Generalized Fuzzy C-Means for Image Segmentation Based on Adaptive Weighted Image Patch

Zhanlong Zhu; Jianbin Dong; Mingliang Li; Yibo Zheng; Yuan Wang

doi:10.3788/LOP57.241006

Journals >Laser & Optoelectronics Progress >Volume 57 >Issue 24 >Page 241006 > Article

Laser & Optoelectronics Progress
Vol. 57, Issue 24, 241006 (2020)

Generalized Fuzzy C-Means for Image Segmentation Based on Adaptive Weighted Image Patch

Zhanlong Zhu^1、2、3, Jianbin Dong^1、2、3, Mingliang Li^1、2、3, Yibo Zheng^2、3、*, and Yuan Wang^2、3

Author Affiliations

¹School of Information Engineering, Hebei GEO University, Shijiazhuang, Hebei 050031, China

²Hebei Key Laboratory of Optoelectronic Information and Geo-Detection Technology, Hebei GEO University, Shijiazhuang, Hebei 050031, China

³Intelligent Sensor Network Engineering Research Center of Hebei Province, Hebei GEO University, Shijiazhuang, Hebei 050031, China

show less

DOI: 10.3788/LOP57.241006 Cite this Article Set citation alerts

Zhanlong Zhu, Jianbin Dong, Mingliang Li, Yibo Zheng, Yuan Wang. Generalized Fuzzy C-Means for Image Segmentation Based on Adaptive Weighted Image Patch[J]. Laser & Optoelectronics Progress, 2020, 57(24): 241006 Copy Citation Text

show less

Examples of weight allocation of image patch. (a) Three selected image blocks; (b) gray value of image block A; (c) gray value of image block B; (d) gray value of image block C; (e) wjr of image block A; (f) wjr of image block B; (g) wjr of image block C

Fig. 1. Examples of weight allocation of image patch. (a) Three selected image blocks; (b) gray value of image block A; (c) gray value of image block B; (d) gray value of image block C; (e) w_jr of image block A; (f) w_jr of image block B; (g) w_jr of image block C

Download full size

Segmentation results of different algorithms on the synthetic image. (a) Original synthetic image; (b) image with SPN(0.1); (c) GFCM algorithm; (d) KGFCM_S1 algorithm; (e) KGFCM_S2 algorithm; (f) EnFCM algorithm; (g) FGFCM algorithm; (h) NDFCM algorithm; (i) WIPFCM algorithm; (j) GFCM_WP algorithm

Fig. 2. Segmentation results of different algorithms on the synthetic image. (a) Original synthetic image; (b) image with SPN(0.1); (c) GFCM algorithm; (d) KGFCM_S1 algorithm; (e) KGFCM_S2 algorithm; (f) EnFCM algorithm; (g) FGFCM algorithm; (h) NDFCM algorithm; (i) WIPFCM algorithm; (j) GFCM_WP algorithm

Download full size

Fig. 3. Segmentation results of different algorithms on the synthetic image. (a) Synthetic image with WGN (0,0.01) & SPN (0.1); (b) GFCM algorithm; (c) KGFCM_S1 algorithm; (d) KGFCM_S2 algorithm; (e) EnFCM algorithm; (f) FGFCM algorithm; (g) NDFCM algorithm; (h) WIPFCM algorithm; (i) GFCM_WP algorithm

Download full size

Fig. 4. Actual images to be segmented. (a) Bird image; (b) standard segmentation image of Bird; (c) House image; (d) Coins image; (e) Rocks image

Download full size

Fig. 5. Segmentation results of different algorithms on the Bird image. (a) Image with SPN(0.2); (b) GFCM algorithm; (c) KGFCM_S1 algorithm; (d) KGFCM_S2 algorithm; (e) EnFCM algorithm; (f) FGFCM algorithm; (g) NDFCM algorithm; (h) WIPFCM algorithm; (i) GFCM_WP algorithm

Download full size

Fig. 6. Segmentation results of different algorithms on the House image. (a)Image with WGN(0,0.002) & SPN(0.05); (b) GFCM algorithm; (c) KGFCM_S1 algorithm; (d) KGFCM_S2 algorithm; (e) EnFCM algorithm; (f) FGFCM algorithm; (g) NDFCM algorithm; (h) WIPFCM algorithm; (i) GFCM_WP algorithm

Download full size

Fig. 7. Segmentation results of different algorithms on the Coins image. (a)Image with SPN(0.1); (b) GFCM algorithm; (c) KGFCM_S1 algorithm; (d) KGFCM_S2 algorithm; (e) EnFCM algorithm; (f) FGFCM algorithm; (g) NDFCM algorithm; (h) WIPFCM algorithm; (i) GFCM_WP algorithm

Download full size

Fig. 8. Segmentation results of different algorithms on the Rocks image. (a)Image with SPN(0.1); (b) GFCM algorithm; (c) KGFCM_S1 algorithm; (d) KGFCM_S2 algorithm; (e) EnFCM algorithm; (f) FGFCM algorithm; (g) NDFCM algorithm; (h) WIPFCM algorithm; (i) GFCM_WP algorithm

Download full size

Algorithm	Parameter setting
Algorithm	m	α	β	λ_α	λ_s	λ_g	T	ε	q
GFCM	2	0.9					300	10^-4
KGFCM_S1	2	0.9	4				300	10^-4	3
KGFCM_S2	2	0.9	4				300	10^-4	3
EnFCM	2		4				300	10^-4	3
FGFCM	2				3	3	300	10^-4	3
NDFCM	2			1	3	3	300	10^-4	3
WIPFCM	2						300	10^-4	3
GFCM_WP	2	0.9					300	10^-4	3

Table 1. Parameters of algorithms

Noise level	Index	GFCM	KGFCM_S1	KGFCM_S2	EnFCM	FGFCM	NDFCM	WIPFCM	GFCM_WP
WGN(0,0.01)	SA	0.8777	0.9561	0.9812	0.9558	0.9790	0.9703	0.9781	0.9903
	ARI	0.8369	0.9415	0.9735	0.9411	0.9720	0.9604	0.9709	0.9871
WGN(0,0.015)	SA	0.8044	0.9517	0.9734	0.9518	0.9731	0.9649	0.9705	0.9810
	ARI	0.7392	0.9355	0.9643	0.9358	0.9642	0.9532	0.9607	0.9747
WGN(0,0.02)	SA	0.7528	0.9421	0.9633	0.9421	0.9647	0.9603	0.9570	0.9659
	ARI	0.6704	0.9228	0.9511	0.9229	0.9530	0.9471	0.9426	0.9546
SPN (0.1)	SA	0.9231	0.8842	0.9701	0.8803	0.9649	0.9624	0.9841	0.9961
	ARI	0.8979	0.8456	0.9602	0.8404	0.9532	0.9499	0.9788	0.9948
SPN (0.2)	SA	0.8507	0.8137	0.9387	0.8211	0.9196	0.9272	0.9745	0.9890
	ARI	0.8009	0.7515	0.9183	0.7615	0.8928	0.9029	0.9660	0.9854
WGN(0,0.01) &SPN(0.1)	SA	0.8156	0.8701	0.9461	0.8696	0.9537	0.9560	0.9424	0.9838
	ARI	0.7542	0.8267	0.9281	0.8262	0.9382	0.9413	0.9232	0.9784

Table 2. Segmentation results of different algorithms on the synthetic image

Image	Noise level	Index	GFCM	KGFCM_S1	KGFCM_S2	EnFCM	FGFCM	NDFCM	WIPFCM	GFCM_WP
	WGN(0,0.02)	SA	0.9007	0.9515	0.9535	0.9616	0.9560	0.9557	0.9556	0.9558
		ARI	0.8013	0.9030	0.9071	0.9031	0.9121	0.9113	0.9112	0.9114
	WGN(0,0.03)	SA	0.8521	0.9496	0.9500	0.9497	0.9546	0.9550	0.9539	0.9550
		ARI	0.7042	0.8992	0.9000	0.8993	0.9085	0.9099	0.9078	0.9099
Bird	SPN (0.1)	SA	0.9146	0.9337	0.9592	0.9343	0.9599	0.9593	0.9597	0.9626
		ARI	0.8292	0.8373	0.9185	0.8686	0.9198	0.9186	0.9194	0.9251
	SPN (0.2)	SA	0.8672	0.8932	0.9548	0.8932	0.9483	0.9517	0.9576	0.9612
		ARI	0.7345	0.7863	0.9096	0.7863	0.8965	0.9035	0.9145	0.9225
	WGN(0,0.02) &SPN(0.1)	SA	0.8768	0.9243	0.9461	0.9244	0.9507	0.9524	0.9460	0.9526
		ARI	0.7536	0.8486	0.8921	0.8488	0.9015	0.9047	0.8920	0.9053
		H_r(I)	5.0977	5.0877	5.0874	5.0898	4.2392	4.2651	4.2328	4.2191
	WGN(0,0.002)	H_l(I)	1.0495	1.0359	1.0419	1.0395	1.0375	1.0399	1.0396	1.0388
		E	6.1473	6.1236	6.1293	6.1293	5.2767	5.3050	5.2723	5.2579
		H_r(I)	5.2411	5.2292	5.2289	5.2305	4.5052	4.4959	4.5028	4.4828
	WGN(0,0.005)	H_l(I)	1.0426	1.0399	1.0442	1.0430	1.0425	1.0412	1.0420	1.0416
		E	6.2837	6.2691	6.2731	6.2735	5.5477	5.5371	5.5448	5.5244
		H_r(I)	4.4690	4.4619	4.4639	4.4688	3.6127	3.6567	3.5908	3.5832
House	SPN (0.05)	H_l(I)	1.0451	1.0350	1.0362	1.0433	1.0372	1.0473	1.0364	1.0351
		E	5.5141	5.4969	5.5001	5.5121	4.6499	4.7040	4.6271	4.6183
		H_r(I)	4.4119	4.4004	4.4034	4.4102	3.6161	3.6738	3.6039	3.5713
	SPN (0.1)	H_l(I)	1.0544	1.0391	1.0424	1.0545	1.0395	1.0552	1.0429	1.0345
		E	5.4663	5.4395	5.4458	5.4647	4.6555	4.7290	4.6468	4.6059
		H_r(I)	5.0679	5.0637	5.0649	5.0676	4.2894	4.2995	4.2564	4.2557
	WGN(0,0.002) &SPN(0.05)	H_l(I)	1.0495	1.0364	1.0445	1.0493	1.0429	1.0486	1.0380	1.0383
		E	6.1174	6.1001	6.1095	6.1169	5.3323	5.3481	5.2945	5.2940
		H_r(I)	4.9222	4.8211	4.8210	4.8218	4.2669	4.2642	4.2665	4.2529
	WGN(0,0.005)	H_l(I)	0.7456	0.7164	0.7068	0.7396	0.7167	0.7366	0.7093	0.7157
		E	5.6678	5.5374	5.5278	5.5613	4.9836	5.0008	4.9758	4.9686
		H_r(I)	4.8321	4.8296	4.8295	4.8298	4.3213	4.3122	4.3229	4.3089
	WGN(0,0.01)	H_l(I)	0.7517	0.7211	0.7151	0.7428	0.7204	0.7413	0.7168	0.7203
		E	5.5838	5.5507	5.5446	5.5726	5.0417	5.0545	5.0396	5.0292
		H_r(I)	3.4357	3.4344	3.4336	3.4353	2.8642	2.8420	2.8516	2.8602
Image	Noise level	Index	GFCM	KGFCM_S1	KGFCM_S2	EnFCM	FGFCM	NDFCM	WIPFCM	GFCM_WP
Coins	SPN (0.05)	H_l(I)	0.8071	0.7190	0.7320	0.8060	0.7181	0.8045	0.7125	0.7015
		E	4.2428	4.1534	4.1655	4.2413	3.5824	3.6465	3.5641	3.5617
		H_r(I)	3.4415	3.4307	3.4307	3.4321	2.8964	2.8665	2.8933	2.8884
	SPN (0.1)	H_l(I)	0.8901	0.7273	0.7754	0.8825	0.7216	0.8837	0.7191	0.7033
		E	4.3316	4.1580	4.2061	4.3146	3.6181	3.7502	3.6130	3.5916
		H_r(I)	4.8011	4.7368	4.7640	4.7644	4.2528	4.2087	4.2445	4.2357
	WGN(0,0.005) & SPN(0.05)	H_l(I)	0.8182	0.7180	0.7409	0.8074	0.7196	0.7982	0.7087	0.7161
		E	5.6193	5.4818	5.5049	5.5718	4.9724	5.0070	4.9532	4.9519
		H_r(I)	5.2001	5.1971	5.1922	5.1985	4.2027	4.2022	4.1805	4.1879
	WGN(0,0.005)	H_l(I)	1.3157	1.3132	1.3135	1.3156	1.3144	1.3136	1.3134	1.3139
		E	6.5158	6.5103	6.5107	6.5141	5.5171	5.5158	5.4939	5.5018
		H_r(I)	5.2566	5.2437	5.2437	5.2452	4.3874	4.3661	4.3839	4.3400
	WGN(0,0.01)	H_l(I)	1.3193	1.3151	1.3181	1.3194	1.3172	1.3169	1.3178	1.3166
		E	6.5759	6.5588	6.5618	6.5646	5.7046	5.6830	5.7017	5.6566
		H_r(I)	4.6550	4.6479	4.6494	4.6545	3.6113	3.6514	3.6288	3.6086
Rocks	SPN (0.05)	H_l(I)	1.3193	1.3122	1.3123	1.3191	1.3120	1.3177	1.3124	1.3104
		E	5.9744	5.9601	5.9617	5.9736	4.9233	4.9691	4.9412	4.9190
		H_r(I)	4.5552	4.5427	4.5452	4.5544	3.6593	3.5989	3.6082	3.5792
	SPN (0.1)	H_l(I)	1.3272	1.3135	1.3144	1.3271	1.3234	1.3151	1.3127	1.3103
		E	5.8824	5.8562	5.8597	5.8814	4.9828	4.9140	4.9209	4.8895
		H_r(I)	5.1239	5.1069	5.1080	5.1111	4.2034	4.2341	4.1884	4.1966
	WGN(0,0.005) & SPN(0.05)	H_l(I)	1.3261	1.3147	1.3176	1.3259	1.3138	1.3220	1.3234	1.3149
		E	6.4500	6.4215	6.4257	6.4370	5.5172	5.5561	5.5118	5.5115

Table 3. Segmentation results of different algorithms on the real images

Download Citation

Set citation alerts for the article

Tools

Set citation alerts for the article

Save the article for my favorites

Paper Information