An Improved Criminisi Image Inpainting Method Based on Information Entropy and Gradient Factor

Fengsui Wang; Zhengnan Liu; Linjun Fu

doi:10.3788/LOP57.221006

Journals >Laser & Optoelectronics Progress >Volume 57 >Issue 22 >Page 221006 > Article

Laser & Optoelectronics Progress
Vol. 57, Issue 22, 221006 (2020)

An Improved Criminisi Image Inpainting Method Based on Information Entropy and Gradient Factor

Fengsui Wang^{1、2、3、*}, Zhengnan Liu^1、2、3, and Linjun Fu^1、2、3

Author Affiliations

¹Key Laboratory of Advanced Perception and Intelligent Control of High-end Equipment, Ministry of Education, Wuhu, Anhui 241000, China

²Anhui Key Laboratory of Electric Drive and Control, Wuhu, Anhui 241000, China

³School of Electrical Engineering, Anhui Polytechnic University, Wuhu, Anhui 241000, China

show less

DOI: 10.3788/LOP57.221006 Cite this Article Set citation alerts

Fengsui Wang, Zhengnan Liu, Linjun Fu. An Improved Criminisi Image Inpainting Method Based on Information Entropy and Gradient Factor[J]. Laser & Optoelectronics Progress, 2020, 57(22): 221006 Copy Citation Text

show less

Fig. 1. Inpainting effect of global search and dynamic range search. (a) Original images; (b) global search; (c) dynamic range search

Download full size

Fig. 2. Inpainting time for global search and dynamic range search

Download full size

Fig. 3. Inpainting effects of different sizes. (a) Original images; (b) size 5×5; (c) size 7×7; (d) size 9×9; (e) size 11×11

Download full size

Fig. 4. Inpainting effects comparison between six different algorithms and the proposed algorithm. (a) Original images; (b) inpainting results by MD algorithm; (c) inpainting results by PDE algorithm; (d) inpainting results by NN algorithm; (e) inpainting results by Ref.[20] method; (f) inpainting results by Criminisi algorithm; (g) inpainting results by Ref.[3] method; (h) inpainting results by proposed algorithm

Download full size

Image	Index	MDmethod	NNmethod	Method inRef.[20]	Criminisimethod	Method inRef.[3]	Proposedmethod
	PSNR /dB	29.5975	25.8096	24.3662	27.3331	27.6764	27.3669
Air	SSIM	0.9628	0.8668	0.7940	0.8684	0.8681	0.8675
	r	0.9708	0.9299	0.9018	0.9504	0.9543	0.9507
	PSNR /dB	15.4539	16.3219	14.4868	16.4791	16.9938	20.7359
Car	SSIM	0.3874	0.7054	0.3692	0.7041	0.7161	0.6427
	r	0.7858	0.8417	0.7587	0.8473	0.8565	0.9291
	PSNR /dB	18.9067	18.1455	17.8397	18.1682	18.5533	18.3531
Lenna	SSIM	0.8987	0.8912	0.8508	0.8919	0.8917	0.8953
	r	0.8131	0.7760	0.7543	0.7779	0.7974	0.7870
	PSNR /dB	23.6609	22.3550	21.8943	22.4453	23.3841	23.5420
Lincoln	SSIM	0.9806	0.9604	0.9554	0.9587	0.9800	0.9804
	r	0.9415	0.9207	0.9107	0.9227	0.9377	0.9399
	PSNR /dB	35.5345	45.4129	24.9755	49.2218	44.7734	44.6925
Pimple	SSIM	0.9979	0.9995	0.9102	0.9997	0.9982	0.9982
	r	0.9964	0.9996	0.9591	0.9998	0.9996	0.9996
	PSNR /dB	24.6307	25.6090	20.7125	26.7295	26.2762	26.9381
Average	SSIM	0.8455	0.8847	0.7759	0.8847	0.8908	0.8768
	r	0.9015	0.8936	0.8569	0.8997	0.9091	0.9213

Table 1. Inpainting effects comparison of evaluation index for different algorithms

Image	Method in Ref.[20]	Criminisi method	Method in Ref.[3]	Proposed method
Air	41.61	523.47	465.41	376.29
Car	124.50	2341.95	1739.68	1520.38
Lena	126.83	44182	34005	32789
Lincoln	48.52	333.37	269.79	201.26
Pimple	10.068	17.90	17.52	17.48

Table 2. Comparison of running time of different algorithms unit: s

Fengsui Wang, Zhengnan Liu, Linjun Fu. An Improved Criminisi Image Inpainting Method Based on Information Entropy and Gradient Factor[J]. Laser & Optoelectronics Progress, 2020, 57(22): 221006

Download Citation

Set citation alerts for the article

Tools

Set citation alerts for the article

Save the article for my favorites

Paper Information