Article Preview
Top1. Introduction
Image inpainting is the reconstruction process of lost or worse portion in images and videos (Ruikar & Ghuge, 2016; Ghuge, et al., 2018). The image inpainting is an antique art which requires the human artists to perform the work by hand. Nowadays, investigators or researchers have developed many techniques for automatic inpainting. Besides to the image, many techniques also need a mask as input for representing the regions in an image. The image inpainting operation acts as an important research area, and the results from these research areas can be applied to cultural relic’s protection, text removal, television and film effect, and in several other specific aspects (Wu & Ruan, 2006). Image inpainting methods are extensively used in the applications of image restoration, object removal, image compressing and various other applications. The aim of reconstructing the injured or the misplaced parts of the images (Jegatheeswari & Deepa, 2019) with the use of the incomplete data that are observed as accurate as possible is termed as image inpainting (Mo & Zhou, 2018). The applications, such as suppresment of the scratches and texts in the ancient drawings, removal of the undesirable area of a film or photograph, restoration of the missed pixel at the time of image transmission (Zheng, et al., 2018) make in use of the concept of image inpainting. The usage of the self-similarity inherent of the images by the image inpainting algorithms helps in the synthesis of the missing data. In addition, the inpainting processing strategy helps in the reduction of noise, demosaicing, and super resolution, and so on (Guo, et al., 2018). The inpainted portion is merged into the image in such a way, which cannot be identified by a viewer (Fadili, et al., 2009). The Image inpainting method has been utilized in image restoration, image compression, object removal, and so on (Fan & Zhang, 2018). Now a day, it has been used in visual effect production, image rectification, image compression, and so on (Jin & Bai, 2018). In addition, image inpainting can be applied to repair the damaged photos, complete the missing regions, image deblurring, and so on (Karaca & Tunga, 2018).
The image inpainting methods are classified into a texture synthesis method and partial differential equation method (Wu & Ruan, 2006). The image-level features are used by conventional inpainting techniques to resolve the problem of hole filling. The Patch-Match (Guo, et al., 2018) is a method used in finding the best matching patches for the reconstruction of the missing area. However, these methods can be used only for low-level features and thus cannot be applied for the high-level features. In addition, they propagate the image from outside to the hole, instead of obtaining the global image structure, which is unwanted by a human (Song, et al., 2018). Two types of image inpainting methods are (i) image editing in undesired or text object removal and (ii) the image restoration in blob and scratch removal from the old image. These methods are further divided into two types, namely exemplar-based techniques for texture synthesis and diffusion-based techniques for structure propagation. In the case of the diffusion-based methods (Bertalmio, et al., 2000–Bornemann & März, 2007), the data is proliferated smoothly from its borderline. These techniques utilize the partial differential equation (PDE) for the propagation of the linear structures in the isophote direction. The PDE based techniques are capable of producing a thin target region, like scratch covered with the smooth region. At the same time, in the case of texture images, these methods make a blurring effect with the presence of smoothing terms (Ghorai, et al., 2018).