基于熵掩蔽的DCT域恰可察觉失真模型

doi:10.11959/j.issn.1000-0801.2023014

Abstract

Abstract:

In order to improve the threshold accuracy of JND (just noticeable distortion) model in DCT (discrete cosine transform) domain and avoid cross-domain operation, entropy masking effect was introduced into DCT-based JND model.Firstly, starting from the free-energy theory and the Bayesian inference, an autoregressive model based on texture-energy similarity in DCT domain was designed to simulate the spontaneous prediction behavior of visual perception.Secondly, the mapping relationship between visual perception and prediction residuals were explored to obtain the disorder intensity in block level.Thirdly, the entropy masking effect was modeled as a JND threshold modulation factor of disorder intensity.Finally, the JND model in DCT domain for the entropy masking was proposed by fusing the contrast sensitivity function, the luminance adaptive masking, the contrast masking.Compared with the existing JND model in DCT domain, the proposed model performed all operations in DCT domain, which was more efficient and concise.The subjective and objective experimental results indicate that the proposed JND model shows greater tolerance to distortion with better perceptual quality.

Key words: JND, human visual system, entropy masking effect, free-energy theory, Bayesian inference

CLC Number:

TN91

Qionghua LUO, Hongkui WANG, Haibing YIN, Yafen XING. Just noticeable distortion model based on entropy masking in DCT domain[J]. Telecommunications Science, 2023, 39(2): 59-70.

Figures/Tables 13

IVC	PSNR				MS-SSIM指数				MOS
IVC	Zeng2019	Liu2020	Li2022	本文	Zeng2019	Liu2020	Li2022	本文	Zeng2019	Liu2020	Li2022	本文
I₀₁	29.80	30.92	31.44	26.19	0.947	0.962	0.977	0.959	-0.61	-0.38	-0.39	-0.20
I₀₂	29.25	28.39	27.79	25.26	0.964	0.966	0.962	0.968	-0.52	-0.50	-0.38	-0.25
I₀₃	30.15	30.02	30.52	25.98	0.961	0.964	0.973	0.962	-0.56	-0.50	-0.36	-0.20
I₀₄	27.85	27.65	25.93	26.04	0.945	0.944	0.930	0.968	-0.52	-0.47	-0.42	-0.29
I₀₅	28.20	28.56	26.77	26.61	0.956	0.963	0.956	0.969	-0.47	-0.37	-0.40	-0.24
I₀₆	30.01	29.42	30.38	25.77	0.964	0.971	0.981	0.966	-0.46	-0.41	-0.47	-0.30
I₀₇	29.31	30.16	29.49	26.67	0.956	0.967	0.968	0.967	-0.54	-0.46	-0.31	-0.19
I₀₈	29.65	30.48	28.86	26.72	0.957	0.968	0.963	0.980	-0.54	-0.46	-0.36	-0.28
I₀₉	30.14	26.22	28.98	25.24	0.982	0.975	0.984	0.964	-0.29	-0.43	-0.28	-0.22
I₁₀	28.90	28.73	27.99	26.86	0.949	0.959	0.951	0.964	-0.41	-0.45	-0.36	-0.23
平均值	29.33	29.05	28.81	$26 . 13$	0.958	0.964	0.965	$0 . 967$	-0.49	-0.44	-0.37	$- 0 . 24$

TID2013	PSNR				MS-SSIM指数				MOS
TID2013	Zeng2019	Liu2020	Li2022	本文	Zeng2019	Liu2020	Li2022	本文	Zeng2019	Liu2020	Li2022	本文
T₀₁	29.50	29.24	27.72	26.83	0.932	0.932	0.910	0.954	-0.50	-0.39	-0.54	-0.25
T₀₂	28.92	29.31	27.40	27.12	0.934	0.941	0.921	0.959	-0.43	-0.28	-0.54	-0.30
T₀₃	29.42	29.27	27.43	26.27	0.972	0.974	0.965	0.978	-0.41	-0.37	-0.46	-0.31
T₀₄	28.37	26.53	26.41	24.56	0.974	0.975	0.968	0.972	-0.40	-0.43	-0.37	-0.27
T₀₅	29.91	29.73	29.20	26.24	0.954	0.963	0.962	0.961	-0.46	-0.38	-0.49	-0.30
T₀₆	30.08	28.73	29.37	26.44	0.961	0.969	0.971	0.963	-0.64	-0.40	-0.51	-0.35
T₀₇	27.12	26.84	25.18	26.65	0.904	0.903	0.874	0.954	-0.49	-0.28	-0.51	-0.28
T₀₈	30.47	30.28	29.19	26.52	0.957	0.962	0.955	0.958	-0.43	-0.36	-0.45	-0.28
T₀₉	30.01	28.11	28.26	25.73	0.972	0.974	0.976	0.967	-0.44	-0.37	-0.43	-0.33
T₁₀	30.15	30.17	29.37	26.91	0.950	0.955	0.950	0.958	-0.44	-0.27	-0.47	-0.27
平均值	29.40	28.82	27.95	$26 . 33$	0.951	0.955	0.945	$0 . 962$	-0.46	-0.35	-0.48	$- 0 . 29$

LabelMe	PSNR				MS-SSIM指数				MOS
LabelMe	Zeng2019	Liu2020	Li2022	本文	Zeng2019	Liu2020	Li2022	本文	Zeng2019	Liu2020	Li2022	本文
L₀₁	31.88	31.21	32.78	29.68	0.946	0.943	0.966	0.980	-0.45	-0.35	-0.56	-0.40
L₀₂	27.25	26.95	25.82	27.21	0.883	0.877	0.866	0.946	-0.26	-0.55	-0.52	-0.35
L₀₃	30.57	30.86	29.56	28.50	0.951	0.952	0.957	0.976	-0.43	-0.45	-0.46	-0.42
L₀₄	29.27	28.80	27.96	27.23	0.945	0.943	0.946	0.967	-0.57	-0.45	-0.48	-0.34
L₀₅	29.13	28.55	27.63	27.88	0.935	0.932	0.924	0.964	-0.47	-0.49	-0.47	-0.34
L₀₆	25.72	24.84	23.96	27.50	0.841	0.830	0.810	0.947	-0.4	-0.47	-0.50	-0.42
L₀₇	31.33	31.25	30.41	28.92	0.957	0.955	0.960	0.970	-0.45	-0.40	-0.42	-0.29
L₀₈	30.35	30.12	29.22	27.14	0.959	0.960	0.959	0.960	-0.5	-0.56	-0.63	-0.32
L₀₉	28.68	28.47	26.89	27.09	0.914	0.914	0.891	0.947	-0.42	-0.53	-0.46	-0.33
L₁₀	26.71	26.26	24.91	26.90	0.872	0.865	0.839	0.939	-0.44	-0.57	-0.59	-0.27
平均值	29.09	28.73	27.91	$26 . 80$	0.920	0.917	0.912	$0 . 960$	-0.44	-0.48	-0.51	$- 0 . 35$
注：加粗字体为最优结果。

References 26

[1]	JAYANT N , JOHNSTON J , SAFRANEK R . Signal compression based on models of human perception[J]. Proceedings of the IEEE, 1993,81(10): 1385-1422.
[2]	KI S , DO J , KIM M . Learning-based JND-directed HDR video preprocessing for perceptually lossless compression with HEVC[J]. IEEE Access, 2020(8): 228605-228618.
[3]	崔帅南, 彭宗举, 邹文辉 ,等. 多特征融合的合成视点立体图像质量评价[J]. 电信科学, 2019,35(5): 104-112.
	CUI S N , PENG Z J , ZOU W H ,et al. Quality assessment of synthetic viewpoint stereo image with multi-feature fusion[J]. Telecommunications Science, 2019,35(5): 104-112.
[4]	LI J , ZHANG H , WANG J ,et al. Orientation-aware saliency guided JND model for robust image watermarking[J]. IEEE Access, 2019(7): 41261-41272.
[5]	XU Y F , ZHANG N , LI L ,et al. Joint learning of super-resolution and perceptual image enhancement for single image[J]. IEEE Access, 2021(9): 48446-48461.
[6]	邢亚芬, 殷海兵, 王鸿奎 ,等. 基于视频时域感知特性的恰可察觉失真模型[J]. 电信科学, 2022,38(2): 92-102.
	XING Y F , YIN H B , WANG H K ,et al. Video temporal perception characteristics based just noticeable difference model[J]. Telecommunications Science, 2022,38(2): 92-102.
[7]	BAE S H , KIM M . A novel generalized DCT-based JND profile based on an elaborate CM-JND model for variable block-sized transforms in monochrome images[J]. IEEE Transactions on Image Processing, 2014,23(8): 3227-3240.
[8]	FRISTON K J , DAUNIZEAU J , KIEBEL S J . Reinforcement learning or active inference?[J]. PloS one, 2009,4(7): e6421.
[9]	FRISTON K . The free-energy principle:a unified brain theory?[J]. Nature Reviews Neuroscience, 2010,11(2): 127-138.
[10]	WU J J , SHI G M , LIN W S ,et al. Just noticeable difference estimation for images with free-energy principle[J]. IEEE Transactions on Multimedia, 2013,15(7): 1705-1710.
[11]	WU J J , LI L D , DONG W S ,et al. Enhanced just noticeable difference model for images with pattern complexity[J]. IEEE Transactions on Image Processing, 2017,26(6): 2682-2693.
[12]	ZENG Z P , ZENG H Q , CHEN J ,et al. Visual attention guide dpixel-wise just noticeable difference model[J]. IEEE Access, 2019(7): 132111-132119.
[13]	LIU X Q , ZHAN X N , WANG M H . A novel edge-pattern-based just noticeable difference model for screen content images[C]// 2020 IEEE 5th International Conference on Signal and Image Processing (ICSIP). Piscataway:IEEE Press, 2020: 386-390.
[14]	WANG H K , YU L , LIANG J H ,et al. Hierarchical predictive coding-based JND estimation for image compression[J]. IEEE Transactions on Image Processing, 2020(30): 487-500.
[15]	WATSON A B , . DC Tune:A technique for visual optimization of DCT quantization matrices for individual images[C]// Sid International Symposium Digest of Technical Papers,Society for Information Display, 1993(24): 946-946.
[16]	WEI Z Y , NGAN K N . Spatio-temporal just noticeable distortion profile for grey scale image/video in DCT domain[J]. IEEE Transactions on Circuits and Systems for Video Technology, 2009,19(3): 337-346.
[17]	ECKERT M P , BRADLEY A P . Perceptual quality metrics applied to still image compression[J]. Signal processing, 1998,70(3): 177-200.
[18]	ZHANG X H , LIN W S , XUE P . Improved estimation for just-noticeable visual distortion[J]. Signal Processing, 2005,85(4): 795-808.
[19]	KNILL D C , POUGET A . The Bayesian brain:the role of uncertainty in neural coding and computation[J]. TRENDS in Neurosciences, 2004,27(12): 712-719.
[20]	EASTTOM C , . On the use of the SSIM algorithm for detecting intellectual property copying in web design[C]// 2021 IEEE 11th Annual Computing and Communication Workshop and Conference (CCWC).Piscataway:IEEE Press. 2021: 0860-0864.
[21]	LE C P , AUTRUSSEAU F . Subjective quality assessment IRCCyN/IVC database[EB]. 2005.
[22]	PONOMARENKO N , JIN L , IEREMEIEV O ,et al. Image database TID2013:peculiarities,results and perspectives[J]. Signal Processing:Image Communication, 2015,30: 57-77.
[23]	JUDD T , EHINGER K , DURAND F ,et al. Learning to predict where humans look[C]// 2009 IEEE 12th International Conference on Computer Vision. Piscataway:IEEE Press, 2009: 2106-2113.
[24]	WANG H K , YU L , YIN H B ,et al. An improved DCT-based JND estimation model considering multiple masking effects[J]. Journal of Visual Communication and Image Representation, 2020(71): 102850.
[25]	LI J L , YU L , WANG H K . Perceptual redundancy model for compression of screen content videos[J]. IET Image Processing, 2022,16(6): 1724-1741.
[26]	WANG J , LIU Y , WEI P ,et al. Fractal image coding using SSIM[C]// 2011 18th IEEE International Conference on Image Processing. Piscataway:IEEE Press, 2011: 241-244.

Metrics

Recommended 0

No Suggested Reading articles found!

Just noticeable distortion model based on entropy masking in DCT domain

RichHTML

PDF下载

Knowledge

Abstract

Cite this article

share this article

Figures/Tables 13

References 26

Related Articles 1

Metrics

Recommended 0