Recent years have witnessed the growth of point cloud based applications for both immersive media as well 3D sensing auto-driving, because its realistic and fine-grained representation objects scenes. However, it is a challenging problem to compress sparse, unstructured, high-precision points efficient communication. In this paper, leveraging sparsity nature cloud, we propose multiscale end-to-end learning framework that hierarchically reconstructs Point Cloud Geometry (PCG) via progressive...

This study develops a unified Point Cloud Geometry (PCG) compression method through the processing of multiscale sparse tensor-based voxelized PCG. We call this SparsePCGC. The proposed SparsePCGC is low complexity solution because it only performs convolutions on sparsely-distributed Most-Probable Positively-Occupied Voxels (MP-POV). representation also allows us to compress scale-wise MP-POVs by exploiting cross-scale and same-scale correlations extensively flexibly. overall efficiency...

Deep learning provides a great potential for in-loop filtering to improve both coding efficiency and subjective quality in video coding. State-of-the-art work focuses on network structure design employs single powerful solve all problems. In contrast, this paper proposes deep based systematic approach that includes an effective Convolutional Neural Network (CNN) structure, hierarchical training strategy, codec oriented switchable mechanism. First, we propose novel CNN i.e.,...

Learning-based point cloud compression has exhibited superior coding performance over the traditional methods such as MEPG G-PCC. Considering that conventional representation formats (e.g., octree or voxel) will introduce additional errors and affect reconstruction quality, we directly use point-based develop a framework leverages transformer upsampling techniques for compression. To extract latent features well characterize an input cloud, build end-to-end learning framework: at encoder...

Orthogonal time frequency space (OTFS) modulation is a promising technique for the next-generation communications in high-mobility scenarios. However, delay-Doppler (DD) domain, received signals suffer from decrease power due to non-coherent superposition of all symbols transmitted through wireless channels. To address this issue, paper proposes incorporation an intelligent reflecting surface (IRS) assist transmission OTFS systems, and jointly designs frame structure IRS phase shifts achieve...

Point cloud geometry (PCG) is used to precisely represent arbitrary-shaped 3D objects and scenes, of great interest vast applications which puts forward the pressing desire high-efficiency PCG compression for transmission storage. Existing coding mostly relies on octree model by point-wise processing applied without exploring nonlocal regional similarity across entire surface. This work, instead, suggests region-wise leverage region exploit inter-region redundancy efficient lossy point...

Boron neutron capture therapy (BNCT) is an emerging approach for treating malignant tumors with binary targeting. However, its clinical application has been hampered by insufficient

Adaptive loop filtering (ALF) is extensively investigated for lossy video coding to mitigate compression noise. Numerous learning-based ALFs have emerged recently and improved the efficiency significantly through use of complexity-intensive, large-scale models trained on excessive samples, making it impractical real-life applications. By contrast, lightweight, small-scale ALF cannot promise convincing performance model generalization. In principle, estimates sample distortion restoration....

This paper presents a hardware design of context-based adaptive binary arithmetic coding (CABAC) for the emerging High efficiency video (HEVC) standard. While aiming at higher compression efficiency, CABAC in HEVC also invests lot effort pursuit parallelism and reducing cost. Simulation results show that our processes 1.18 bins per cycle on average. It can work 357 MHz with 48.940K gates targeting 0.13 μm CMOS process. processing rate support real-time encoding all sequences under common...

Convolution neural network (CNN) has shown its great success in video quality enhancement. Existing methods mainly conduct enhancement tasks the spatial domain, exploring pixel correlations within one frame. Taking advantage of similarity across successive frames, this paper develops a learning-based multi-frame approach, with an aim to explore greatest potential for leveraging temporal correlation. First, we apply optical flow compensate motion neighboring frames. Afterwards, deep CNN...

This article presents a parallel and memory optimized hardware architecture for intra prediction of the High Efficiency Video Coding (HEVC) standard. The consists 64 reconfigurable Processing Elements as datapaths supports all 35 modes sizes from 4×4 to 64×64. In order avoid implementing large area memory-datapaths interconnections save usage, maximum number reference registers is reduced 129 72 by reclassifying into 3 general categories. addition, stage hierarchical processing including an...

The in-loop filter, which constitutes an important part in modern video coding, improves both subjective and objective quality of reconstructed frames. Lately, Convolutional Neural Network (CNN) has demonstrated its superiority over traditional methods addressing filtering problem. In this paper, we develop a CNN-based namely Wide Activation Residual (WARN), for AV1 encoder. On top the plain (ResNet), introduce wide activation to each residual block, making more reasonable allocation network...

Convolutional neural networks (CNNs)-based video quality enhancement generally employs optical flow for pixelwise motion estimation and compensation, followed by utilizing motion-compensated frames jointly exploring the spatiotemporal correlation across to facilitate enhancement. This method, called optical-flow-based method (OPT), usually achieves high accuracy at expense of computational complexity. In this article, we develop a new framework, referred as biprediction-based multiframe...

Given sparse point clouds, this paper develops a perturbation learning-based cloud upsampling method to generate uniform, clean, and dense clouds. We build simple yet efficient neural network framework including feature extraction, learning, coordinate reconstruction operations. In the extraction task, shallow-and-wide connections are applied present latent geometric information. Subsequently, extracted features expanded for learning. According theory of differential geometry surfaces,...

This work extends the multiscale structure originally developed for point cloud geometry compression to attribute compression. To losslessly encode while maintaining a low bitrate, accurate probability prediction is critical. With this aim, we extensively exploit cross-scale, cross-group, and cross-color correlations of ensure estimation thus high coding efficiency. Specifically, first generate tensors through average pooling, by which, any two consecutive scales, decoded lower-scale can be...

While convolution and self-attention are extensively used in learned image compression (LIC) for transform coding, this paper proposes an alternative called Contextual Clustering based LIC (CLIC) which primarily relies on clustering operations local attention correlation characterization compact representation of image. As seen, CLIC expands the receptive field into entire intra-cluster feature aggregation. Afterward, features reordered to their original spatial positions pass through units...

A universal multiscale conditional coding framework, Unicorn, is proposed to compress the geometry and attribute of any given point cloud. Geometry compression addressed in Part I this paper, while discussed II. We construct sparse tensors each voxelized cloud frame properly leverage lower-scale priors current (previously processed) temporal reference frames improve probability approximation or content-aware predictive reconstruction occupancy compression. Unicorn a versatile, learning-based...