Solving an image separation problem using space-scale analysis by wavelets

点击翻译按钮获取中文摘要

When we observe mixtures of original images, we want to separate original images. This inverse problem is called an image separation problem. This problem appears, analyzing satellite images. In this research, we studied two-dimensional wavelet analysis for an image separation problem.We proposed a continuous multiwavelet transformation with annular sector multiwavelets, which is the most suitable transformation for solving an image separation problem. When mixtures are superpositions of original images without translations, we can separate ten original images from eleven mixtures, using the proposed transformation and Gaussian elimination. When mixtures are superpositions of original images with translations, we can separate four original images from four mixtures, using complex-valued correlation functions of continuous multiwavelet transforms.