The Non-Separable Transform (NST) is an effective transform algorithm for improving coding efficiency in H.266/Versatile Video Coding (VVC) and Beyond VVC. In this paper, we propose an adaptive reordering strategy and fast pruning techniques on the encoder side to optimize the transform kernel selection of NST. The absolute sum values of forward transform coefficients (CoefAbsSum) are used as a coefficients energy criterion for evaluating transform kernels: a smaller CoefAbsSum indicates a better transformation of the kernel. Two key technical efforts are designed: we first propose an adaptive reordering for NST transform kernels, which sorts the three transform kernels in each prediction mode's transform set according to CoefAbsSum. It thereby prioritizes kernels with better potential Rate-Distortion (RD) performance. Subsequently, we design three fast algorithms for the sorted kernels, including adjusted RD cost threshold pruning, CoefAbsSum-based pre-screening, and historical RD performance comparison to skip modes with poor preestimates. Experimental results, conducted based on ECM 17.0, show that the proposed framework achieves an average encoding time reduction of 1.3 % (with the encoding time ratio of 98.7 %); meanwhile, it maintains a BD-rate change of 0.00 % for the Y component, 0.03 % for the Cb component, and 0.01 % for the Cr component. The Beyond VVC is an exploration platform with extremely high coding complexity, and our work reduces the complexity while introducing nearly no coding loss.

Adaptive loop filter (ALF), including luma ALF, chroma ALF, and cross-component adaptive loop filter (CCALF), has been adopted in H.266/versatile video coding (VVC). It can enhance the quality of reconstructed videos based on the Wiener filtering principle. In-depth analyses reveal that the efficiency of chroma ALF and CCALF is significantly lower than that of luma ALF, primarily due to the insufficient number of chroma-related filter sets in the adaptation parameter sets (APSs). To address this limitation, this paper focuses on the efficient design of ALF APSs. Specifically, we propose a luma-guided rate-distortion optimization (RDO) criterion for chroma components to improve the efficiency of the new ALF APS. Furthermore, an improved ALF APS list management is introduced to extend the lifetime of chroma-related filter sets in the ALF APS list.

Video transcoding is essential for compatibility, visual quality, and bit-rate reduction. However, the intermediate video produced by initial compression is inherently unfriendly to subsequent re-compression. Existing methods typically treat this problem either as post-processing for the initial compression or as pre-processing for the re-compression, lacking an integrated solution that jointly optimizes video processing, initial and recompression. In this work, we propose a joint optimization algorithm that enables the video processing to bridge the initial and re-compression.

In this paper, we derive a theoretical model for the temporal propagation of perceptual distortion and propose a novel adaptive Quantization Parameter (QP) adjustment strategy to enhance the video coding performance. Temporal adaptive quantization (AQ) models and exploits the complex temporal dependencies introduced by inter-frame prediction, with the objective of Peak Signal-to-Noise Ratio (PSNR) optimization. In contrast to temporal AQ, spatial AQ primarily focuses on intra-frame characteristics and targets Structural Similarity (SSIM) optimization. Motivated by these complementary properties, we propose an algorithm that integrates both aspects.

Motion Estimation (ME) is one of the most computationally intensive module in H.266/VVC, and Fractional Motion Estimation (FME) accounts for the majority of its complexity. While substantial research efforts have been devoted to affine motion estimation (AME) and integer motion estimation (IME), studies on FME optimization remain limited. In this paper, a fast FME algorithm is proposed based on motion vector (MV) correlation from spatially adjacent coding units (CUs). The overall pipeline of the algorithm is illustrated in Fig. 1. The algorithm utilizes FME results from neighboring CUs to guide the FME process for the current CU. A buffer is maintained to store motion information from recently encoded CUs. During FME processing, the current CU retrieves information about adjacent CUs from this buffer. For half-pixel search, the algorithm first identifies integer motion vector (IMV)-similar CUs from the buffer, defined as CUs whose MVs differ from the current CU's IMV by no more than 0.75 pixels in both horizontal and vertical dimensions. For each IMV-similar CU, the motion vector offset (MVO), calculated as the difference between the cached CU's MV and the current CU's IMV, is computed and mapped to one of nine half-pixel search directions. The occurrences of each direction are accumulated, and if the count for any direction satisfies a statistically derived threshold, the half-pixel search is restricted to that specific direction. When the fast half-pixel search condition is satisfied, quarter-pixel search is performed utilizing the selected IMV-similar CUs. For each of these CUs, the quarter-pixel motion vector offset (QMVO) is calculated as the difference between the cached CU's MV and the predicted half-pixel MV obtained from the fast half-pixel search. Each QMVO is mapped to a specific quarter-pixel search pattern, and if the occurrence count for any pattern meets a statistically derived threshold, the quarter-pixel search is restricted to that pattern. Experimental results under the VVenC 1.12.0 medium preset show that the proposed algorithm achieves an average encoding time saving of 1.44% with only a 0.20% BD-rate increase.

Point clouds serve as a fundamental format for virtual and augmented reality applications. However, their substantial data volumes present considerable challenges regarding data storage and transmission. This paper introduces a novel point transformer-based approach for point cloud attribute compression, termed as CST-PCAC, which utilizes a transformer architecture designed to exploit correlations across multiple scales. Our method is grounded in a variational autoencoder framework, enhanced by stacked sparse convolution (SConv) and the proposed Cross-Scale Point Transformer (CSFormer). CSFormer incorporates two types of attention mechanisms: Within-Scale Attention (WSA) and Cross-Scale Attention (CSA). WSA constructs attention maps within K-nearest neighbours at the current scale, while CSA operates across adjacent upper or lower scales. By leveraging varied sparsity distributions and the geometry distortions inherent in consecutive scales, CSFormer enhances the receptive field and feature representation, thereby effectively extracting correlations to minimize redundancy. Experimental results demonstrate that CST-PCAC achieves average Bjøntegaard Delta bitrate gains exceeding 30%, 10%, and 20% in comparison to Sparse-PCAC, Scalable-PCAC, and G-PCC, respectively. In terms of computational efficiency, CST-PCAC exhibits a time complexity two orders of magnitude lower than that of deep learning-based methods including Sparse-PCAC and Scalable-PCAC, while maintaining a comparable complexity to traditional methods like G-PCC.

Cloud gaming is a popular paradigm in the gaming industry, allowing real-time gaming experiences powered by cloud computing. Due to a strict latency sensitivity, efficient video encoding is crucial for its wide application. Cloud gaming videos are produced by game engines, which provide auxiliary data such as rendered generative motion vectors (RG-MVs). These motion vectors, calculated by 3D geometry and representing ground-truth motion, offer significant potential for fast video coding design.

Blendshapes have been widely adopted as a key method for generating facial animation on avatars due to their ease of manipulation, flexibility in capturing diverse facial expressions, and compatibility with real-time rendering. However, current frameworks lack efficient methods for compressing blendshape (BS) animation parameters, which are critical for optimizing data transmission. This study introduces a pioneer compression scheme leveraging the amount of BS to be transmitted, their quantization as well as their transmission frequency. Subjective evaluation using the ITU-R BT.500-15 recommendation demonstrates that the proposed method significantly reduces the amount of data to transmit while preserving acceptable visual quality. This approach extends prior findings on reduced BS sets[kang2023effects] and addresses a significant gap in avatar media coding[avril2023morgan]. This work establishes a baseline for efficient facial animation and, serves as a foundation for further exploration on adaptive rate-allocation strategies and advanced compression strategies tailored for diverse avatar animation scenarios.

This paper investigates the impact of standardized Geometry-based Point Cloud Compression (G-PCC) on LiDAR-based object detection tasks. It evaluates how compression distortion types, and distortion levels, together with point cloud density, and object semantics affect various 3D object detection methods, including point-based, voxel-based, and hybrid approaches, using the KITTI object detection dataset.

The Moving Picture Experts Group (MPEG) is developing a new generation of the mesh coding standard, called Video-based Dynamic Mesh Coding (V-DMC). This standard represents a dense original mesh using a sparser base mesh, combined with a subdivision scheme and a set of displacement vectors that indicate the disparities between the subdivided base mesh and the deformed mesh, which approximates the original mesh. However, the lossy compression of the base mesh introduces displacement biases. Moreover, the one-dimensional encoding of these vectors, while saving bits, also leads to a loss of displacement information. In this paper, we propose a modified encoding framework to enhance the representation of displacement, thereby correcting biases and reducing information loss. Our method involves post-deformation to ensure the deformed mesh aligns with the reconstructed base mesh. Furthermore, by optimizing the deformed mesh corresponding to the reconstructed base mesh, more information is preserved in the encoded displacements, thus improving the quality of the reconstructed mesh. Experimental results show significant BD-rate gains over the V-DMC test model version 8.0, leading to its adoption in the V-DMC test model version 9.0.

The continuous evolution of cloud gaming makes it crucial for efficient video compression to relieve the growing bandwidth pressure. This paper presents a coding optimization algorithm for cloud gaming videos based on three-dimensional image warping, considering the significance of reference frames in inter-frame prediction. By leveraging camera parameters and motion information from the game engine, we generate a virtual reference frame through three-dimensional image warping on the selected key frame and the coloring operation. The resulting virtual reference frame exhibits minimal spatial differences from the frame to be encoded and is utilized to update the original reference frame list in two different methods, achieving accurate inter-frame prediction and significantly reducing coding residuals. Experimental results demonstrate the superiority of our algorithm compared to the mainstream cloud gaming encoder in various camera movements, achieving an average BD-rate reduction of 5.60% and BD-PSNR gain of 0.55dB.

Rate-distortion optimization (RDO) is widely applied in video coding, which aims at minimizing the coding distortion under a target coding rate. Conventionally, RDO in video coding does not take into account the coding complexity. However, because of the diversity of video applications, the video encoders in different applications may have different requirements of or limitation on the computational complexity. Therefore, it is desirable for video encoders to perform RDO in flexible computational complexity. In this paper, we propose a novel RDO scheme under the given computational complexity for the latest H.265/HEVC standard. A model for prediction of the rate-distortion cost (RD cost) is first established based on a pre-searching process. Then according to the predicted RD cost, the rate-distortion-complexity (R-D-C) characteristics of different coding tree units (CTUs) are analyzed. Finally, the total complexity budget is properly allocated to different CTUs according to their R-D-C characteristics. Experimental results demonstrate that, compared with x265, the proposed algorithm can reduce, on average, the BD-rate by 18.8% under the same requirements of encoding speed.