Studies of Some Algorithms for Digital Picture Processing

摘要

Some algorithms in digital picture processing, characterized by the fact that they work on local windows of the data are discussed. The problem of transposing large matrices with limited high-speed storage, that occurs in connection with the computation of two dimensional integral transforms and convolutions, is addressed. Four such algorithms are presented and their complexity investigated. The use of probabilistic relaxation methods for pixel classification in multispectral pictures is then discussed. It is shown that the approach tends to give better results than some other commonly applied techniques. A local method for image smoothing without blurring and a method to construct structural descriptions of waveforms are also described.