China
Africa
Explore chapters and articles related to this topic
Image Formation Using Deep Convolutional Generative Adversarial Networks
Published in Vijay Kumar, Mangey Ram, Predictive Analytics, 2021
Unsupervised representation learning is an erudite concept, especially in the realm of research analysing images. In the area of research, it has been conventional to learn reusable feature representations from a large dataset of images. One can employ a multitudinous aggregate of unspecified videos and images to study magnificent intermediary representations. As a consequence, the concerned can perform particular image-related tasks such as image editing or image construction. Some of these tasks embrace the generation and restoration of images. Image generation is examining a hierarchy of representations and their applications, in particular, to employ them to construct new images from a set of already existing ones. Likewise, image restoration [1], also known as image inpainting, is a repairing technique which practices filling in deteriorated parts to structure a complete image. Regeneration of damaged regions is a content-aware computation. The term content-aware signifies filling the targeted portions using their neighbouring pixel information. In the realm of image processing, image inpainting [2] is an operative research content. In this chapter, the focus is on achieving the previously explained image concomitant intents using deep convolutional generative adversarial networks (DCGAN) [3].
Virtual Restoration of Antique Books and Photographs
Published in Filippo Stanco, Sebastiano Battiato, Giovanni Gallo, Digital Imaging for Cultural Heritage Preservation, 2017
Filippo Stanco, Sebastiano Battiato, Giovanni Gallo
Inpainting algorithms [3,8,13,38,41] propagate both the gradient direction and the colors of a band surrounding the hole inside the area to be filled in. Its basic aim is to replace the unrecoverable image data under the opaque foxing layer with values which show good continuity, with respect to the luminance of the area exterior to the stain. Isophote (region with the same level lines) directions are obtained by computing at each pixel along the inpainting contour a gradient vector and by rotating the resulting vector by 90 degrees. This intends to propagate the information while preserving the edges. After few iterations of the inpainting process, the algorithm performs an anisotropic diffusion run to preserve boundaries across the inpainted region.
Generative Adversarial Network
Published in S. Kanimozhi Suguna, M. Dhivya, Sara Paiva, Artificial Intelligence (AI), 2021
Image translation into textual data has been achieved in DCGAN, StackGAN [22], AttnGAN, GAN-INT-CLS, and OP-GAN. Image inpainting can also be performed in GAN. In some images, certain parts of the pictures are missing or hidden by overlapping objects. Image inpainting is a process where these images are translated into images that are filled with accurate details in missing regions. Image inpainting can also be used to remove watermarks in images.
Optimization enabled deep learning approach with probabilistic fusion for image inpainting
Published in The Imaging Science Journal, 2023
Generally, the conventional image inpainting methods are split into two types, such as patch-dependent and diffusion-dependent techniques. The traditional image [13] inpainting approach fills the region by making copies of image patches and semantically-coherent patches from the input image while eliminating the visual as well as semantic plausibility is highly needed. Recently, the DL techniques have reached a remarkable accomplishment in the development of image [14, 15] inpainting. These techniques have filled the missing region through the distribution of learned data. These techniques can generate a coherent model in missing areas, which is impossible in other traditional approaches. Here, the diffusion-dependent techniques send the background information into the missing area by following a diffusive method classically demonstrated via differential operators. Similarly, the Patch-dependent approaches fill the missing region through the patches from an assemblance of source images, which augments the patch similarity [16, 17]. However, these approaches have failed to reconstruct the missed region from complex images [1]. Thus, high-quality image inpainting techniques are needed to reconstruct the missed region, which requires not only visual information but it requires more information about the missed region [18, 19]. The major challenge of image inpainting is the reconstruction of the missed region from the complex latent feature [8].
Irregular mask image inpainting based on progressive generative adversarial networks
Published in The Imaging Science Journal, 2023
Hong-an Li, Liuqing Hu, Jing Zhang
Digital image processing is the processing of image information to meet the needs of human visual psychology and practical applications [1]. With the continuous development of science and technology, people have more requirements for image quality, image inpainting has become an important branch in the field of image processing. The purpose of image inpainting is to fill in the missing pixels in a corrupted image with a given mask [2], in order to achieve consistency in the overall texture structure as well as semantic visual realism. This task has received extensive attention and become a valuable and popular research topic for decades [3–5]. High-quality image inpainting can be useful in a wide range of applications, such as inpainting of old photographs, object removal, image inpainting, image processing, image denoising, etc.[6–11]. Currently, image inpainting methods can be divided into traditional image inpainting methods and deep learning-based image inpainting methods.
The image inpainting algorithm based on pruning samples by referring to four-domains
Published in The Imaging Science Journal, 2019
Image inpainting process refers to the filling process of missing image area. The consistency that the repaired image can keep the structure information continuity and texture information, and it is an important research topic in computer graphs, computer vision and image processing. At present, the digital image inpainting technology is mainly been divided into two categories: one is based on Partial Differential Equation (PDE) from image inpainting technology, and it is mainly applied to repair small damaged image area. The other is the sample image completion based on texture synthesis technology. Repairing the damaged image area with large scale area, the present image repairing technology mainly includes the following two methods: the repairing technology and the method of image decomposition to fill the object area. The testing target block has known information by texture synthesis technique based on block, and selected the best sample blocks from the sample to repairing operation [1–8].