Recently, deep learning based thin cloud removal methods have led to remarkable results. However, these deep learning models often have intricate structures, numerous parameters, and entail substantial training costs, rendering them impractical for widespread implementation in real-world applications. To overcome these challenges, a light-weight network for thin cloud removal, called PM-LSMN(a Physical-Model-based Lightweight Self-attention Multiscale convolution Network), was proposed in this letter. The LSMN module integrates spatial attention mechanism, channel attention mechanism, and multi-scale convolution net. This enhances the network’s ability to capture the distribution of thin clouds in the cloudy images. Base on the physical model of clouds in optical remote sensing images, the network performs element addition operations between the thin cloud distribution images and cloudy images to generate the final cloud-free image. To ensure color consistency in the reconstructed image, a color loss is incorporated into the design of the loss function. On the RICE1 datasets, the proposed method achieves satisfactory results with 25.90 dB in PSNR, 0.93 in SSIM, 4.115 in SAM, and 0.400 in ERGAS. Additionally, the network accomplishes this with a more efficient parameter with 0.33M in Params and computational footprint with 96.2M in FLOPs. The code is available at https://github.com/xizimi/PM-LSMN.git.

This study develops a unified Point Cloud Geometry (PCG) compression method through the processing of multiscale sparse tensor-based voxelized PCG. We call this compression method SparsePCGC. The proposed SparsePCGC is a low complexity solution because it only performs the convolutions on sparsely-distributed Most-Probable Positively-Occupied Voxels (MP-POV). The multiscale representation also allows us to compress scale-wise MP-POVs by exploiting cross-scale and same-scale correlations extensively and flexibly. The overall compression efficiency highly depends on the accuracy of estimated occupancy probability for each MP-POV. Thus, we first design the Sparse Convolution-based Neural Network (SparseCNN) which stacks sparse convolutions and voxel sampling to best characterize and embed spatial correlations. We then develop the SparseCNN-based Occupancy Probability Approximation (SOPA) model to estimate the occupancy probability either in a single-stage manner only using the cross-scale correlation, or in a multi-stage manner by exploiting stage-wise correlation among same-scale neighbors. Besides, we also suggest the SparseCNN based Local Neighborhood Embedding (SLNE) to aggregate local variations as spatial priors in feature attribute to improve the SOPA. Our unified approach not only shows state-of-the-art performance in both lossless and lossy compression modes across a variety of datasets including the dense object PCGs (8iVFB, Owlii, MUVB) and sparse LiDAR PCGs (KITTI, Ford) when compared with standardized MPEG G-PCC and other prevalent learning-based schemes, but also has low complexity which is attractive to practical applications.