Diffusion-Based Image Inpainting Forensics Via Gradient Domain Guided Filtering Enhancement

Tingting Liu; Yujin Zhang; Fei Wu; Shiting Xiong

doi:10.3788/LOP57.081003

Laser & Optoelectronics Progress, Volume. 57, Issue 8, 081003(2020)

Diffusion-Based Image Inpainting Forensics Via Gradient Domain Guided Filtering Enhancement

Tingting Liu¹, Yujin Zhang^1,2、*, Fei Wu¹, and Shiting Xiong¹

Author Affiliations

¹School of Electrical and Electronic Engineering, Shanghai University of Engineering Science, Shanghai 201620, China

²Shanghai Key Laboratory of Integrated Administration Technologies for Information Security, Shanghai Jiao Tong University, Shanghai 200240, China

show less

Abstract Get PDF(in Chinese)

Figures & Tables(9)

Fig. 1. Training and detection flow chart

Download full size

Fig. 2. Artifacts left after the image has been inpainted. (a) Original image; (b) ground truth; (c) δ_Δ_I map

Download full size

Fig. 3. Comparison of repair image positioning results. (a) Original image; (b) ground truth image; (c) location result obtained by algorithm used in reference [10]; (d) location result obtained by our algorithm

Download full size

Table 1. Comparison of F1 scores based on Isotropic repair algorithm

View table

Table 1. Comparison of F1 scores based on Isotropic repair algorithm

Pixel size	Algorithm
	Square		Circular		Irregular
	Reference[10]	Proposed	Reference[10]	Proposed	Reference[10]	Proposed
64×64	0.8896	0.9088	0.8863	0.9056	0.8757	0.8947
32×32	0.7908	0.8192	0.7893	0.8176	0.7645	0.7913
16×16	0.6585	0.6880	0.6664	0.6960	0.6215	0.6489
8×8	0.4063	0.4381	0.4858	0.5174	0.1990	0.2137

Table 2. Comparison of F1 score based on Edge-oriented repair algorithm

View table

Table 2. Comparison of F1 score based on Edge-oriented repair algorithm

Pixel size	Algorithm
	Square		Circular		Irregular
	Reference[10]	Proposed	Reference[10]	Proposed	Reference[10]	Proposed
64×64	0.8821	0.9011	0.8790	0.8977	0.8661	0.8846
32×32	0.7808	0.8082	0.7757	0.8031	0.7480	0.7735
16×16	0.6465	0.6779	0.6442	0.6729	0.5940	0.6208
8×8	0.3171	0.3388	0.4065	0.4315	0.1726	0.1833

Table 3. Comparison of F1 scores based on Delaunay-oriented repair algorithm

View table

Table 3. Comparison of F1 scores based on Delaunay-oriented repair algorithm

Pixel size	Algorithm
	Square		Circular		Irregular
	Reference[10]	Proposed	Reference[10]	Proposed	Reference[10]	Proposed
64×64	0.8147	0.8318	0.8610	0.8788	0.8297	0.8474
32×32	0.6189	0.6388	0.6953	0.7164	0.6768	0.6977
16×16	0.3138	0.3352	0.3184	0.3390	0.3833	0.4063
8×8	0.1007	0.1074	0.1234	0.1305	0.0784	0.0784

Table 4. Comparison of F1 scores based on Isotropic repair algorithm (1.1×)

View table

Table 4. Comparison of F1 scores based on Isotropic repair algorithm (1.1×)

Pixel size	Algorithm
	Square		Circular		Irregular
	Reference[10]	Proposed	Reference[10]	Proposed	Reference[10]	Proposed
64×64	0.8106	0.8336	0.8097	0.8335	0.7974	0.8194
32×32	0.6514	0.6828	0.6526	0.6852	0.6289	0.6602
16×16	0.4582	0.4915	0.4661	0.5007	0.4301	0.4626
8×8	0.2051	0.2217	0.2742	0.2982	0.1246	0.1373

Table 5. Comparison of F1 score based on Edge-oriented repair algorithm (1.1×)

View table

Table 5. Comparison of F1 score based on Edge-oriented repair algorithm (1.1×)

Pixel size	Algorithm
	Square		Circular		Irregular
	Reference[10]	Proposed	Reference[10]	Proposed	Reference[10]	Proposed
64×64	0.8001	0.8237	0.7981	0.8203	0.7850	0.8059
32×32	0.6351	0.6659	0.6342	0.6663	0.6078	0.6366
16×16	0.4447	0.4777	0.4427	0.4769	0.4079	0.4365
8×8	0.1582	0.1729	0.2222	0.2437	0.0972	0.1085

Table 6. Comparison of F1 scores based on Delaunay-oriented repair algorithm (1.1×)

View table

Table 6. Comparison of F1 scores based on Delaunay-oriented repair algorithm (1.1×)

Pixel size	Algorithm
	Square		Circular		Irregular
	Reference[10]	Proposed	Reference[10]	Proposed	Reference[10]	Proposed
64×64	0.7374	0.7574	0.7687	0.7879	0.7466	0.7689
32×32	0.5033	0.5256	0.5425	0.5714	0.5398	0.5653
16×16	0.2311	0.2488	0.2280	0.2476	0.2588	0.2819
8×8	0.0450	0.0517	0.0603	0.0653	0.0436	0.0479

Tools

Get Citation

Copy Citation Text

Tingting Liu, Yujin Zhang, Fei Wu, Shiting Xiong. Diffusion-Based Image Inpainting Forensics Via Gradient Domain Guided Filtering Enhancement[J]. Laser & Optoelectronics Progress, 2020, 57(8): 081003

Download Citation

EndNote(RIS)BibTex Plain Text

Set citation alerts for article

Save article for my favorites

Paper Information

Category: Image Processing

Received: Aug. 13, 2019

Accepted: Sep. 6, 2019

Published Online: Apr. 3, 2020

The Author Email: Yujin Zhang (yjzhang@sues.edu.cn)

DOI:10.3788/LOP57.081003

Topics

laser devices and laser physics

Lasers and Laser Optics

Laser physics

laser manufacturing

Instrumentation, Measurement and Metrology