Normal orthogonal transformation algorithm of 2D image

Authors

  • A. I. Rybin National Technical University of Ukraine "Kyiv Polytechnic Institute", Ukraine
  • S. M. Litvintsev National Technical University of Ukraine "Kyiv Polytechnic Institute", Ukraine
  • I. O. Sushko National Technical University of Ukraine "Kyiv Polytechnic Institute", Ukraine

DOI:

https://doi.org/10.20535/RADAP.2015.63.21-32

Keywords:

2D normal orthogonal transformation, pattern recognition, transform coefficient

Abstract

A new 2D transformation algorithm based on algorithm of matrix operator formation in 1D discrete orthogonal transformation is presented. Complexity 2D algorithm creation is a high order of matrix operator when 2D image is presented as a sequence of rows (columns). In this case the order of matrix operator is N2 for image matrix of N order. As result, the number of its elements is equal to N4, which is equivalent of huge figure for image, having size N = 256…1024. A simple algorithm for creation of matrix operator in 2D discrete transformation was obtained. It allows reduce to N3 the memory volume, required for transform coefficient calculation. It makes possible to classify images having matrix of order N ≈ 256…1024. The algorithm is illustrated on the example selected from the ease of inspection results.

Author Biographies

A. I. Rybin, National Technical University of Ukraine "Kyiv Polytechnic Institute"

Rybin A. I.

S. M. Litvintsev, National Technical University of Ukraine "Kyiv Polytechnic Institute"

Litvintsev S. N.

I. O. Sushko, National Technical University of Ukraine "Kyiv Polytechnic Institute"

Sushko I. A.

Published

2015-12-30

Issue

Section

Computing methods in radio electronics