Generative models have recently produced good results in image inpainting. Generally, structure, texture, and semantics are essential aspects for an image to be visually appealing. Most of the existing methods fail to consider all these three aspects, producing unwanted artifacts and fuzzy textures, especially for images with large damaged regions. Being an ill posed inverse problem, it is appreciable to solve the problem of inpainting from the perception of prior information. So, the proposed work suggests an inpainting framework named TSSI-GAN: Texture, Structure, and Semantic-based Inpainting - Generative Adversarial Network, which aims to extract the texture, structure, and semantic context of the degraded input image and adopt this knowledge as priors to perform the inpainting process. To obtain photo-realistic images as output, the optimization of the discriminator used in the inpainting module is an essential factor. So, the presented work introduces a novel and persuasive classification approach named Dolphin-Chicken Swarm Optimizer (DCSO), which is used to optimize the discriminator of the inpainting module. The usage of DCSO enhances the inpainting process to produce optimal results such that the inpainted images look visually appealing, and also the similarity between the ground truth and inpainted images is highly appreciable. The proposed DCSO incorporates the Dolphin EchoLocation (DEL) Algorithm with the chicken swarm behavior of the Chicken Swarm Optimization (CSO) algorithm. The proposed framework is evaluated using Flickr Faces-HQ (FFHQ) and DIV2K datasets and the results prove that the suggested framework excels the state-of-the-art models.


    Access

    Check access

    Check availability in my library

    Order at Subito €


    Export, share and cite



    Title :

    Improving Image Inpainting with Emphasis on Texture, Structure and Semantic Features Using BioInspired Generative Adversarial Network


    Contributors:
    Sumathi, G (author) / Devi, M Uma (author)


    Publication date :

    2023-11-22


    Size :

    985031 byte




    Type of media :

    Conference paper


    Type of material :

    Electronic Resource


    Language :

    English



    Multiview Scene Image Inpainting Based on Conditional Generative Adversarial Networks

    Yuan, Zefeng / Li, Hengyu / Liu, Jingyi et al. | IEEE | 2020


    VeIGAN: Vectorial Inpainting Generative Adversarial Network for Depth Maps Object Removal

    Matias, Lucas P. N. / Sons, Marc / Souza, Jefferson R. et al. | IEEE | 2019


    Simultaneous structure and texture image inpainting

    Bertalmio, M. / Vese, L. / Sapiro, G. et al. | IEEE | 2003


    Simultaneous Structure and Texture Image Inpainting

    Bertalmio, M. / Vese, L. / Sapiro, G. et al. | British Library Conference Proceedings | 2003


    Image-to-Image Translation Using Generative Adversarial Network

    Lata, Kusam / Dave, Mayank / Nishanth, K N | IEEE | 2019