웨이브렛 프레임과 공간 정보를 이용한 질감 영상 분할

Abstract

This paper introduces a texture segmentation method that relies on envelope values of discrete wavelet frame coefficients and spatial information. A texture is characterized by a set of channel variances estimated at the output of the corresponding filter bank. Wavelet transform that is combination of filter banks is usefully tool for feature extraction of texture. We use envelope of discrete wavelet frame for feature extraction of texture using wavelet transform. But this method is sensitive to filter selection and have problem that calculation is complicated.

The proposed method removes these problems. We use wavelet frame instead of wavelet packet frame to reduce expenditure and add Coordinate Values to feature for efficient segmentation. In experiment, this method is proved that is efficient than the previous method using wavelet packet frame.