GALVÃO, J.; LIMA, E.A.; MADEIRO, F.

doi:10.5540/tema.2017.018.02.0197

Acessibilidade / Reportar erro

Brasil

Español English

sumário « anterior atual seguinte »

Sumário

Articles • TEMA (São Carlos) 18 (2) • May-Aug 2017 • https://doi.org/10.5540/tema.2017.018.02.0197 copy

Atribuição de Índices para QV Robusta usando o Algoritmo Variable Neighborhood Search

Authorship SCIMAGO INSTITUTIONS RANKINGS

ABSTRACT

This work addresses the problem of robust vector quantization in the scenario of image transmission over a binary symmetric channel. The application of the algorithm Variable Neighborhood Search (VNS) is introduced for the purpose of index assignment of codevectors. Simulation results reveal that the algorithm VNS, when compared to Simulated Annealing and Genetic algorithm, produces codebooks with the same robustness to channel erros with the spent of minor time.

Keywords:
Robust Vector Quantization; optimization algorithms; digital image processing

N	t ₀	_^tf	α	Avaliações	Δ_^Id
32	10.000	1	0,95	180.000	53,57%
64	10.000	1	0,95	180.000	56,46%
128	10.000	1	0,95	177.653	61,63%
256	50.000	1	0,97	129.484	69,50%
512	50.000	10	0,97	146.174	71,42%

N	P	Tc	Tm	Ps	Avaliações	Δ_^Id
32	10	90%	5%	4	70.000	58,71%
64	8	80%	0,5%	4	70.000	58,98%
128	8	80%	0,4%	4	70.000	62,60%
256	8	90%	0,5%	6	120.000	69,24%
512	8	90%	0,2%	4	150.000	69,54%

N	Vmax	Vpts	Pl	Avaliações	Δ_^Id
32	30	50	18	22.301	58,84%
64	10	80	20	26.191	59,14%
128	15	80	10	48.801	62,74%
256	10	80	20	196.878	69,79%
512	10	70	18	329.883	71,59%

Método	∈ = 0,0	∈ = 10^-4	∈ = 5 × 10^-4	∈ = 10^-3	∈ = 5 × 10^-3	∈ = 10^-2	∈ = 5 × 10^-2	∈ = 10^-1
Original	27,59	27,46	27,00	26,56	23,87	21,88	16,32	13,89
SA	27,59	27,54	27,34	27,12	25,59	24,21	19,29	16,66
AG	27,59	27,54	27,34	27,11	25,58	24,19	19,28	16,65
VNS	27,59	27,54	27,34	27,12	25,63	24,27	19,33	16,70

Método	∈ = 0,0	∈ = 10^-4	∈ = 5 × 10^-4	∈ = 10^-3	∈ = 5 × 10^-3	∈ = 10^-2	∈ = 5 × 10^-2	∈ = 10^-1
Original	28,76	28,60	28,00	27,30	24,04	21,88	16,04	13,64
SA	28,76	28,70	28,45	28,16	26,35	24,78	19,53	16,82
AG	28,76	28,71	28,46	28,17	26,31	24,79	19,50	16,80
VNS	28,76	28,70	28,45	28,17	26,37	24,83	19,57	16,86

Método	∈ = 0,0	∈ = 10^-4	∈ = 5 × 10^-4	∈ = 10^-3	∈ = 5 × 10^-3	∈ = 10^-2	∈ = 5 × 10^-2	∈ = 10^-1
Original	29,92	29,69	28,67	27,88	24,30	21,68	15,63	13,43
SA	29,92	29,84	29,59	29,30	27,23	25,44	19,89	17,27
AG	29,92	28,85	29,45	29,22	27,19	25,31	19,75	17,07
VNS	29,92	28,85	29,49	29,24	27,28	25,34	19,85	17,14

Método	∈ = 0,0	∈ = 10^-4	∈ = 5 × 10^-4	∈ = 10^-3	∈ = 5 × 10^-3	∈ = 10^-2	∈ = 5 × 10^-2	∈ = 10^-1
Original	30,82	30,44	29,22	28,05	23,52	21,01	14,95	12,68
SA	30,82	30,72	30,32	29,91	27,37	25,45	19,63	16,83
AG	30,82	30,71	30,28	29,80	27,15	25,20	19,37	16,62
VNS	30,82	30,72	30,31	29,86	27,38	25,46	19,63	16,83

Método	∈ = 0,0	∈ = 10^-4	∈ = 5 × 10^-4	∈ = 10^-3	∈ = 5 × 10^-3	∈ = 10^-2	∈ = 5 × 10^-2	∈ = 10^-1
Original	31,55	31,09	29,69	28,31	23,53	20,92	14,91	12,76
SA	31,55	31,44	31,00	30,54	27,90	25,94	20,10	17,31
AG	31,55	31,45	30,98	30,51	27,76	25,82	19,93	17,17
VNS	31,55	31,44	31,00	30,53	27,84	25,93	20,05	17,28

Sociedade Brasileira de Matemática Aplicada e Computacional Rua Maestro João Seppe, nº. 900, 16º. andar - Sala 163 , 13561-120 São Carlos - SP, Tel. / Fax: (55 16) 3412-9752 - São Carlos - SP - Brazil
E-mail: sbmac@sbmac.org.br

Acompanhe os números deste periódico no seu leitor de RSS

[1] *Autor correspondente: Jair Galvão - E-mail: jair.galvao@upe.br