Studies on Genetic Algorithms for Image Processing

• Main Authors: Chih-Ching Lai, 賴智錦
• Other Authors: Din-Chang Tseng
• Format: Others
• Language: en_US
• Published: 1999
• Online Access: http://ndltd.ncl.edu.tw/handle/43176158577665048452

博士 === 國立中央大學 === 資訊工程研究所 === 87 === Image processing is a rapidly evolving research field with growing applications in science and engineering. In this study, we solve four image-processing problems from the viewpoint of optimization. Genetic algorithms (GAs) are inspired by biological evolution and they are widely believed to be effective global optimization algorithms. GAs provide versatile problem-solving mechanism for search, adaptation, and learning in a variety of application domains, especially for those problems in which heuristic methods lead to unsatisfactory results. Hence, the motivation of this dissertation study is to apply the emerging optimization techniques for solving image-processing problems. In the first topic, a least-square model-based halftoning technique using a GA is proposed to produce halftone images with less gray-level distortion. The GA is used for searching the visually pleasing placement of black pixels in halftone images. In the second topic, an L-filter combining a GA for reducing the blocking artifacts in the compressed images is proposed. The GA is used to search the near-optimal parameters for the L-filter. The proposed approach has the advantage of combining the powerful enhancement of the L-filter and the global solution exploration of the GA. In the third topic, an unsupervised segmentation approach based on Markov random field (MRF) model for multispectral textured images is proposed. We use a GA to provide a better initialization for the iterated conditional modes (ICM) algorithm in the pixel-labeling process. The proposed hybrid approach has the advantage that combines the fast convergence of the MRF-based iterative algorithm and the powerful global exploration of the GA. In the last topic, a multilevel optimal thresholding based on a GA is proposed. The GA is used to find the parameters in a joint probability density function that best fit the given histogram. The experimental results reveal that the proposed approaches based on the GA actually provide appropriate solutions for the considered image-processing problems.