ABSTRACT
This chapter examines the basic methods of image analysis are complemented by several more advanced approaches. Independent component analysis is aimed at separating linearly combined components of a measured/observed signal or image. The emphasis is on the property of independence; in case of images, it may be formulated by saying that independent image components are generated by different stochastic fields. Independently of the used optimization procedure, the optimization criterion will need to be chosen reasonably, utilizing the prior knowledge on mutual independence among the sought-after image components. The basic idea of deep learning, as used in image analysis, is automating the derivation of gradually more and more abstract description of the given image data, this way building models that may be recognized, classified, or formulated in a structured way for further use. Among the goals of the higher-level analysis of individual images or of analysis based on mono- or multimodal fusion of a set of images.
