Terrain Image Segmentation Based on Wavelet Transform and Partial Least Squares

Image segmentation is a fundamental stage in several domains of knowledge, in particular, remote sensing applications. Its main purpose is to partition an image into constituent regions of interest through characteristics such as pixel intensity or color, texture, and shape of objects. This work presents a two-stage image segmentation method, allowing an overall computation cost reduction while maintaining an adequate discrimination of details. Initially, wavelet transforms are applied to image blocks to extract a number of features and identify homogeneous regions. Then, Euclidean distance and partial least squares classifiers are evaluated for effectiveness on the second stage of the method.