Emphasis of Geomorphometric Features using Wavelet Domain Markov Trees

Wavelet domain hidden Markov models (WHMM) have proven to be useful tools for statistical image processing capturing the key features of the wavelet coefficients of the images. The main drawback of the WHMM is the need for an iterative training, which is computationally expensive. The images (2D signals), we used in this work, are TIF format DEMs (digital elevation models). The DEM´s surface-specific geomorphometric parameters are modeled with conic sections (described by quadratic functions). This modeling method is sensitive to noise data contained by the image. The WHMM estimation algorithm removes the unimportant data within the image and emphasizes some geomorphometric features (especially the linear ones: ridges and channels). The implementation was made using Matlab and Java (for the wavelet domain WHMM estimation algorithm)