複素数ウェーブレット変換を基にした新たな方向性ウェーブレット変換の提案と医用画像認識への応用

Abstract

In this paper, we propose the novel directional Wavelet Transform based on the Complex Discrete Wavelet Transform(CDWT). The CDWT is one of the Wavelet Transform and widely used as space-frequency analysis method for the signal and the image. The CDWT decomposes the image into directional components from the frequency analysis. And then, we can get the border, discontinuities and edges in the image. This property is called Directional Selection and this property is expected to give us shape information and geometric features. But, previous Directional Selection only offers the six directional components and features. Thus we firstly design the directional filter that can extract many directional feature rather than previous one. Secondly we propose the novel directional Wavelet Transform combining with previous CDWT and designed directional filters. Finally, we apply the proposed method into medical image processing.