Otimização do algoritmo Non Local Means mediante agrupamento por similaridade no domínio da frequência

• Ano de defesa: 2017
• Autor(a) principal: Meneses, Douglas Andrade de
• Tipo de documento: Dissertação
• Tipo de acesso: Acesso aberto
• Idioma: por
• Instituição de defesa: Universidade Federal da Paraíba; Brasil; Informática; Programa de Pós-Graduação em Informática; UFPB
• Palavras-chave em Português: Remoção de ruído; Non Local Means; Transformada discreta do cosseno; Denoising; Discrete cosine transform; CNPQ::CIENCIAS EXATAS E DA TERRA::CIENCIA DA COMPUTACAO
• Link de acesso: https://repositorio.ufpb.br/jspui/handle/123456789/12910

Among the image denoising methods in state-of-the-art the Non Local Means (NLM) is highlighted due to its efficiency. However, its computational complexity becomes an obstacle to most applications. In this paper, we propose a new approach which aims to reduce processing time, without resorting to search windows, thus preserving the non-local calculations that characterize NLM. The proposed method uses K-means clustering to group pixels with similar neighborhood in the frequency domain by Discrete Cosine Transform. In order to avoid harsh transitions at cluster borders, one pixel can be assigned to different clusters. After this initial step, the NLM algorithm performs a cluster-based search. Experimental results testify a reduction in the computational time around 19 times and, in some cases, improvements in the Mean Squared Error values, when compared to the original algorithm.