Wen-Hsiao Peng

Yung-Han Ho, Chih-Peng Chang, Peng-Yu Chen et al. · pku

This paper presents an end-to-end learning-based video compression system, termed CANF-VC, based on conditional augmented normalizing flows (CANF). Most learned video compression systems adopt the same hybrid-based coding architecture as the traditional codecs. Recent research on conditional coding has shown the sub-optimality of the hybrid-based coding and opens up opportunities for deep generative models to take a key role in creating new coding frameworks. CANF-VC represents a new attempt that leverages the conditional ANF to learn a video generative model for conditional inter-frame coding. We choose ANF because it is a special type of generative model, which includes variational autoencoder as a special case and is able to achieve better expressiveness. CANF-VC also extends the idea of conditional coding to motion coding, forming a purely conditional coding framework. Extensive experimental results on commonly used datasets confirm the superiority of CANF-VC to the state-of-the-art methods. The source code of CANF-VC is available at https://github.com/NYCU-MAPL/CANF-VC.

David Alexandre, Hsueh-Ming Hang, Wen-Hsiao Peng

Typical video compression systems consist of two main modules: motion coding and residual coding. This general architecture is adopted by classical coding schemes (such as international standards H.265 and H.266) and deep learning-based coding schemes. We propose a novel B-frame coding architecture based on two-layer Conditional Augmented Normalization Flows (CANF). It has the striking feature of not transmitting any motion information. Our proposed idea of video compression without motion coding offers a new direction for learned video coding. Our base layer is a low-resolution image compressor that replaces the full-resolution motion compressor. The low-resolution coded image is merged with the warped high-resolution images to generate a high-quality image as a conditioning signal for the enhancement-layer image coding in full resolution. One advantage of this architecture is significantly reduced computational complexity due to eliminating the motion information compressor. In addition, we adopt a skip-mode coding technique to reduce the transmitted latent samples. The rate-distortion performance of our scheme is slightly lower than that of the state-of-the-art learned B-frame coding scheme, B-CANF, but outperforms other learned B-frame coding schemes. However, compared to B-CANF, our scheme saves 45% of multiply-accumulate operations (MACs) for encoding and 27% of MACs for decoding. The code is available at https://nycu-clab.github.io.

Shih-Po Lee, Niraj Prakash Kini, Wen-Hsiao Peng et al.

This paper introduces a novel human pose estimation benchmark, Human Pose with Millimeter Wave Radar (HuPR), that includes synchronized vision and radio signal components. This dataset is created using cross-calibrated mmWave radar sensors and a monocular RGB camera for cross-modality training of radar-based human pose estimation. There are two advantages of using mmWave radar to perform human pose estimation. First, it is robust to dark and low-light conditions. Second, it is not visually perceivable by humans and thus, can be widely applied to applications with privacy concerns, e.g., surveillance systems in patient rooms. In addition to the benchmark, we propose a cross-modality training framework that leverages the ground-truth 2D keypoints representing human body joints for training, which are systematically generated from the pre-trained 2D pose estimation network based on a monocular camera input image, avoiding laborious manual label annotation efforts. The framework consists of a new radar pre-processing method that better extracts the velocity information from radar data, Cross- and Self-Attention Module (CSAM), to fuse multi-scale radar features, and Pose Refinement Graph Convolutional Networks (PRGCN), to refine the predicted keypoint confidence heatmaps. Our intensive experiments on the HuPR benchmark show that the proposed scheme achieves better human pose estimation performance with only radar data, as compared to traditional pre-processing solutions and previous radio-frequency-based methods.

Su-Kai Chen, Hung-Lin Yen, Yu-Lun Liu et al. · nvidia

Deep learning is commonly used to reconstruct HDR images from LDR images. LDR stack-based methods are used for single-image HDR reconstruction, generating an HDR image from a deep learning-generated LDR stack. However, current methods generate the stack with predetermined exposure values (EVs), which may limit the quality of HDR reconstruction. To address this, we propose the continuous exposure value representation (CEVR), which uses an implicit function to generate LDR images with arbitrary EVs, including those unseen during training. Our approach generates a continuous stack with more images containing diverse EVs, significantly improving HDR reconstruction. We use a cycle training strategy to supervise the model in generating continuous EV LDR images without corresponding ground truths. Our CEVR model outperforms existing methods, as demonstrated by experimental results.

Mu-Jung Chen, Yi-Hsin Chen, Wen-Hsiao Peng · pku

Over the past few years, learning-based video compression has become an active research area. However, most works focus on P-frame coding. Learned B-frame coding is under-explored and more challenging. This work introduces a novel B-frame coding framework, termed B-CANF, that exploits conditional augmented normalizing flows for B-frame coding. B-CANF additionally features two novel elements: frame-type adaptive coding and B*-frames. Our frame-type adaptive coding learns better bit allocation for hierarchical B-frame coding by dynamically adapting the feature distributions according to the B-frame type. Our B*-frames allow greater flexibility in specifying the group-of-pictures (GOP) structure by reusing the B-frame codec to mimic P-frame coding, without the need for an additional, separate P-frame codec. On commonly used datasets, B-CANF achieves the state-of-the-art compression performance as compared to the other learned B-frame codecs and shows comparable BD-rate results to HM-16.23 under the random access configuration in terms of PSNR. When evaluated on different GOP structures, our B*-frames achieve similar performance to the additional use of a separate P-frame codec.

Yung-Han Ho, Chih-Hsuan Lin, Peng-Yu Chen et al. · pku

This paper proposes a learning-based video compression framework for variable-rate coding on YUV 4:2:0 content. Most existing learning-based video compression models adopt the traditional hybrid-based coding architecture, which involves temporal prediction followed by residual coding. However, recent studies have shown that residual coding is sub-optimal from the information-theoretic perspective. In addition, most existing models are optimized with respect to RGB content. Furthermore, they require separate models for variable-rate coding. To address these issues, this work presents an attempt to incorporate the conditional inter-frame coding for YUV 4:2:0 content. We introduce a conditional flow-based inter-frame coder to improve the inter-frame coding efficiency. To adapt our codec to YUV 4:2:0 content, we adopt a simple strategy of using space-to-depth and depth-to-space conversions. Lastly, we employ a rate-adaption net to achieve variable-rate coding without training multiple models. Experimental results show that our model performs better than x265 on UVG and MCL-JCV datasets in terms of PSNR-YUV. However, on the more challenging datasets from ISCAS'22 GC, there is still ample room for improvement. This insufficient performance is due to the lack of inter-frame coding capability at a large GOP size and can be mitigated by increasing the model capacity and applying an error propagation-aware training strategy.

Yung-Han Ho, Chia-Hao Kao, Wen-Hsiao Peng et al.

This paper presents a reinforcement learning (RL) framework that utilizes Frank-Wolfe policy optimization to solve Coding-Tree-Unit (CTU) bit allocation for Region-of-Interest (ROI) intra-frame coding. Most previous RL-based methods employ the single-critic design, where the rewards for distortion minimization and rate regularization are weighted by an empirically chosen hyper-parameter. Recently, the dual-critic design is proposed to update the actor by alternating the rate and distortion critics. However, its convergence is not guaranteed. To address these issues, we introduce Neural Frank-Wolfe Policy Optimization (NFWPO) in formulating the CTU-level bit allocation as an action-constrained RL problem. In this new framework, we exploit a rate critic to predict a feasible set of actions. With this feasible set, a distortion critic is invoked to update the actor to maximize the ROI-weighted image quality subject to a rate constraint. Experimental results produced with x265 confirm the superiority of the proposed method to the other baselines.

Mu-Jung Chen, Hong-Sheng Xie, Cheng Chien et al.

This paper introduces a learned hierarchical B-frame coding scheme in response to the Grand Challenge on Neural Network-based Video Coding at ISCAS 2023. We address specifically three issues, including (1) B-frame coding, (2) YUV 4:2:0 coding, and (3) content-adaptive variable-rate coding with only one single model. Most learned video codecs operate internally in the RGB domain for P-frame coding. B-frame coding for YUV 4:2:0 content is largely under-explored. In addition, while there have been prior works on variable-rate coding with conditional convolution, most of them fail to consider the content information. We build our scheme on conditional augmented normalized flows (CANF). It features conditional motion and inter-frame codecs for efficient B-frame coding. To cope with YUV 4:2:0 content, two conditional inter-frame codecs are used to process the Y and UV components separately, with the coding of the UV components conditioned additionally on the Y component. Moreover, we introduce adaptive feature modulation in every convolutional layer, taking into account both the content information and the coding levels of B-frames to achieve content-adaptive variable-rate coding. Experimental results show that our model outperforms x265 and the winner of last year's challenge on commonly used datasets in terms of PSNR-YUV.

Chia-Hao Kao, Yi-Hsin Chen, Cheng Chien et al.

This paper presents a Transformer-based image compression system that allows for a variable image quality objective according to the user's preference. Optimizing a learned codec for different quality objectives leads to reconstructed images with varying visual characteristics. Our method provides the user with the flexibility to choose a trade-off between two image quality objectives using a single, shared model. Motivated by the success of prompt-tuning techniques, we introduce prompt tokens to condition our Transformer-based autoencoder. These prompt tokens are generated adaptively based on the user's preference and input image through learning a prompt generation network. Extensive experiments on commonly used quality metrics demonstrate the effectiveness of our method in adapting the encoding and/or decoding processes to a variable quality objective. While offering the additional flexibility, our proposed method performs comparably to the single-objective methods in terms of rate-distortion performance.

Chih-Hsuan Lin, Yi-Hsin Chen, Wen-Hsiao Peng

This paper introduces an online motion rate adaptation scheme for learned video compression, with the aim of achieving content-adaptive coding on individual test sequences to mitigate the domain gap between training and test data. It features a patch-level bit allocation map, termed the $α$-map, to trade off between the bit rates for motion and inter-frame coding in a spatially-adaptive manner. We optimize the $α$-map through an online back-propagation scheme at inference time. Moreover, we incorporate a look-ahead mechanism to consider its impact on future frames. Extensive experimental results confirm that the proposed scheme, when integrated into a conditional learned video codec, is able to adapt motion bit rate effectively, showing much improved rate-distortion performance particularly on test sequences with complicated motion characteristics.

Yung-Han Ho, Yun Liang, Chia-Hao Kao et al.

This paper presents a reinforcement learning (RL) framework that leverages Frank-Wolfe policy optimization to address frame-level bit allocation for HEVC/H.265. Most previous RL-based approaches adopt the single-critic design, which weights the rewards for distortion minimization and rate regularization by an empirically chosen hyper-parameter. More recently, the dual-critic design is proposed to update the actor network by alternating the rate and distortion critics. However, the convergence of training is not guaranteed. To address this issue, we introduce Neural Frank-Wolfe Policy Optimization (NFWPO) in formulating the frame-level bit allocation as an action-constrained RL problem. In this new framework, the rate critic serves to specify a feasible action set, and the distortion critic updates the actor network towards maximizing the reconstruction quality while conforming to the action constraint. Experimental results show that when trained to optimize the video multi-method assessment fusion (VMAF) metric, our NFWPO-based model outperforms both the single-critic and the dual-critic methods. It also demonstrates comparable rate-distortion performance to the 2-pass average bit rate control of x265.

Niraj Prakash Kini, Shiau-Rung Tsai, Guan-Hsun Lin et al.

Millimeter-wave radar offers a privacy-preserving and lighting-invariant alternative to RGB sensors for Human Pose Estimation (HPE) task. However, the radar signals are often sparse due to specular reflection, making the extraction of robust features from radar signals highly challenging. To address this, we present milliMamba, a radar-based 2D human pose estimation framework that jointly models spatio-temporal dependencies across both the feature extraction and decoding stages. Specifically, given the high dimensionality of radar inputs, we adopt a Cross-View Fusion Mamba encoder to efficiently extract spatio-temporal features from longer sequences with linear complexity. A Spatio-Temporal-Cross Attention decoder then predicts joint coordinates across multiple frames. Together, this spatio-temporal modeling pipeline enables the model to leverage contextual cues from neighboring frames and joints to infer missing joints caused by specular reflections. To reinforce motion smoothness, we incorporate a velocity loss alongside the standard keypoint loss during training. Experiments on the TransHuPR and HuPR datasets demonstrate that our method achieves significant performance improvements, exceeding the baselines by 11.0 AP and 14.6 AP, respectively, while maintaining reasonable complexity. Code: https://github.com/NYCU-MAPL/milliMamba

Chia-Hao Kao, Cheng Chien, Yu-Jen Tseng et al.

This paper presents the first-ever study of adapting compressed image latents to suit the needs of downstream vision tasks that adopt Multimodal Large Language Models (MLLMs). MLLMs have extended the success of large language models to modalities (e.g. images) beyond text, but their billion scale hinders deployment on resource-constrained end devices. While cloud-hosted MLLMs could be available, transmitting raw, uncompressed images captured by end devices to the cloud requires an efficient image compression system. To address this, we focus on emerging neural image compression and propose a novel framework with a lightweight transform-neck and a surrogate loss to adapt compressed image latents for MLLM-based vision tasks. Given the huge scale of MLLMs, our framework excludes the entire downstream MLLM except part of its visual encoder from training our system. This stands out from most existing coding for machine approaches that involve downstream networks in training and thus could be impractical when the networks are MLLMs. The proposed framework is general in that it is applicable to various MLLMs, neural image codecs, and multiple application scenarios, where the neural image codec can be (1) pre-trained for human perception without updating, (2) fully updated for joint human and machine perception, or (3) fully updated for only machine perception. Extensive experiments on different neural image codecs and various MLLMs show that our method achieves great rate-accuracy performance with much less complexity.

Sang NguyenQuang, Xiem HoangVan, Wen-Hsiao Peng

Learned B-frame codecs with hierarchical temporal prediction often encounter the domain-shift issue due to mismatches between the Group-of-Pictures (GOP) sizes for training and testing, leading to inaccurate motion estimates, particularly for large motion. A common solution is to turn large motion into small motion by downsampling video frames during motion estimation. However, determining the optimal downsampling factor typically requires costly rate-distortion optimization. This work introduces lightweight classifiers to predict downsampling factors. These classifiers leverage simple state signals from current and reference frames to balance rate-distortion performance with computational cost. Three variants are proposed: (1) a binary classifier (Bi-Class) trained with Focal Loss to choose between high and low resolutions, (2) a multi-class classifier (Mu-Class) trained with novel soft labels based on rate-distortion costs, and (3) a co-class approach (Co-Class) that combines the predictive capability of the multi-class classifier with the selective search of the binary classifier. All classifier methods can work seamlessly with existing B-frame codecs without requiring codec retraining. Experimental results show that they achieve coding performance comparable to exhaustive search methods while significantly reducing computational complexity. The code is available at: https://github.com/NYCU-MAPL/Fast-OMRA.git.

Huu-Phu Do, Hao-Chien Hsueh, Tzu-Hao Chiang et al.

Photo exposure correction is widely investigated, but fewer studies focus on correcting under- and over-exposed images simultaneously. Three issues remain open to handle and correct both under- and over-exposed images in a unified way. First, a locally-adaptive exposure adjustment may be more flexible instead of learning a global mapping. Second, it is an ill-posed problem to determine the suitable exposure values locally. Third, photos with the same content but different exposures may not reach consistent adjustment results. To this end, we proposed a novel exposure correction network, ExReg, to address the challenges by formulating exposure correction as a multi-dimensional regression process. Given an input image, a compact multi-exposure generation network is introduced to generate images with different exposure conditions for multi-dimensional regression and exposure correction in the next stage. An auxiliary module is designed to predict the region-wise exposure values, guiding the proposed Encoder-Decoder ANP (Attentive Neural Processes) to regress the final corrected image. The experimental results show that ExReg can generate well-exposed results and outperform the SOTA method in PSNR for extensive exposure problems. Furthermore, the processing speed, with 0.05 seconds per image on an RTX 3090, is efficient. When tested on the same image under various exposure levels, ExReg also yields results that are visually consistent and physically accurate.

Shun-Yi Pan, Cheng-You Lu, Shih-Po Lee et al.

This work addresses weakly-supervised image semantic segmentation based on image-level class labels. One common approach to this task is to propagate the activation scores of Class Activation Maps (CAMs) using a random-walk mechanism in order to arrive at complete pseudo labels for training a semantic segmentation network in a fully-supervised manner. However, the feed-forward nature of the random walk imposes no regularization on the quality of the resulting complete pseudo labels. To overcome this issue, we propose a Graph Convolutional Network (GCN)-based feature propagation framework. We formulate the generation of complete pseudo labels as a semi-supervised learning task and learn a 2-layer GCN separately for every training image by back-propagating a Laplacian and an entropy regularization loss. Experimental results on the PASCAL VOC 2012 dataset confirm the superiority of our scheme to several state-of-the-art baselines. Our code is available at https://github.com/Xavier-Pan/WSGCN.

Cheng-Hsun Lei, Yi-Hsin Chen, Wen-Hsiao Peng et al.

In this paper, we address the problem of distillation-based class-incremental learning with a single head. A central theme of this task is to learn new classes that arrive in sequential phases over time while keeping the model's capability of recognizing seen classes with only limited memory for preserving seen data samples. Many regularization strategies have been proposed to mitigate the phenomenon of catastrophic forgetting. To understand better the essence of these regularizations, we introduce a feature-graph preservation perspective. Insights into their merits and faults motivate our weighted-Euclidean regularization for old knowledge preservation. We further propose rectified cosine normalization and show how it can work with binary cross-entropy to increase class separation for effective learning of new classes. Experimental results on both CIFAR-100 and ImageNet datasets demonstrate that our method outperforms the state-of-the-art approaches in reducing classification error, easing catastrophic forgetting, and encouraging evenly balanced accuracy over different classes. Our project page is at : https://github.com/yhchen12101/FGP-ICL.

Kai-Hsiang Hsieh, Monyneath Yim, Wen-Hsiao Peng et al.

Joint compression of point cloud geometry and attributes is essential for efficient 3D data representation. Existing methods often rely on post-hoc recoloring procedures and manually tuned bitrate allocation between geometry and attribute bitstreams in inference, which hinders end-to-end optimization and increases system complexity. To overcome these limitations, we propose MEGA-PCC, a fully end-to-end, learning-based framework featuring two specialized models for joint compression. The main compression model employs a shared encoder that encodes both geometry and attribute information into a unified latent representation, followed by dual decoders that sequentially reconstruct geometry and then attributes. Complementing this, the Mamba-based Entropy Model (MEM) enhances entropy coding by capturing spatial and channel-wise correlations to improve probability estimation. Both models are built on the Mamba architecture to effectively model long-range dependencies and rich contextual features. By eliminating the need for recoloring and heuristic bitrate tuning, MEGA-PCC enables data-driven bitrate allocation during training and simplifies the overall pipeline. Extensive experiments demonstrate that MEGA-PCC achieves superior rate-distortion performance and runtime efficiency compared to both traditional and learning-based baselines, offering a powerful solution for AI-driven point cloud compression.

Yu-Ting Zhan, Cheng-Yuan Ho, Hebi Yang et al.

3D Gaussian Splatting (3DGS) has recently emerged as a promising 3D representation. Much research has been focused on reducing its storage requirements and memory footprint. However, the needs to compress and transmit the 3DGS representation to the remote side are overlooked. This new application calls for rate-distortion-optimized 3DGS compression. How to quantize and entropy encode sparse Gaussian primitives in the 3D space remains largely unexplored. Few early attempts resort to the hyperprior framework from learned image compression. But, they fail to utilize fully the inter and intra correlation inherent in Gaussian primitives. Built on ScaffoldGS, this work, termed CAT-3DGS, introduces a context-adaptive triplane approach to their rate-distortion-optimized coding. It features multi-scale triplanes, oriented according to the principal axes of Gaussian primitives in the 3D space, to capture their inter correlation (i.e. spatial correlation) for spatial autoregressive coding in the projected 2D planes. With these triplanes serving as the hyperprior, we further perform channel-wise autoregressive coding to leverage the intra correlation within each individual Gaussian primitive. Our CAT-3DGS incorporates a view frequency-aware masking mechanism. It actively skips from coding those Gaussian primitives that potentially have little impact on the rendering quality. When trained end-to-end to strike a good rate-distortion trade-off, our CAT-3DGS achieves the state-of-the-art compression performance on the commonly used real-world datasets.

Xiem HoangVan, Hieu Bui Minh, Sang NguyenQuang et al.

The latest video coding standard H.266/VVC has shown its great improvement in terms of compression performance when compared to its predecessor HEVC standard. Though VVC was implemented with many advanced techniques, it still met the same challenges as its predecessor due to the need for even higher perceptual quality demand at the decoder side as well as the compression performance at the encoder side. The advancement of Artificial Intelligence (AI) technology, notably the deep learning-based video quality enhancement methods, was shown to be a promising approach to improving the perceptual quality experience. In this paper, we propose a novel Omniscient video quality enhancement Network for VVC compressed Videos. The Omniscient Network for compressed video quality enhancement was originally designed for HEVC compressed videos in which not only the spatial-temporal features but also cross-frequencies information were employed to augment the visual quality. Inspired by this work, we propose a modification of the OVQE model and integrate it into the lasted STD-VVC (Standard Versatile Video Coding) decoder architecture. As assessed in a rich set of test conditions, the proposed OVQE-VVC solution is able to achieve significant PSNR improvement, notably around 0.74 dB and up to 1.2 dB with respect to the original STD-VVC codec. This also corresponds to around 19.6% of bitrate saving while keeping a similar quality observation.

Yu-Ting Zhan, He-bi Yang, Cheng-Yuan Ho et al.

3D Gaussian Splatting (3DGS) has shown immense potential for novel view synthesis. However, achieving rate-distortion-optimized compression of 3DGS representations for transmission and/or storage applications remains a challenge. CAT-3DGS introduces a context-adaptive triplane hyperprior for end-to-end optimized compression, delivering state-of-the-art coding performance. Despite this, it requires prolonged training and decoding time. To address these limitations, we propose CAT-3DGS Pro, an enhanced version of CAT-3DGS that improves both compression performance and computational efficiency. First, we introduce a PCA-guided vector-matrix hyperprior, which replaces the triplane-based hyperprior to reduce redundant parameters. To achieve a more balanced rate-distortion trade-off and faster encoding, we propose an alternate optimization strategy (A-RDO). Additionally, we refine the sampling rate optimization method in CAT-3DGS, leading to significant improvements in rate-distortion performance. These enhancements result in a 46.6% BD-rate reduction and 3x speedup in training time on BungeeNeRF, while achieving 5x acceleration in decoding speed for the Amsterdam scene compared to CAT-3DGS.

Sang NguyenQuang, Zong-Lin Gao, Kuan-Wei Ho et al.

Most learned B-frame codecs with hierarchical temporal prediction suffer from the domain shift issue caused by the discrepancy in the Group-of-Pictures (GOP) size used for training and test. As such, the motion estimation network may fail to predict large motion properly. One effective strategy to mitigate this domain shift issue is to downsample video frames for motion estimation. However, finding the optimal downsampling factor involves a time-consuming rate-distortion optimization process. This work introduces lightweight classifiers to determine the downsampling factor. To strike a good rate-distortion-complexity trade-off, our classifiers observe simple state signals, including only the coding and reference frames, to predict the best downsampling factor. We present two variants that adopt binary and multi-class classifiers, respectively. The binary classifier adopts the Focal Loss for training, classifying between motion estimation at high and low resolutions. Our multi-class classifier is trained with novel soft labels incorporating the knowledge of the rate-distortion costs of different downsampling factors. Both variants operate as add-on modules without the need to re-train the B-frame codec. Experimental results confirm that they achieve comparable coding performance to the brute-force search methods while greatly reducing computational complexity.

Yu-Jen Tseng, Chia-Hao Kao, Jing-Zhong Chen et al.

We present the first unified framework for rate-distortion-optimized compression and segmentation of 3D Gaussian Splatting (3DGS). While 3DGS has proven effective for both real-time rendering and semantic scene understanding, prior works have largely treated these tasks independently, leaving their joint consideration unexplored. Inspired by recent advances in rate-distortion-optimized 3DGS compression, this work integrates semantic learning into the compression pipeline to support decoder-side applications--such as scene editing and manipulation--that extend beyond traditional scene reconstruction and view synthesis. Our scheme features a lightweight implicit neural representation-based hyperprior, enabling efficient entropy coding of both color and semantic attributes while avoiding costly grid-based hyperprior as seen in many prior works. To facilitate compression and segmentation, we further develop compression-guided segmentation learning, consisting of quantization-aware training to enhance feature separability and a quality-aware weighting mechanism to suppress unreliable Gaussian primitives. Extensive experiments on the LERF and 3D-OVS datasets demonstrate that our approach significantly reduces transmission cost while preserving high rendering quality and strong segmentation performance.

Cheng-Yuan Ho, He-Bi Yang, Jui-Chiu Chiang et al.

Building on the success of 3D Gaussian Splatting (3DGS) in static 3D scene representation, its extension to dynamic scenes, commonly referred to as 4DGS or dynamic 3DGS, has attracted increasing attention. However, designing more compact and efficient deformation schemes together with rate-distortion-optimized compression strategies for dynamic 3DGS representations remains an underexplored area. Prior methods either rely on space-time 4DGS with overspecified, short-lived Gaussian primitives or on canonical 3DGS with deformation that lacks explicit temporal control. To address this, we present TED-4DGS, a temporally activated and embedding-based deformation scheme for rate-distortion-optimized 4DGS compression that unifies the strengths of both families. TED-4DGS is built on a sparse anchor-based 3DGS representation. Each canonical anchor is assigned learnable temporal-activation parameters to specify its appearance and disappearance transitions over time, while a lightweight per-anchor temporal embedding queries a shared deformation bank to produce anchor-specific deformation. For rate-distortion compression, we incorporate an implicit neural representation (INR)-based hyperprior to model anchor attribute distributions, along with a channel-wise autoregressive model to capture intra-anchor correlations. With these novel elements, our scheme achieves state-of-the-art rate-distortion performance on several real-world datasets. To the best of our knowledge, this work represents one of the first attempts to pursue a rate-distortion-optimized compression framework for dynamic 3DGS representations.

Hao-Chien Hsueh, Chi-En Yen, Wen-Hsiao Peng et al.

Diffusion probabilistic models have achieved remarkable success in generative tasks across diverse data types. While recent studies have explored alternative degradation processes beyond Gaussian noise, this paper bridges two key diffusion paradigms: hot diffusion, which relies entirely on noise, and cold diffusion, which uses only blurring without noise. We argue that hot diffusion fails to exploit the strong correlation between high-frequency image detail and low-frequency structures, leading to random behaviors in the early steps of generation. Conversely, while cold diffusion leverages image correlations for prediction, it neglects the role of noise (randomness) in shaping the data manifold, resulting in out-of-manifold issues and partially explaining its performance drop. To integrate both strengths, we propose Warm Diffusion, a unified Blur-Noise Mixture Diffusion Model (BNMD), to control blurring and noise jointly. Our divide-and-conquer strategy exploits the spectral dependency in images, simplifying score model estimation by disentangling the denoising and deblurring processes. We further analyze the Blur-to-Noise Ratio (BNR) using spectral analysis to investigate the trade-off between model learning dynamics and changes in the data manifold. Extensive experiments across benchmarks validate the effectiveness of our approach for image generation.

Xiem HoangVan, Dang BuiDinh, Sang NguyenQuang et al.

Compressed video quality enhancement (CVQE) is crucial for improving user experience with lossy video codecs like H.264/AVC, H.265/HEVC, and H.266/VVC. While deep learning based CVQE has driven significant progress, existing surveys still suffer from limitations: lack of systematic classification linking methods to specific standards and artifacts, insufficient comparative analysis of architectural paradigms across coding types, and underdeveloped benchmarking practices. To address these gaps, this paper presents three key contributions. First, it introduces a novel taxonomy classifying CVQE methods across architectural paradigms, coding standards, and compressed-domain feature utilization. Second, it proposes a unified benchmarking framework integrating modern compression protocols and standard test sequences for fair multi-criteria evaluation. Third, it provides a systematic analysis of the critical trade-offs between reconstruction performance and computational complexity observed in state-of-the-art methods and highlighting promising directions for future research. This comprehensive review aims to establish a foundation for consistent assessment and informed model selection in CVQE research and deployment.

Sang NguyenQuang, Cheng-Wei Chen, Xiem HoangVan et al.

Learned video coding (LVC) has recently achieved superior coding performance. In this paper, we model the rate-quality (R-Q) relationship for learned video coding by a parametric function. We learn a neural network, termed RQNet, to characterize the relationship between the bitrate and quality level according to video content and coding context. The predicted (R,Q) results are further integrated with those from previously coded frames using the least-squares method to determine the parameters of our R-Q model on-the-fly. Compared to the conventional approaches, our method accurately estimates the R-Q relationship, enabling the online adaptation of model parameters to enhance both flexibility and precision. Experimental results show that our R-Q model achieves significantly smaller bitrate deviations than the baseline method on commonly used datasets with minimal additional complexity.

Chia-Hao Kao, Ying-Chieh Weng, Yi-Hsin Chen et al.

This paper proposes a transformer-based learned image compression system. It is capable of achieving variable-rate compression with a single model while supporting the region-of-interest (ROI) functionality. Inspired by prompt tuning, we introduce prompt generation networks to condition the transformer-based autoencoder of compression. Our prompt generation networks generate content-adaptive tokens according to the input image, an ROI mask, and a rate parameter. The separation of the ROI mask and the rate parameter allows an intuitive way to achieve variable-rate and ROI coding simultaneously. Extensive experiments validate the effectiveness of our proposed method and confirm its superiority over the other competing methods.

Yung-Han Ho, Chih-Chun Chan, Wen-Hsiao Peng et al.

This paper introduces an end-to-end learned image compression system, termed ANFIC, based on Augmented Normalizing Flows (ANF). ANF is a new type of flow model, which stacks multiple variational autoencoders (VAE) for greater model expressiveness. The VAE-based image compression has gone mainstream, showing promising compression performance. Our work presents the first attempt to leverage VAE-based compression in a flow-based framework. ANFIC advances further compression efficiency by stacking and extending hierarchically multiple VAE's. The invertibility of ANF, together with our training strategies, enables ANFIC to support a wide range of quality levels without changing the encoding and decoding networks. Extensive experimental results show that in terms of PSNR-RGB, ANFIC performs comparably to or better than the state-of-the-art learned image compression. Moreover, it performs close to VVC intra coding, from low-rate compression up to nearly-lossless compression. In particular, ANFIC achieves the state-of-the-art performance, when extended with conditional convolution for variable rate compression with a single model.

Yung-Han Ho, Guo-Lun Jin, Yun Liang et al.

This paper introduces a dual-critic reinforcement learning (RL) framework to address the problem of frame-level bit allocation in HEVC/H.265. The objective is to minimize the distortion of a group of pictures (GOP) under a rate constraint. Previous RL-based methods tackle such a constrained optimization problem by maximizing a single reward function that often combines a distortion and a rate reward. However, the way how these rewards are combined is usually ad hoc and may not generalize well to various coding conditions and video sequences. To overcome this issue, we adapt the deep deterministic policy gradient (DDPG) reinforcement learning algorithm for use with two critics, with one learning to predict the distortion reward and the other the rate reward. In particular, the distortion critic works to update the agent when the rate constraint is satisfied. By contrast, the rate critic makes the rate constraint a priority when the agent goes over the bit budget. Experimental results on commonly used datasets show that our method outperforms the bit allocation scheme in x265 and the single-critic baseline by a significant margin in terms of rate-distortion performance while offering fairly precise rate control.

Yan-Cheng Huang, Yi-Hsin Chen, Cheng-You Lu et al.

This paper addresses the video rescaling task, which arises from the needs of adapting the video spatial resolution to suit individual viewing devices. We aim to jointly optimize video downscaling and upscaling as a combined task. Most recent studies focus on image-based solutions, which do not consider temporal information. We present two joint optimization approaches based on invertible neural networks with coupling layers. Our Long Short-Term Memory Video Rescaling Network (LSTM-VRN) leverages temporal information in the low-resolution video to form an explicit prediction of the missing high-frequency information for upscaling. Our Multi-input Multi-output Video Rescaling Network (MIMO-VRN) proposes a new strategy for downscaling and upscaling a group of video frames simultaneously. Not only do they outperform the image-based invertible model in terms of quantitative and qualitative results, but also show much improved upscaling quality than the video rescaling methods without joint optimization. To our best knowledge, this work is the first attempt at the joint optimization of video downscaling and upscaling.

Shih-Po Lee, Si-Cun Chen, Wen-Hsiao Peng

This paper addresses fast semantic segmentation on video.Video segmentation often calls for real-time, or even fasterthan real-time, processing. One common recipe for conserving computation arising from feature extraction is to propagate features of few selected keyframes. However, recent advances in fast image segmentation make these solutions less attractive. To leverage fast image segmentation for furthering video segmentation, we propose a simple yet efficient propagation framework. Specifically, we perform lightweight flow estimation in 1/8-downscaled image space for temporal warping in segmentation outpace space. Moreover, we introduce a guided spatially-varying convolution for fusing segmentations derived from the previous and current frames, to mitigate propagation error and enable lightweight feature extraction on non-keyframes. Experimental results on Cityscapes and CamVid show that our scheme achieves the state-of-the-art accuracy-throughput trade-off on video segmentation.

Hui-Po Wang, Wen-Hsiao Peng, Wei-Jan Ko

Most deep latent factor models choose simple priors for simplicity, tractability or not knowing what prior to use. Recent studies show that the choice of the prior may have a profound effect on the expressiveness of the model,especially when its generative network has limited capacity. In this paper, we propose to learn a proper prior from data for adversarial autoencoders(AAEs). We introduce the notion of code generators to transform manually selected simple priors into ones that can better characterize the data distribution. Experimental results show that the proposed model can generate better image quality and learn better disentangled representations than AAEs in both supervised and unsupervised settings. Lastly, we present its ability to do cross-domain translation in a text-to-image synthesis task.

Yen-Wei Chang, Wen-Hsiao Peng

This paper tackles the problem of learning a questioner in the goal-oriented visual dialog task. Several previous works adopt model-free reinforcement learning. Most pretrain the model from a finite set of human-generated data. We argue that using limited demonstrations to kick-start the questioner is insufficient due to the large policy search space. Inspired by a recently proposed information theoretic approach, we develop two analytic experts to serve as a source of high-quality demonstrations for imitation learning. We then take advantage of reinforcement learning to refine the model towards the goal-oriented objective. Experimental results on the GuessWhat?! dataset show that our method has the combined merits of imitation and reinforcement learning, achieving the state-of-the-art performance.

David Alexandre, Chih-Peng Chang, Wen-Hsiao Peng et al.

This paper introduces the notion of soft bits to address the rate-distortion optimization for learning-based image compression. Recent methods for such compression train an autoencoder end-to-end with an objective to strike a balance between distortion and rate. They are faced with the zero gradient issue due to quantization and the difficulty of estimating the rate accurately. Inspired by soft quantization, we represent quantization indices of feature maps with differentiable soft bits. This allows us to couple tightly the rate estimation with context-adaptive binary arithmetic coding. It also provides a differentiable distortion objective function. Experimental results show that our approach achieves the state-of-the-art compression performance among the learning-based schemes in terms of MS-SSIM and PSNR.

Wei-Lun Chang, Hui-Po Wang, Wen-Hsiao Peng et al.

In this paper we tackle the problem of unsupervised domain adaptation for the task of semantic segmentation, where we attempt to transfer the knowledge learned upon synthetic datasets with ground-truth labels to real-world images without any annotation. With the hypothesis that the structural content of images is the most informative and decisive factor to semantic segmentation and can be readily shared across domains, we propose a Domain Invariant Structure Extraction (DISE) framework to disentangle images into domain-invariant structure and domain-specific texture representations, which can further realize image-translation across domains and enable label transfer to improve segmentation performance. Extensive experiments verify the effectiveness of our proposed DISE model and demonstrate its superiority over several state-of-the-art approaches.

David Alexandre, Chih-Peng Chang, Wen-Hsiao Peng et al.

We propose a lossy image compression system using the deep-learning autoencoder structure to participate in the Challenge on Learned Image Compression (CLIC) 2018. Our autoencoder uses the residual blocks with skip connections to reduce the correlation among image pixels and condense the input image into a set of feature maps, a compact representation of the original image. The bit allocation and bitrate control are implemented by using the importance maps and quantizer. The importance maps are generated by a separate neural net in the encoder. The autoencoder and the importance net are trained jointly based on minimizing a weighted sum of mean squared error, MS-SSIM, and a rate estimate. Our aim is to produce reconstructed images with good subjective quality subject to the 0.15 bits-per-pixel constraint.