Deep Learning Image Inpainting Techniques: An Overview

Images can have flaws when captured, may become impaired, or degrade over time. Paintings and other art works symbolize a significant part of one’s legacy and modern culture. But, because of the type of supplies used in these art works, they are susceptible to impairment and they are gradually degraded. Traditionally, these were repaired manually to improve image quality. With the advent of digital age we can avoid manual restoration work by using digital image restoration techniques. The primary aim of inpainting process is recreating the missing portions so that the restored image resembles the original by predicting the missing part of the image by using the details from the other parts of the same image or by referring related images from the bulk data set. This paper presents an overview and comparison of image inpainting techniques using deep learning over the past decade. The steps involved and the pros and cons of each method is discussed. This will serve as a reference for the researchers in choosing the suitable method for their problem.