SSIM Metrics Research Articles

Joint Optimization-Based Texture and Geometry Enhancement Method for Single-Image-Based 3D Content Creation

Recent studies have explored the generation of three-dimensional (3D) meshes from single images. A key challenge in this area is the difficulty of improving both the generalization and detail simultaneously in 3D mesh generation. To address this issue, existing methods utilize fixed-resolution mesh features to train networks for generalization. This approach is capable of generating the overall 3D shape without limitations on object categories. However, the generated shape often exhibits a blurred surface and suffers from suboptimal texture resolution due to the fixed-resolution mesh features. In this study, we propose a joint optimization method that enhances geometry and texture by integrating generalized 3D mesh generation with adjustable mesh resolution. Specifically, we apply an inverse-rendering-based remeshing technique that enables the estimation of complex-shaped mesh estimations without relying on fixed-resolution structures. After remeshing, we enhance the texture to improve the detailed quality of the remeshed mesh via a texture enhancement diffusion model. By separating the tasks of generalization, detailed geometry estimation, and texture enhancement and adapting different target features for each specific network, the proposed joint optimization method effectively addresses the characteristics of individual objects, resulting in increased surface detail and the generation of high-quality textures. Experimental results on the Google Scanned Objects and ShapeNet datasets demonstrate that the proposed method significantly improves the accuracy of 3D geometry and texture estimation, as evaluated by the PSNR, SSIM, LPIPS, and CD metrics.

A Deep Learning-Based Two-Branch Generative Adversarial Network for Image De-Raining

Raindrops can scatter and absorb light, causing images to become blurry or distorted. To improve image quality by reducing the impact of raindrops, this paper proposes a novel generative adversarial network for image de-raining. The network comprises two parts: a generative network and an adversarial network. The generative network performs image de-raining. The adversarial network determines whether the input image is rain-free or de-rained. The generative network comprises two branches: the A branch, which follows a traditional convolutional network structure, and the U branch, which utilizes a U-Net architecture. The A branch includes a multi-scale module for extracting information at different scales and a residual attention module to reduce redundant information interference. The U branch contains an encoder module designed to address the loss of details and local information caused by conventional down-sampling. To improve the performance of the generative network in image de-raining, this paper employs a relative discriminator incorporating a mean squared error loss. This discriminator better measures the differences between rainy and rain-free images while effectively preventing the occurrence of gradient vanishing. Finally, this study performs visual and quantitative comparisons of the proposed method and existing methods on three established rain image datasets. In the quantitative experiments, the proposed method outperforms existing methods regarding PSNR, SSIM, and VIF metrics. Specifically, our method achieves an average PSNR, SSIM, and VIF of approximately 5%, 3%, and 4% higher than the MFAA-GAN method, respectively. These results indicate that the de-rained images generated via the proposed method are closer to rain-free images.

Robust image tamper detection and recovery with self-embedding watermarking using SPIHT and LDPC

In today’s digital landscape, the pervasive use of digital images across diverse domains has led to growing concerns regarding their authenticity and reliability. The potential for malicious manipulation of these images underscores the critical need to develop robust methods for detecting tampering and ensuring their integrity. Fragile watermarking has been found to have extensive applications for tamper detection and recovery. An image watermarking technique for tamper detection, correction, and recovery is presented in this study. The proposed method employs a self-embedding method to generate the reference watermark from the original image, which has the advantage of superior tamper detection, localization, recovery capabilities, and robustness against attacks. Set Partitioning in the Hierarchical Tree (SPIHT) algorithm is applied to generate a reference watermark of the original image. Low-Density Parity Check (LDPC) is employed for error correction, providing higher-quality reconstruction to recover the original image. Schur decomposition processed the watermarked image blocks to generate authentication bits for each block to enhance tampering detection. The proposed watermarking method was evaluated using PSNR, SSIM, BER, and NC metrics for grayscale and colored images. The technique demonstrated high robustness in tamper detection and recovery against various malicious attacks. Comparative analysis with existing methods shows the efficacy of the proposed method.

Enhancing Steganography in 256×256 Colored Images with U-Net: A Study on PSNR and SSIM Metrics with Variable-Sized Hidden Images

Enhancing Steganography in 256×256 Colored Images with U-Net: A Study on PSNR and SSIM Metrics with Variable-Sized Hidden Images

Machine Learning for Pedestrian-Level Wind Comfort Analysis

(1) Background: Artificial intelligence (AI) and machine learning (ML) techniques are being more widely employed in the field of wind engineering. Nevertheless, there is a scarcity of research on the comfort of pedestrians in terms of wind conditions with respect to building design, particularly in historic sites. (2) Objectives: This research aims to evaluate ML- and computational fluid dynamics (CFD)-based pedestrian wind comfort (PWC) analysis outputs using a novel method that relies on the sophisticated handling of image data. The goal is to propose a novel assessment method to enhance the efficiency of AI models over different urban scenarios. (3) Methodology: The stages include the analysis of climate data, CFD analysis with OpenFOAM, ML analysis using Autodesk Forma, and comparisons of the CFD and ML results using a novel image similarity assessment method based on the SSIM, MSE, and PSNR metrics. (4) Conclusions: This study effectively demonstrates the considerable potential of utilizing ML as a supplementary tool for evaluating PWC. It maintains a high degree of accuracy and precision, allowing for rapid and effective assessments. The methodology for precise comparison of two visual outputs in the absence of numerical data allows for more objective and pertinent comparisons, as it eliminates any potential distortions. (5) Recommendations: Additional research can explore the integration of ML models with climate data and different case studies, thus expanding the scope of wind comfort studies.

Enhanced Wind Field Spatial Downscaling Method Using UNET Architecture and Dual Cross-Attention Mechanism

Before 2008, China lacked high-coverage regional surface observation data, making it difficult for the China Meteorological Administration Land Data Assimilation System (CLDAS) to directly backtrack high-resolution, high-quality land assimilation products. To address this issue, this paper proposes a deep learning model named UNET_DCA, based on the UNET architecture, which incorporates a Dual Cross-Attention module (DCA) for multiscale feature fusion by introducing Channel Cross-Attention (CCA) and Spatial Cross-Attention (SCA) mechanisms. This model focuses on the near-surface 10-m wind field and achieves spatial downscaling from 6.25 km to 1 km. We conducted training and validation using data from 2020–2021, tested with data from 2019, and performed ablation experiments to validate the effectiveness of each module. We compared the results with traditional bilinear interpolation methods and the SNCA-CLDASSD model. The experimental results show that the UNET-based model outperforms SNCA-CLDASSD, indicating that the UNET-based model captures richer information in wind field downscaling compared to SNCA-CLDASSD, which relies on sequentially stacked CNN convolution modules. UNET_CCA and UNET_SCA, incorporating cross-attention mechanisms, outperform UNET without attention mechanisms. Furthermore, UNET_DCA, incorporating both Channel Cross-Attention and Spatial Cross-Attention mechanisms, outperforms UNET_CCA and UNET_SCA, which only incorporate one attention mechanism. UNET_DCA performs best on the RMSE, MAE, and COR metrics (0.40 m/s, 0.28 m/s, 0.93), while UNET_DCA_ars, incorporating more auxiliary information, performs best on the PSNR and SSIM metrics (29.006, 0.880). Evaluation across different methods indicates that the optimal model performs best in valleys, followed by mountains, and worst in plains; it performs worse during the day and better at night; and as wind speed levels increase, accuracy decreases. Overall, among various downscaling methods, UNET_DCA and UNET_DCA_ars effectively reconstruct the spatial details of wind fields, providing a deeper exploration for the inversion of high-resolution historical meteorological grid data.

Deformable registration of magnetic resonance images using unsupervised deep learning in neuro-/radiation oncology

PurposeAccurate deformable registration of magnetic resonance imaging (MRI) scans containing pathologies is challenging due to changes in tissue appearance. In this paper, we developed a novel automated three-dimensional (3D) convolutional U-Net based deformable image registration (ConvUNet-DIR) method using unsupervised learning to establish correspondence between baseline pre-operative and follow-up MRI scans of patients with brain glioma.MethodsThis study involved multi-parametric brain MRI scans (T1, T1-contrast enhanced, T2, FLAIR) acquired at pre-operative and follow-up time for 160 patients diagnosed with glioma, representing the BraTS-Reg 2022 challenge dataset. ConvUNet-DIR, a deep learning-based deformable registration workflow using 3D U-Net style architecture as a core, was developed to establish correspondence between the MRI scans. The workflow consists of three components: (1) the U-Net learns features from pairs of MRI scans and estimates a mapping between them, (2) the grid generator computes the sampling grid based on the derived transformation parameters, and (3) the spatial transformation layer generates a warped image by applying the sampling operation using interpolation. A similarity measure was used as a loss function for the network with a regularization parameter limiting the deformation. The model was trained via unsupervised learning using pairs of MRI scans on a training data set (n = 102) and validated on a validation data set (n = 26) to assess its generalizability. Its performance was evaluated on a test set (n = 32) by computing the Dice score and structural similarity index (SSIM) quantitative metrics. The model’s performance also was compared with the baseline state-of-the-art VoxelMorph (VM1 and VM2) learning-based algorithms.ResultsThe ConvUNet-DIR model showed promising competency in performing accurate 3D deformable registration. It achieved a mean Dice score of 0.975 ± 0.003 and SSIM of 0.908 ± 0.011 on the test set (n = 32). Experimental results also demonstrated that ConvUNet-DIR outperformed the VoxelMorph algorithms concerning Dice (VM1: 0.969 ± 0.006 and VM2: 0.957 ± 0.008) and SSIM (VM1: 0.893 ± 0.012 and VM2: 0.857 ± 0.017) metrics. The time required to perform a registration for a pair of MRI scans is about 1 s on the CPU.ConclusionsThe developed deep learning-based model can perform an end-to-end deformable registration of a pair of 3D MRI scans for glioma patients without human intervention. The model could provide accurate, efficient, and robust deformable registration without needing pre-alignment and labeling. It outperformed the state-of-the-art VoxelMorph learning-based deformable registration algorithms and other supervised/unsupervised deep learning-based methods reported in the literature.

Non-local sparse attention based swin transformer V2 for image super-resolution

In single image super resolution tasks, distortion measurement (such as PSNR, SSIM) and perceptual quality (such as PI, NIQE) are contradictory, and methods that perform well in perceptual quality often perform poorly in distortion measurement, and vice versa. In this article, we propose a method of balancing the two, which is divided into three stages. Firstly, this article proposes an image super-resolution model NLSAV2 that focuses on PSNR and SSIM metrics. The entire NLSAV2 consists of three modules: shallow feature extraction, deep feature extraction, and high-quality image reconstruction. In the shallow feature extraction module, non-local sparse attention is used to identify the most abundant feature information in mapping input from low-dimensional space to high-dimensional space, and the deep feature extraction module mainly consists of residual Swin Transformer V2 Block. Then, NLSAV2 is used as a generator and a relative discriminator is introduced to further train the model, which is called NLSAV2-GAN. The experimental results indicate that NLSAV2 and NLSAV2-GAN exhibit advantages in distortion measurement and perceptual quality respectively. Finally, network interpolation and image interpolation strategies are used to continuously adjust the reconstruction style and smoothness to achieve a balance between the two.

Building Edge Detection Technology From Remote Sensing Image Based On NSCT And Tensor Voting

An edge detection technology based on the combination of non-downsampling contour wave transform (NSCT) and tensor voting is proposed, which aims to obtain more accurate and detailed edge information of buildings in remote sensing images. Firstly, NSCT is used for image decomposition to obtain the subband frequency information of different scales and angles. Then, position encoding is performed on these subband coefficients to obtain second-order symmetric tensors at the corresponding positions. Tensors of different scales and angles at the same position are weighted and summed to complete feature fusion. Finally, the edge features of the image are obtained based on tensor voting theory. The experimental results show that compared to common edge detection technologies, such as Canny, Fast Edge (Fast edge detection using structured forests) and HED (Holistically-Nested Edge Detection), our method can more accurately and intensively reflect the boundaries of buildings and the edge information of roofs, providing better support for the analysis of building types and architectural styles. Compared to the HED method, which is based on deep learning, our method improves PSNR and SSIM metrics by 0.98 and 0.03, respectively.

CGIHE-VDSR: Color global image histogram equalization with very deep super resolution networks for color image super resolution

Medical and satellite image analysis require incredibly high resolution. Super-resolution combines several low-resolution images of the same scene to generate a high-resolution image. The Super resolution employing deep learning techniques still has an illumination issue. This paper proposes a novel CGIHE-VDSR algorithm that integrates the Very Deep Super Resolution (VDSR) Network with Color Global Image Histogram Equalization (CGIHE) to improve image resolution. In the proposed method, the low-resolution image is first histogram equalized using the CGIHE algorithm. Then, the VDSR network is applied to the histogram equalized image for super-resolution. The comparison of real-time data with the benchmark images is done using the proposed algorithm in the MATLAB platform. The PSNR and SSIM metrics demonstrate that the super resolution image obtained using the proposed method is significantly better than the existing methods.

MMDCP: An Image Enhancement Algorithm Incorporating Multi-Channel Phase Activation and Multi-Constrained Dark Channel Prior

The quality of visual media is critically impacted by low illumination and the presence of airborne particulates, leading to challenges in brightness balance, color saturation, and texture clarity which are detrimental to various applications in image processing and computer vision. Addressing these challenges, this study introduces a novel image enhancement algorithm that significantly improves the quality of degraded images. Our proposed method, the multi-channel phase activation and multi-constraint dark channel prior (MMDCP), leverages an innovative approach by integrating the phase-adjusted Gaussian kernel function for brightness channel optimization in the Fourier transform frequency domain. This optimization is enhanced through the application of a saturated dark channel prior, achieving simultaneous brightness enhancement and color fidelity. Furthermore, we refine the dark channel prior deblurring algorithm by incorporating intensity, brightness, and color constraints to correct overexposure issues and color offsets in the reconstructed images. The efficacy of the MMDCP algorithm is demonstrated through extensive experimentation, comparing it against six contemporary image enhancement algorithms using two types of objective indicators and subjective assessments across four public datasets. The MMDCP algorithm consistently outperforms the existing methods, with a notable average improvement of 20% in PSNR and 19.6% in SSIM metrics, substantiating its superiority in enhancing brightness, detail, and color accuracy. This study’s results underline the MMDCP algorithm’s robustness and versatility in improving image quality across various conditions, including daytime, nighttime, indoor, and outdoor settings.

Edge Boosted Global Awared Low-light Image Enhancement Network

Low-light images are captured in situations where the lighting is poor or the camera hardware is not capable of producing good quality images. These types of images tend to have low contrast, blurry details, noise, and color distortion. In computer vision applications, image brightness plays a crucial role, and therefore, low-light image enhancement is used as a preprocessing step. In this study, we have improved the Low-Light Enhancement Network with Global Awareness (GLADNet) method by adding a UNet-based edge information extraction unit. The channel attention mechanism was also incorporated into the edge information extraction unit to achieve color preservation. Our experiments show that our proposed method has achieved higher PSNR, SSIM, and FSIM metrics compared to reference images. Additionally, it has produced lower NIQE and BRISQUE values for non-reference performance evaluation. Moreover, our proposed method removes noise better and produces visual results that are closer to the target images.

DSG-GAN:A dual-stage-generator-based GAN for cross-modality synthesis from PET to CT

PET/CT devices typically use CT images for PET attenuation correction, leading to additional radiation exposure. Alternatively, in a standalone PET imaging system, attenuation and scatter correction cannot be performed due to the absence of CT images. Therefore, it is necessary to explore methods for generating pseudo-CT images from PET images. However, traditional PET-to-CT synthesis models encounter conflicts in multi-objective optimization, leading to disparities between synthetic and real images in overall structure and texture. To address this issue, we propose a staged image generation model. Firstly, we construct a dual-stage generator, which synthesizes the overall structure and texture details of images by decomposing optimization objectives and employing multiple loss functions constraints. Additionally, in each generator, we employ improved deep perceptual skip connections, which utilize cross-layer information interaction and deep perceptual selection to effectively and selectively leverage multi-level deep information and avoid interference from redundant information. Finally, we construct a context-aware local discriminator, which integrates context information and extracts local features to generate fine local details of images and reasonably maintain the overall coherence of the images. Experimental results demonstrate that our approach outperforms other methods, with SSIM, PSNR, and FID metrics reaching 0.8993, 29.6108, and 29.7489, respectively, achieving the state-of-the-art. Furthermore, we conduct visual experiments on the synthesized pseudo-CT images in terms of image structure and texture. The results indicate that the pseudo-CT images synthesized in this study are more similar to real CT images, providing accurate structure information for clinical disease analysis and lesion localization.

Colorful image reconstruction from neuromorphic event cameras with biologically inspired deep color fusion neural networks.

Neuromorphic event-based cameras communicate transients in luminance instead of frames, providing visual information with a fine temporal resolution, high dynamic range and high signal-to-noise ratio. Enriching event data with color information allows for the reconstruction of colorful frame-like intensity maps, supporting improved performance and visually appealing results in various computer vision tasks. In this work, we simulated a biologically inspired color fusion system featuring a three-stage convolutional neural network for reconstructing color intensity maps from event data and sparse color cues. While current approaches for color fusion use full RGB frames in high resolution, our design uses event data and low-spatial and tonal-resolution quantized color cues, providing a high-performing small model for efficient colorful image reconstruction. The proposed model outperforms existing coloring schemes in terms of SSIM, LPIPS, PSNR, and CIEDE2000 metrics. We demonstrate that auxiliary limited color information can be used in conjunction with event data to successfully reconstruct both color and intensity frames, paving the way for more efficient hardware designs.

StainSWIN: Vision transformer-based stain normalization for histopathology image analysis

Stain normalization is a key preprocessing step that has been shown to significantly improve the segmentation and classification performance of computer-aided diagnosis (CAD) systems. In recent advancements, numerous approaches have demonstrated significant progress in the domain of stain normalization; however, the most of these approaches are based on Generative Adversarial Networks. In this paper, we propose a novel vision transformer-based model, termed as StainSWIN, that combines the strengths of swin transformer and the architecture of super resolution to achieve improved performance in stain normalization task. The key concept behind the StainSWIN is the utilization of swin transformer blocks that exploit content-based interactions to capture long-range dependencies. The proposed model is equipped with two key blocks, including residual stain swin block (ResStainSWIN) and swin transformer block (STB). The StainSWIN has a residual super resolution architecture, in which the high-level features, extracted by STB, are combined to ResStainSWIN block. The performance of the StainSWIN model was compared with other state-of-the-art methods on a commonly used MITOS-ATYPIA14 histopathology dataset. The StainSWIN outperformed other methods in stain normalization with a large margin in terms of PSNR, SSIM, and RMSE metrics. The StainSWIN model achieved PSNR of 26.667 ± 3.492, SSIM of 0.943 ± 0.037, and RMSE of 6.206 ± 1.973. Additionally, we evaluated the model’s impact to the segmentation performance of the MICCAI GlaS’16 dataset. The results demonstrates a 4.3% improvement in segmentation accuracy, attributed to a reduction in stain color variation. The proposed method has the ability to greatly assist CAD systems in maintaining consistent performance despite color variations.

Neural radiance fields-based multi-view endoscopic scene reconstruction for surgical simulation.

In virtual surgery, the appearance of 3D models constructed from CT images lacks realism, leading to potential misunderstandings among residents. Therefore, it is crucial to reconstruct realistic endoscopic scene using multi-view images captured by an endoscope. We propose an Endoscope-NeRF network for implicit radiance fields reconstruction of endoscopic scene under non-fixed light source, and synthesize novel views using volume rendering. Endoscope-NeRF network with multiple MLP networks and a ray transformer network represents endoscopic scene as implicit field function with color and volume density at continuous 5D vectors (3D position and 2D direction). The final synthesized image is obtained by aggregating all sampling points on each ray of the target camera using volume rendering. Our method considers the effect of distance from the light source to the sampling point on the scene radiance. Our network is validated on the lung, liver, kidney and heart of pig collected by our device. The results show that the novel views of endoscopic scene synthesized by our method outperform existing methods (NeRF and IBRNet) in terms of PSNR, SSIM, and LPIPS metrics. Our network can effectively learn a radiance field function with generalization ability. Fine-tuning the pre-trained model on a new endoscopic scene to further optimize the neural radiance fields of the scene, which can provide more realistic, high-resolution rendered images for surgical simulation.

HATF: Multi-Modal Feature Learning for Infrared and Visible Image Fusion via Hybrid Attention Transformer

Current CNN-based methods for infrared and visible image fusion are limited by the low discrimination of extracted structural features, the adoption of uniform loss functions, and the lack of inter-modal feature interaction, which make it difficult to obtain optimal fusion results. To alleviate the above problems, a framework for multimodal feature learning fusion using a cross-attention Transformer is proposed. To extract rich structural features at different scales, residual U-Nets with mixed receptive fields are adopted to capture salient object information at various granularities. Then, a hybrid attention fusion strategy is employed to integrate the complementing information from the input images. Finally, adaptive loss functions are designed to achieve optimal fusion results for different modal features. The fusion framework proposed in this study is thoroughly evaluated using the TNO, FLIR, and LLVIP datasets, encompassing diverse scenes and varying illumination conditions. In the comparative experiments, HATF achieved competitive results on three datasets, with EN, SD, MI, and SSIM metrics reaching the best performance on the TNO dataset, surpassing the second-best method by 2.3%, 18.8%, 4.2%, and 2.2%, respectively. These results validate the effectiveness of the proposed method in terms of both robustness and image fusion quality compared to several popular methods.

FDSR: An Interpretable Frequency Division Stepwise Process Based Single-Image Super-Resolution Network.

Deep learning has excelled in single-image super-resolution (SISR) applications, yet the lack of interpretability in most deep learning-based SR networks hinders their applicability, especially in fields like medical imaging that require transparent computation. To address these problems, we present an interpretable frequency division SR network that operates in the image frequency domain. It comprises a frequency division module and a step-wise reconstruction method, which divides the image into different frequencies and performs reconstruction accordingly. We develop a frequency division loss function to ensure that each reconstruction module (ReM) operates solely at one image frequency. These methods establish an interpretable framework for SR networks, visualizing the image reconstruction process and reducing the black box nature of SR networks. Additionally, we revisited the subpixel layer upsampling process by deriving its inverse process and designing a displacement generation module. This interpretable upsampling process incorporates subpixel information and is similar to pre-upsampling frameworks. Furthermore, we develop a new ReM based on interpretable Hessian attention to enhance network performance. Extensive experiments demonstrate that our network, without the frequency division loss, outperforms state-of-the-art methods qualitatively and quantitatively. The inclusion of the frequency division loss enhances the network's interpretability and robustness, and only slightly decreases the PSNR and SSIM metrics by an average of 0.48 dB and 0.0049, respectively.

Image Enhancement Network Architecture for Multidimensional Fusion of Medical Imaging Data under Intense Light Interference

Abstract Due to the limitations of imaging equipment and environment, the acquired medical images usually have a certain degree of noise and artifacts, which leads to the degradation of the quality of medical images and affects the doctors’ clinical diagnosis of the condition. In this paper, the Gauss-Laplace operator is used to perform normalized filtering on medical images to reduce the influence of noise and improve the convolution effect of images. Through the CLAHE algorithm, the histogram is optimized for equalization, and the network architecture of the image is designed in this way. The quality of the enhanced image is evaluated through experimental design and dataset processing. In the evaluation of subjective and objective metrics, the PSNR and SSIM metrics of the images in SR × 2 are improved by 1.576 dB and 0.997 dB, respectively, on the BraTS dataset. This algorithm’s subjective score is the most high among the four enhancement algorithms, with an average score of 8.25, which aligns with the objective evaluation results. Among the image enhancement results, this paper’s algorithm better adjusts the histogram distribution with h(k) distribution ranging from 0.526-4.215, which is better than other enhancement algorithms in detail enhancement.

A Cloud Coverage Image Reconstruction Approach for Remote Sensing of Temperature and Vegetation in Amazon Rainforest

Remote sensing involves actions to obtain information about an area located on Earth. In the Amazon region, the presence of clouds is a common occurrence, and the visualization of important terrestrial information in the image, like vegetation and temperature, can be difficult. In order to estimate land surface temperature (LST) and the normalized difference vegetation index (NDVI) from satellite images with cloud coverage, the inpainting approach will be applied to remove clouds and restore the image of the removed region. This paper proposes the use of the neural network LaMa (large mask inpainting) and the scalable model named Big LaMa for the automatic reconstruction process in satellite images. Experiments are conducted on Landsat-8 satellite images of the Amazon rainforest in the state of Acre, Brazil. To evaluate the architecture’s accuracy, the RMSE (root mean squared error), SSIM (structural similarity index) and PSNR (peak signal-to-noise ratio) metrics were used. The LST and NDVI of the reconstructed image were calculated and compared qualitatively and quantitatively, using scatter plots and the chosen metrics, respectively. The experimental results show that the Big LaMa architecture performs more effectively and robustly in restoring images in terms of visual quality. And the LaMa network shows minimal superiority for the measured metrics when addressing medium marked areas. When comparing the results achieved in NDVI and LST of the reconstructed images with real cloud coverage, great visual results were obtained with Big LaMa.