Multiresolutional fault-tolerant sensor integration and object recognition in images.

摘要

This dissertation applies multiresolution methods to two important problems in signal analysis. The problem of fault-tolerant sensor integration in distributed sensor networks is addressed, and an efficient multiresolutional algorithm for estimating the sensors' effective output is proposed.;The problem of object/shape recognition in images is addressed in a multiresolutional setting using pyramidal decomposition of images with respect to an orthonormal wavelet basis. A new approach to efficient template matching to detect objects using computational geometric methods is put forward. An efficient paradigm for object recognition is described.