This paper discusses methods of segmenting complex texture images. A method of multiscale wavelet segmentation of color textures has been developed, based on their characteristic attributes and chromaticity characteristics. Multiscale image analysis is used to form the texture attributes. The method makes it possible to combine estimates of the texture characteristics on the basis of the discrete wavelet transformation and the color characteristics of the texture. Estimates of the segmentation efficiency are obtained from the number of iterations of the wavelet transformation, the type of wavelet basis, the form of the color-coordinate space, and the size of the filtering window for estimating the segmentation attributes.
展开▼
Branch FGUP Kometa Central Scientific Research Institute, Scientific Production Center of Optoelectronic Observation Complexes, St. Petersburg, Russian Federation;