LINC: 轻量化隐式帧间编码的近无损医学影像压缩框架

王续烨; 赵文博; 柏园超; 王凯; 刘贤明

doi:10.11999/JEIT260244

LINC: 轻量化隐式帧间编码的近无损医学影像压缩框架

doi: 10.11999/JEIT260244 cstr: 32379.14.JEIT260244

计算学部，哈尔滨工业大学哈尔滨市 150000

基金项目: 国家自然科学基金(62401176)

详细信息

通讯作者:
赵文博　wbzhao@hit.edu.cn

中图分类号: TN919.81
计量
- 文章访问数: 10
- HTML全文浏览量: 2
- PDF下载量: 0
- 被引次数: 0
出版历程
- 收稿日期: 2026-03-06
- 修回日期: 2026-10-20
- 录用日期: 2026-10-28
- 网络出版日期: 2026-10-31

LINC: Lightweight Implicit Inter-Frame Coding Framework for Near-Lossless Medical Image Compression

The Faculty of Computing, Harbin Institute of Technology, Harbin 150000, China

Funds: Natural Science Foundation of China under Grant (62401176)

摘要

摘要: 在AI驱动的智能医疗体系与6G超高速网络基础设施快速发展的背景下，大规模医学影像数据的高效传输与存储成为制约远程诊疗与智能辅助决策的重要瓶颈。传统无损压缩方法受限于信息熵下界，压缩率提升空间有限；而现有近无损压缩方法多针对单帧图像独立建模，未能充分利用医学影像序列中的帧间相关性，导致压缩效率仍有提升空间。针对上述问题，本文提出一种基于隐式帧间建模的轻量化近无损医学影像压缩框架 LINC。在严格的误差约束条件下，该方法将有损编码与残差近无损编码统一于概率建模框架之中，并通过引入跨帧条件建模机制增强条件概率估计能力。在有损阶段，利用参考帧构建跨帧特征融合网络，提升时序冗余去除能力；在残差阶段，通过区间量化与块级并行自回归建模实现误差受控的高效编码。实验结果表明，在多个医学影像数据集上，本文方法相较现有近无损方法平均实现约15%的压缩性能提升，为AI与6G融合背景下的智能医疗提供了一种高效的数据传输途径。
- 近无损压缩 /
- 帧间编码 /
- 概率建模 /
- 医学影像压缩
Abstract: Objective With the rapid advancement of AI-enabled healthcare systems and high-speed communication networks, the explosive growth of medical imaging data from large-scale CT and MRI examinations imposes significant pressure on storage and transmission infrastructures. Unlike natural images, medical images require strict diagnostic fidelity, making irreversible compression schemes unsuitable due to potential clinical risks. Although lossless compression guarantees exact reconstruction, it is fundamentally limited by the entropy bound, resulting in modest compression ratios. Near-lossless compression relaxes this constraint by bounding per-pixel reconstruction error within a predefined threshold τ, enabling improved coding efficiency while preserving diagnostic reliability. However, existing near-lossless methods typically model each frame independently and fail to fully exploit inter-frame correlations inherent in volumetric medical image sequences. To address these limitations, a lightweight implicit inter-frame near-lossless compression framework, termed LINC, is proposed. The framework unifies lossy modeling and near-lossless residual coding within a probabilistic modeling paradigm while effectively leveraging temporal dependencies across adjacent slices. Methods Given a medical image sequence, the objective is to achieve near-lossless compression under a strict pixel-wise error bound τ while maintaining lightweight model complexity. The proposed framework consists of two stages: lossy inter-frame coding and near-lossless residual coding. In the lossy stage, the previously decoded frame is used as a reference to construct a cross-frame conditional modeling mechanism. Instead of block-wise independent compression, full-frame feature fusion is performed to capture long-range structural correlations. Inter-frame dependencies are implicitly represented through learned features and conditional probability distributions, without explicitly estimating or transmitting motion vectors, optical flow, or deformation fields. A lightweight encoder–decoder architecture with attention modules extracts fused representations from the current and reference frames (Fig. 2). A hyperprior model is employed to estimate latent variable distributions, and arithmetic coding is applied to encode quantized latent representations. The theoretical rate decomposition follows probabilistic modeling principles, where the total rate is expressed as the sum of latent representation cost and conditional residual cost (Eqs. 4–5). After lossy reconstruction is obtained, it is restored to the original pixel scale, rounded to the nearest integer, and clipped to the valid pixel range before residual calculation (Eq. 6). The residual is partitioned into blocks of size L×L to enable parallel processing. To satisfy the near-lossless constraint, interval quantization is applied (Eq. 7), ensuring that reconstruction error does not exceed τ. Within each block, a parallel autoregressive scanning order is adopted (Fig. 3), and masked convolution is used to extract contextual information from previously encoded residuals. The conditional probability mass function of residuals is modeled using a discretized logistic mixture model (Eq. 9), and the near-lossless PMF is derived through interval aggregation (Eq. 10). Arithmetic coding is then applied to encode quantized residuals. At the decoder, the recovered quantized residuals are added to the integer-valued lossy reconstruction, and the result is clipped to the valid pixel range without increasing the reconstruction error. During training, both lossless and near-lossless branches are retained to stabilize probability estimation (Eqs. 12–13). A fixed GOP structure (I–P–P–…–P) with length S=8 is adopted during deployment, where I-frames are encoded using a pretrained single-frame model and P-frames utilize LINC. Dynamic residual range adjustment is introduced to accelerate entropy coding. Result and Discussion Extensive experiments are conducted on four medical datasets: MosMed, CHAOS, MRNet, and ACDC (Section 4.1). Near-lossless compression performance under τ=1–4 is compared with JPEG-LS, CALIC, and DLPR (Table 1). The proposed method consistently achieves the lowest average bits per pixel (bpp) across all distortion levels. Across the tested distortion levels, DLPR’s mean bpp is approximately 15% higher than that of LINC on average. Performance gain varies across datasets due to different levels of inter-frame correlation. Visualization of adjacent-frame differences and reconstruction residuals (Fig. 4) illustrates weaker inter-frame correlation in the selected MRNet example than in the MosMed example, consistent with the smaller improvement observed on MRNet. The measured maximum absolute reconstruction errors remain within the prescribed thresholds on all four medical datasets. In strict lossless mode, comparisons with JPEG2000, JPEG-XL, HEVC, JPEG-LS, CALIC, and DLPR demonstrate that LINC achieves the lowest average bpp, with DLPR’s mean bpp approximately 7.6% higher than that of LINC (Table 2), verifying the effectiveness of the framework in both compression modes. Complexity and resource analysis (Table 3) shows that LINC increases parameter counts and peak GPU memory usage, while reducing the floating-point operation count of lossy encoding and slightly increasing that of lossy decoding. Nevertheless, the measured average encoding and decoding times are lower than those of DLPR (Table 4). Rate allocation analysis (Table 5) further indicates that inter-frame modeling reduces both latent and residual coding costs in P-frames. An ablation study on residual block size (Table 6) reveals the trade-off between parallel efficiency and modeling accuracy. Increasing the block size from 64 to 256 reduces residual bpp by only approximately 0.09%, while increasing residual encoding and decoding times by approximately 279% and 283%, respectively. Therefore, L=64 is selected as a practical compromise between efficiency and performance. Boundary and non-boundary regions exhibit similar MAE and PSNR values on the evaluated CHAOS sequence, with no pronounced concentration of reconstruction error at block boundaries (Table 7). Further ablation experiments show that full-frame modeling without inter-frame information improves compression over DLPR, that introducing inter-frame information into either stage provides additional gains, and that using it in both stages achieves the lowest average bpp (Table 8). Experiments on BVI-DVC, JHMDB, and Xiph demonstrate consistently lower bpp than DLPR across the tested distortion thresholds. However, LINC remains inferior to CALIC on Xiph under τ=1–4, indicating limitations in its applicability to natural video compression (Table 9). Conclusions This paper proposes LINC, a lightweight implicit inter-frame near-lossless compression framework for medical images. By unifying lossy coding and residual near-lossless modeling within a probabilistic framework and introducing cross-frame conditional modeling, the proposed method effectively exploits temporal redundancy in medical image sequences. Experimental results demonstrate that LINC achieves lower average bpp and shorter average encoding and decoding times than DLPR, although it uses more parameters and slightly more peak GPU memory. Furthermore, the framework supports strict lossless compression with measurable gains. Although improvements over DLPR are also observed on natural video datasets, its competitiveness against traditional codecs remains dataset-dependent. The proposed approach provides a practical and efficient solution for large-scale medical image storage and transmission under strict per-pixel reconstruction error constraints.
- Near-lossless compression /
- Inter-frame coding /
- Probabilistic modeling /
- Medical image compression

HTML全文

图 1 本文提出的编码框架，由有损编码与残差编码两部分组成。

下载: 全尺寸图片幻灯片

图 2 有损编码器的结构

下载: 全尺寸图片幻灯片

图 3 $ L=9 $时的编码顺序与掩膜卷积示意图，数字代表编码顺序，蓝色表示已编码的残差，红色表示当前待编码的残差，黄色表示掩膜卷积范围内的残差。

下载: 全尺寸图片幻灯片

图 4 $ \tau =4 $时，MosMed与MRNet的帧间差异与重建残差对比图

下载: 全尺寸图片幻灯片

表 1 不同方法的近无损压缩bpp指标对比，最好的结果加粗表示

失真上界	方法	MosMed	CHAOS	MRNet	ACDC	平均		最大失真
$ \tau =1 $	JPEG-LS	3.74	3.82	3.26	2.27	3.27	+12.0%	1
	CALIC	4.24	4.06	3.11	2.13	3.39	+16.1%	1
	DLPR	4.29	3.76	2.82	2.30	3.29	+12.7%	1
	Ours	3.58	3.14	2.82	2.12	2.92		1
$ \tau =2 $	JPEG-LS	3.19	3.14	2.61	1.79	2.68	+15.5%	2
	CALIC	3.73	3.40	2.45	1.61	2.80	+20.7%	2
	DLPR	3.66	3.04	2.21	1.72	2.66	+14.7%	2
	Ours	2.97	2.52	2.20	1.57	2.32		2
$ \tau =3 $	JPEG-LS	2.85	2.73	2.21	1.52	2.33	+19.0%	3
	CALIC	3.33	2.97	2.06	1.33	2.42	+24.1%	3
	DLPR	3.12	2.69	1.84	1.39	2.26	+15.9%	3
	Ours	2.58	2.12	1.81	1.28	1.95		3
$ \tau =4 $	JPEG-LS	2.59	2.43	1.92	1.34	2.07	+20.3%	4
	CALIC	3.09	2.69	1.79	1.15	2.18	+26.7%	4
	DLPR	2.69	2.50	1.58	1.20	1.99	+15.7%	4
	Ours	2.32	1.89	1.55	1.10	1.72		4

下载: 导出CSV

表 2 不同方法的无损压缩bpp指标对比，最好的结果加粗表示

方法	MosMed	CHAOS	MRNet	ACDC	平均		最大失真
JPEG2000	5.32	5.27	4.85	3.73	4.79	+14.2%	0
JPEG-XL	4.76	5.04	4.59	3.29	4.42	+5.3%	0
HEVC	6.63	6.79	5.67	4.77	5.97	+42.1%	0
JPEG-LS	4.96	5.35	4.74	3.52	4.64	+10.5%	0
CALIC	5.49	5.58	4.61	3.41	4.77	+13.6%	0
DLPR	5.07	5.29	4.27	3.44	4.52	+7.6%	0
Ours	4.72	4.52	4.25	3.28	4.20		0

下载: 导出CSV

表 3 本文方法与DLPR的编解码复杂度及资源开销对比

模型	参数量(M)		平均有损GFLOPs		平均残差GFLOPs		峰值显存(MiB)
DLPR-编码	22.2	–47%	741.3	+33%	216.9	+0.15%	734.2	–3.67%
Ours-编码	42.0		558.5		216.6		762.2
DLPR-解码	15.6	–44%	466.4	–2.5%	1459.9	+0.02%	708.6	–6.86%
Ours-解码	27.9		478.4		1459.6		760.8

下载: 导出CSV

表 4 本文方法与DLPR的编解码时间对比

失真	方法	编码(秒)				解码(秒)
失真	方法	有损		残差		有损		残差
$ \tau =1 $	DLPR	0.046	+0.016	0.330	+0.027	0.022	+0.008	0.453	+0.036
$ \tau =1 $	Ours	0.030		0.303		0.015		0.417
$ \tau =2 $	DLPR	0.046	+0.017	0.305	+0.025	0.023	+0.008	0.424	+0.032
$ \tau =2 $	Ours	0.030		0.280		0.015		0.391
$ \tau =3 $	DLPR	0.046	+0.017	0.292	+0.022	0.022	+0.008	0.410	+0.030
$ \tau =3 $	Ours	0.030		0.270		0.015		0.380
$ \tau =4 $	DLPR	0.046	+0.016	0.285	+0.018	0.022	+0.007	0.402	+0.025
$ \tau =4 $	Ours	0.030		0.267		0.015		0.377

下载: 导出CSV

表 5 本文方法与DLPR在不同阶段的bpp对比

失真	方法	有损	残差	总和
$ \tau =1 $	DLPR	0.440	3.851	4.292
	Ours-I帧	0.447	3.886	4.333
	Ours-P帧	0.207	3.277	3.484
$ \tau =2 $	DLPR	0.440	3.223	3.664
	Ours-I帧	0.447	3.256	3.704
	Ours-P帧	0.207	2.667	2.875
$ \tau =3 $	DLPR	0.438	2.682	3.121
	Ours-I帧	0.447	2.654	3.101
	Ours-P帧	0.207	2.308	2.515
$ \tau =4 $	DLPR	0.440	2.247	2.688
	Ours-I帧	0.447	2.280	2.728
	Ours-P帧	0.207	2.057	2.265

下载: 导出CSV

表 6 CHAOS数据集在不同块尺寸时的残差编码性能与速度对比

块尺寸	残差编码时间(秒)		残差解码时间(秒)		残差bpp
$ L=64 $	0.295		0.481		2.077
$ L=128 $	0.532	+80%	0.891	+85%	2.072	–0.21%
$ L=256 $	1.118	+279%	1.842	+283%	2.075	–0.09%

下载: 导出CSV

表 7 不同失真阈值下块边界区域与非边界区域的重建质量对比

失真	区域	PSNR(dB)		MAE
$ \tau =1 $	边界	74.006	–0.014	0.667	+0.003
$ \tau =1 $	非边界	74.020		0.664
$ \tau =2 $	边界	69.237	–0.014	1.199	+0.003
$ \tau =2 $	非边界	69.251		1.196
$ \tau =3 $	边界	66.232	–0.013	1.712	+0.005
$ \tau =3 $	非边界	66.245		1.707
$ \tau =4 $	边界	64.032	–0.021	2.213	+0.010
$ \tau =4 $	非边界	64.053		2.203

下载: 导出CSV

表 8 MosMed数据集在不同网络结构下的bpp对比，最好的结果加粗

失真上界	DLPR	无帧间整帧	仅有损阶段	仅残差阶段	原始模型
$ \tau =1 $	4.29	3.89	3.70	3.69	3.58
$ \tau =2 $	3.66	3.40	3.00	3.04	2.97
$ \tau =3 $	3.12	2.84	2.61	2.61	2.58
$ \tau =4 $	2.69	2.63	2.36	2.34	2.32
平均	3.44	3.19	2.92	2.92	2.86

下载: 导出CSV

表 9 不同方法在自然视频数据集上的性能对比，最好的结果加粗表示

失真上界	方法	BVI-DVC	JHMDB	Xiph	平均		最大失真
$ \tau =0 $	JPEG-LS	4.07	3.34	4.24	3.88	+13.2%	0
	CALIC	3.90	3.14	4.08	3.71	+8.2%	0
	DLPR	4.47	3.24	3.99	3.90	+13.8%	0
	Ours	3.65	2.77	3.86	3.43		0
$ \tau =1 $	JPEG-LS	2.74	2.24	2.83	2.60	+10.4%	1
	CALIC	2.57	2.05	2.67	2.43	+3.0%	1
	DLPR	3.29	2.37	3.58	3.08	+30.7%	1
	Ours	2.27	1.73	3.08	2.36		1
$ \tau =2 $	JPEG-LS	2.21	1.80	2.25	2.08	+19.0%	2
	CALIC	2.01	1.58	2.08	1.89	+7.9%	2
	DLPR	2.76	1.92	3.00	2.56	+46.2%	2
	Ours	1.73	1.23	2.29	1.75		2
$ \tau =3 $	JPEG-LS	1.89	1.54	1.90	1.78	+23.7%	3
	CALIC	1.70	1.31	1.76	1.59	+10.5%	3
	DLPR	2.50	1.74	2.74	2.33	+61.9%	3
	Ours	1.42	1.01	1.88	1.44		3
$ \tau =4 $	JPEG-LS	1.67	1.37	1.68	1.57	+25.8%	4
	CALIC	1.47	1.13	1.52	1.38	+10.0%	4
	DLPR	2.36	1.66	2.63	2.21	+77.2%	4
	Ours	1.23	0.88	1.64	1.25		4

下载: 导出CSV

参考文献(34)

[1]	WALLACE G K. The JPEG still picture compression standard[J]. Communications of the ACM, 1991, 34(4): 30–44. doi: 10.1145/103085.103089.
[2]	SULLIVAN G J, OHM J R, HAN W J, et al. Overview of the high efficiency video coding (HEVC) standard[J]. IEEE Transactions on Circuits and Systems for Video Technology, 2012, 22(12): 1649–1668. doi: 10.1109/TCSVT.2012.2221191.
[3]	CHEN Yue, MURHERJEE D, HAN Jingning, et al. An overview of core coding tools in the AV1 video codec[C]. Proceedings of the 2018 Picture Coding Symposium, San Francisco, USA, 2018: 41–45. doi: 10.1109/PCS.2018.8456249.
[4]	盛庆华, 陶泽浩, 黄小芳, 等. 一种面向AV1粗模式决策的高吞吐量硬件设计方法[J]. 电子与信息学报, 2025, 47(4): 1202–1214. doi: 10.11999/JEIT240823. SHENG Qinghua, TAO Zehao, HUANG Xiaofang, et al. A high-throughput hardware design for AV1 rough mode decision[J]. Journal of Electronics & Information Technology, 2025, 47(4): 1202–1214. doi: 10.11999/JEIT240823.
[5]	唐述, 周广义, 谢显中, 等. 一种快速的多尺度多输入编码树单元互补分类网络[J]. 电子与信息学报, 2024, 46(9): 3646–3653. doi: 10.11999/JEIT240223. TANG Shu, ZHOU Guangyi, XIE Xianzhong, et al. A multi-scale-multi-input complementation classification network for fast coding tree unit partition[J]. Journal of Electronics & Information Technology, 2024, 46(9): 3646–3653. doi: 10.11999/JEIT240223.
[6]	SHANNON C E. A mathematical theory of communication[J]. The Bell System Technical Journal, 1948, 27(3): 379–423. doi: 10.1002/j.1538-7305.1948.tb01338.x.
[7]	WEINBERGER M J, SEROUSSI G, and SAPIRO G. The LOCO-I lossless image compression algorithm: Principles and standardization into JPEG-LS[J]. IEEE Transactions on Image Processing, 2000, 9(8): 1309–1324. doi: 10.1109/83.855427.
[8]	WU Xiaolin and MEMON N. CALIC-a context based adaptive lossless image codec[C]. Proceedings of the 1996 IEEE International Conference on Acoustics, Speech, and Signal Processing Conference Proceedings, Atlanta, USA, 1996: 1890–1893. doi: 10.1109/ICASSP.1996.544819.
[9]	柏园超, 刘文昌, 江俊君, 等. 深度神经网络图像压缩方法进展综述[J]. 电子与信息学报, 2025, 47(11): 4112–4128. doi: 10.11999/JEIT250567. BAI Yuanchao, LIU Wenchang, JIANG Junjun, et al. Advances in deep neural network based image compression: A survey[J]. Journal of Electronics & Information Technology, 2025, 47(11): 4112–4128. doi: 10.11999/JEIT250567.
[10]	LIAO Hengrui and LI Yue. LFC-UNet: Learned lossless medical image fast compression with U-Net[J]. PeerJ Computer Science, 2024, 10: e1924. doi: 10.7717/peerj-cs.1924.
[11]	ZHANG Zhe, WANG Huairui, CHEN Zhenzhong, et al. Learned lossless image compression based on bit plane slicing[C]. Proceedings of the 2024 IEEE/CVF Conference on Computer Vision and Pattern Recognition, Seattle, USA, 2024: 27569–27578. doi: 10.1109/CVPR52733.2024.02604.
[12]	LI Daxin, BAI Yuanchao, WANG Kai, et al. CALLIC: Content adaptive learning for lossless image compression[C]. Proceedings of the 39th Annual AAAI Conference on Artificial Intelligence, Philadelphia, USA, 2025: 4679–4688. doi: 10.1609/aaai.v39i5.32494.
[13]	WANG Kai, BAI Yuanchao, LI Daxin, et al. Learning lossless compression for high bit-depth volumetric medical image[J]. IEEE Transactions on Image Processing, 2025, 34: 113–125. doi: 10.1109/TIP.2024.3513156.
[14]	CHEN Jietao, CHEN Weijie, XING Qianjian, et al. Multi-scale contextual coding for human-machine vision of volumetric medical images[J]. IEEE Access, 2025, 13: 145663–145679. doi: 10.1109/ACCESS.2025.3597008.
[15]	LIU Xiangrui, WANG Meng, WANG Shiqi, et al. Bilateral context modeling for residual coding in lossless 3D medical image compression[J]. IEEE Transactions on Image Processing, 2024, 33: 2502–2513. doi: 10.1109/TIP.2024.3378910.
[16]	XUE Dongmei, MA Haichuan, LI Li, et al. aiWave: Volumetric image compression with 3-D trained affine wavelet-like transform[J]. IEEE Transactions on Medical Imaging, 2023, 42(3): 606–618. doi: 10.1109/TMI.2022.3212780.
[17]	WANG Kai, BAI Yuanchao, LI Daxin, et al. 3D-SLARM: Practical lossless volumetric image compression via a 3D-scanning lightweight autoregressive model[J]. IEEE Transactions on Image Processing, 2026, 35: 1174–1189. doi: 10.1109/TIP.2025.3648557.
[18]	黄胜, 向思皓, 胡峰, 等. 自适应不规则纹理的胶囊内镜图像无损压缩[J]. 计算机工程, 2022, 48(10): 21–27,36. doi: 10.19678/j.issn.1000-3428.0063208. HUANG Sheng, XIANG Sihao, HU Feng, et al. Adaptive lossless compression of capsule endoscopy image with irregular texture[J]. Computer Engineering, 2022, 48(10): 21–27,36. doi: 10.19678/j.issn.1000-3428.0063208.
[19]	薛冬梅. 基于类小波变换的生物医学图像和视频智能编码研究[D]. [博士论文], 中国科学技术大学, 2025. doi: 10.27517/d.cnki.gzkju.2025.000809. XUE Dongmei. Research on biomedical image and video intelligent coding based on wavelet-like transform[D]. [Ph. D. dissertation], University of Science and Technology of China, 2025. doi: 10.27517/d.cnki.gzkju.2025.000809.
[20]	BAI Yuanchao, LIU Xianming, WANG Kai, et al. Deep lossy plus residual coding for lossless and near-lossless image compression[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2024, 46(5): 3577–3594. doi: 10.1109/TPAMI.2023.3348486.
[21]	KINGMA D P and WELLING M. Auto-encoding variational bayes[C]. Proceedings of the 2nd International Conference on Learning Representations, Banff, Canada, 2014.
[22]	JIA Zhaoyang, LI Bin, LI Jiahao, et al. Towards practical real-time neural video compression[C]. Proceedings of the 2025 IEEE/CVF Conference on Computer Vision and Pattern Recognition, Nashville, USA, 2025: 12543–12552. doi: 10.1109/CVPR52734.2025.01170.
[23]	LIU Ze, LIN Yutong, CAO Yue, et al. Swin transformer: Hierarchical vision transformer using shifted windows[C]. Proceedings of the 2021 IEEE/CVF International Conference on Computer Vision, Montreal, Canada, 2021: 9992–10002. doi: 10.1109/ICCV48922.2021.00986.
[24]	BALLÉ J, MINNEN D, SINGH S, et al. Variational image compression with a scale hyperprior[C]. Proceedings of the 6th International Conference on Learning Representations, Vancouver, Canada, 2018.
[25]	SALIMANS T, KARPATHY A, CHEN Xi, et al. PixelCNN++: Improving the PixelCNN with discretized logistic mixture likelihood and other modifications[C]. Proceedings of the 5th International Conference on Learning Representations, Toulon, France, 2017.
[26]	MOROZOV S P, ANDREYCHENKO A E, BLOKHIN I A, et al. MosMedData: Data set of 1110 chest CT scans performed during the COVID-19 epidemic[J]. Digital Diagnostics, 2020, 1(1): 49–59. doi: 10.17816/DD46826.
[27]	KAVUR A E, GEZER N S, BARIŞ M, et al. CHAOS challenge-combined (CT-MR) healthy abdominal organ segmentation[J]. Medical Image Analysis, 2021, 69: 101950. doi: 10.1016/j.media.2020.101950.
[28]	BIEN N, RAJPURKAR P, BALL R L, et al. Deep-learning-assisted diagnosis for knee magnetic resonance imaging: Development and retrospective validation of MRNet[J]. PLoS Medicine, 2018, 15(11): e1002699. doi: 10.1371/journal.pmed.1002699.
[29]	BERNARD O, LALANDE A, ZOTTI C, et al. Deep learning techniques for automatic MRI cardiac multi-structures segmentation and diagnosis: Is the problem solved?[J]. IEEE Transactions on Medical Imaging, 2018, 37(11): 2514–2525. doi: 10.1109/TMI.2018.2837502.
[30]	SKODRAS A, CHRISTOPOULOS C, and EBRAHIMI T. The JPEG 2000 still image compression standard[J]. IEEE Signal Processing Magazine, 2001, 18(5): 36–58. doi: 10.1109/79.952804.
[31]	ALAKUIJALA J, VAN ASSELDONK R, BOUKORTT S, et al. JPEG XL next-generation image compression architecture and coding tools[C]. Proceedings of SPIE 11137, Applications of Digital Image Processing XLII, San Diego, USA, 2019: 111370K. doi: 10.1117/12.2529237.
[32]	MA Di, ZHANG Fan, and BULL D R. BVI-DVC: A training database for deep video compression[J]. IEEE Transactions on Multimedia, 2022, 24: 3847–3858. doi: 10.1109/TMM.2021.3108943.
[33]	JHUANG H, GALL J, ZUFFI S, et al. Towards understanding action recognition[C]. Proceedings of the 2013 IEEE International Conference on Computer Vision, Sydney, Australia, 2013: 3192–3199. doi: 10.1109/ICCV.2013.396.
[34]	Xiph. org Foundation. Xiph. org Video test media: Derf’s collection[EB/OL]. https://media.xiph.org/video/derf/, 2026. (查阅网上资料,未找到本条文献年份信息,请确认).