Highlighted area extraction by using a basis function network

This paper describes an effective image extraction method by using a basis function network. The basis function network is represented as a linear combination of weighted basis functions with different support size. The method which employs a compactly supported function as a basis is superior lo an image feature approximation in a local space. The weights are set to optimize the evaluation function. In this paper, the headline area extraction can be realized by specializing the evaluation function to the feature extraction. We try to extract headline areas from newspaper by using the proposed method and discuss its validity.