Metric of global texture direction Based on wavelet modulus maxima threshold selection and Power Spectrum

摘要

The present study proposes a metric algorithm for image texture orientation to establish the corresponding mathematical model. First, the image power spectrum is obtained through wavelet maximum edge detection and edge expansion of the texture image. Then threshold image power is set to acquire multiple sets of texture power vector. Then the sum of texture power vector sets makes global power vector, quantitative value of global texture direction. At last, precision of sector and global texture direction under different threshold values is tested on multiple linear stripe and wood texture images. Test results on these two texture databases have a high consistency with human visual perception. It can be concluded that the proposed metric algorithm of texture direction measurement can satisfy the need of high accuracy of man-machine texture recognition model.