Two-stage filtration algorithm with interframe causal processing for multichannel image with presence of uncorrelated noise
Author(s) -
Oleksii Liashuk,
S. V. Vishnevyy,
S. Ya. Zhuk
Publication year - 2015
Publication title -
visnyk ntuu kpi seriia - radiotekhnika radioaparatobuduvannia
Language(s) - English
Resource type - Journals
eISSN - 2310-0397
pISSN - 2310-0389
DOI - 10.20535/radap.2015.63.46-54
Subject(s) - algorithm , noise (video) , pixel , image processing , frame (networking) , computer science , mathematics , channel (broadcasting) , a priori and a posteriori , gaussian , image (mathematics) , artificial intelligence , telecommunications , computer network , philosophy , physics , epistemology , quantum mechanics
In many modern information and technical systems multichannel images have severalapplications. The multichannel image can be presented as a set of images of an object of aresearch which are received in various areas of a frequency range. Such images can be exposedto the distorting influence of the noise. One of the most widespread types of the noise is theadditive uncorrelated noise. Optimal filtration algorithms are characterized by hugecomputing complexity that often limits their practical realization. Therefore development ofalgorithms of a filtration of multichannel images which provide the acceptable precisioncharacteristics at moderate computing expenses is an important task. With use of property ofconditional independence an expression for calculation of a posteriori density of probability ofpixels of the multichannel image at a two-stage filtration with non causal frame and causalinterframe processing in the presence of an additive uncorrelated noise is created. Expressionsfor the image pixel estimate calculation and dispersion of estimate error at a two-stage noncausalframe and causal interframe filtration for a Gaussian images filtration are created.Developed algorithm allows to lower a mean square deviation of estimate error more thantwice in comparison with an algorithm of interframe averaging for a model example when processing the sequence of the Gaussian images distorted by additive white Gaussian noise.One-dimensional processing along each of coordinates is carried out at the first stage of analgorithm, and at the second stage union of the obtained data is made. The created algorithmis concretized for a case of processing of the Gaussian images distorted by additive whiteGaussian noise. The two-stage approach implemented in the synthesized algorithm allows toreduce significantly computing complexity in comparison with an optimal algorithm,providing at the same time the acceptable precision characteristics and considering dimensionof the multichannel image.
Accelerating Research
Robert Robinson Avenue,
Oxford Science Park, Oxford
OX4 4GP, United Kingdom
Address
John Eccles HouseRobert Robinson Avenue,
Oxford Science Park, Oxford
OX4 4GP, United Kingdom