In order to solve the defects of edge artifacts and semantic discontinuity when inpainting large and irregular distortions in an image with fine texture and complex background, a residual connection image restoration algorithm based on parallel generation convolution was proposed. The damaged image was inputted into a two-column parallel convolutional structure inpainting network to obtain two image components with different sizes of reception fields. The two image components were combined by the shared decoding and the L2 loss of the output was calculated to optimize the network. The output of the coarse network was sent into the fine inpainting network which contained the residual connection and the attention mechanism to fuse context information and improve the ability of repairing fine details. The global and local discriminators and visual geometry group net were used to calculate the loss and to optimize the overall discrimination network and enhance the global and local consistency of the inpainting result. The performance of the proposed algorithm was validated on internationally recognized databases, and experimental results show that the proposed algorithm can effectively repair large and irregular missing areas under complex background and fine texture, improve the authenticity and integrity of image details, semantics and structure. Its peak signal-to-noise ratio and structural similarity are superior to the state-of-the-art.

To effectively solve the problems of low feature utilization and poor image structure coherence occurred when existing algorithms are used to repair large irregularly missing images, this study proposed an image repair algorithm based on dense feature inference (DFR) and hybrid loss function. The repair network consists of multiple inference modules (FRs) densely connected. Firstly, after the image to be restored was fed into the first inference module for feature inference, the output feature map channels were merged and sent to the next inference module. The input of each subsequent inference module was the inferred features from all the previous inference modules and so on, so as to make full use of the feature information captured by each reasoning module. Subsequently, a propagation consistent attention (PCA) mechanism was proposed to improve the overall consistency of the patched regions with the known regions. Finally, a hybrid loss function (ML) was proposed to optimize the structural coherence of the repair results. The whole DFR network adopted group normalization (GN), and excellent repair results can be achieved even using small training batches. The performance of the proposed algorithm was verified on Paris StreetView and CelebA face datasets, which are internationally recognized datasets. The objective and subjective experimental results show that the proposed algorithm can effectively repair large irregular missing images, improve feature utilization and structural coherence. Its average peak signal-to-noise ratio (PSNR), average structural similarity ( SSIM), mean square error (MSE), Fréchet distance (FID) and learning perceptual image block similarity (LPIPS) metrics all outperform the comparison algorithms.