遺伝的アルゴリズムによる画像セグメンテーションの最適化

概要

論文の詳細を見る
A method is proposed for automatic optimization of image segmentation procedures, which is based on the genetic algorithm and an original technique called "Male-female intersection" An image segmentation module corresponds to an individual in GA and it consists of connected filter primitives that realize fundamental image processing functions with specific parameters. Common boundary condition on each module is easily given by the pair of an input and goal image that enables users to clarify their requirement to the optimizer. A population includes modules with different combination of parameters. Effective evolution with reasonable computational cost can be performed by MF intersection over the populations. Furthermore, an image processing hardware is utilized in a prototype system in order to speed up image processing computation. From many experiments with images of the real world, effectiveness and practical availability of the proposed method is shown.
公益社団法人精密工学会の論文
2000-06-05