Based Inpainting Technique
M. Varalakshmamma, Tangala Venkateswarlu · International journal of imaging and robotics · 2018
In digital photographs, it is desirable to remove visually-annoying fences when they are occluded. Manually detecting and removing of these fences is tedious and time-consuming. This paper presents a novel approach for the detection of the fence and restoration of the masked fence area from the images. Detection of the fence is done by the threshold based image segmentation and then the detected fence is refined by using the morphological operations. Restoration of the occluded region is done by using the modified Wavelet-based inpainting technique. This algorithm works for removal of regular and irregular fences from digital photographs. Experiments were carried out on the PSU-NRT dataset and De-fenced dataset. The comparative results with the state-of-art image de-fencing and inpainting techniques are provided.