Kopenkov,V.
Abstract:
The article deal with technology of digital images processing on the base of non-linear algorithms. This approach allows to construct the efficient procedure of local image processing. The aim of this research is to workout an algorithm of processing images with predetermined computational complexity and the best quality of processing on the existing data set avoiding a problem of retraining or lesstraining. To achieve this aim we use local discrete wavelet transformation and hierarchical regression to construct local image processing procedure on the base of a training dataset. Moreover, we workout method to estimate the necessity of finishing or continuing the training process. This method is based on of the functional of full cross-validation control which allow to construct processing procedure with predetermined computational complexity and veracity, and with the best quality (for an existing training dataset).