Obtenção de tabela de quantização para compressão de imagens utilizando algoritmos genéticos

• Ano de defesa: 2005
• Autor(a) principal: Costa, Leonardo Faria
• Tipo de documento: Dissertação
• Tipo de acesso: Acesso aberto
• Idioma: por
• Instituição de defesa: Universidade Federal de Uberlândia; BR; Programa de Pós-graduação em Engenharia Elétrica; Engenharias; UFU
• Palavras-chave em Português: Algoritmo genético; Relação sinal/ruído; Seleção natural; Função de aptidão; Algoritmos genéticos; Genetic algorithm; Signal noise relation; Natural selection; Fitness function; CNPQ::ENGENHARIAS::ENGENHARIA ELETRICA
• Link de acesso: https://repositorio.ufu.br/handle/123456789/14642

The purpose of this work is to generate a better Quantization Table for a group of natural images and medical images using the Genetic Algorithm method in the process of compression JPEG. The fitness function of the population will be analyzed and the probability of selection the better pairs of chromosomes or matrix, and we will also breach means to improve the results through modifications of the input parameters. In the previous processing of the coded image it is necessary to know very well the Quantization Table to achieve an image with less loss and therefore, better final quality. The method of Genetic Algorithm applied to this program is based on mechanisms of natural selection and reproduction for a set of natural and medical images and furthermore it allowed for the creation of not one but various Tables with SNR higher than those produced by the JPEG Table since it stimulated the principal characteristics of this method in programming such as: codification of defined parameters; the process of search for better matrix starting from a group of matrix and not a simple matrix; information obtained of an objective determined function (fitness function) and the use of the probabilistic rules of transition. Based on the results obtained in this work, we can recommend the use of Genetic Algorithm for natural images and human eye images with a compression rate up to 30:1 and a great reconstructed image quality.