Increasing the resolution of low contrast images in the space of phase characteristics of singular decomposition

Authors

  • Lyudmila Akhmetshina
  • Artyom Yegorov
  • Kostyantin Akhmetshin

DOI:

https://doi.org/10.34185/1562-9945-1-126-2020-02

Keywords:

слабоконтрастные изображения, сегментация, ортогональные преобразования,, сингулярное разложение, визуальный анализ изображений

Abstract

Increasing the sensitivity and veracity of images analyzing (in the sense of increasing probability of detecting of objects of interest) is one of the main goals during images processing. It depends a lot on source data quality especially in the case of low-contrast images processing which make analyzing much harder. One of approaches for solving this problem is transition to new characteristics space. But way of performing this procedure depends on images type and solving task.
One of the ways of transition to new characteristics space is applying one of orthogonalization methods which are often used for solving various tasks during images processing like compressing visual information, determining characteristics for objects recognition decreasing calculations during image processing and so on.
This article describes PCA (Principal Component Analysis) usage for transition to new characteristics space with farther calculation phase characteristics (their obtaining based on eigenvectors matrix) for segmentation sensitivity and resolution increasing in the case of grayscale low-contrast images processing.
In this article transition to new characters space performing was based on scores matrix (T) of PCA transformation usage. In our experiments PCA transformation was based on SVD (singular value decomposition), so scores matrix T was obtained as multiplication of matrices of left singular vectors and eigenvalues. After getting scores matrix T we interpreted it as multidimensional image and used first K most important spectral components (components with maximal eigenvalues) for transition to complex area that gave us new characteristics space. This transformation gives another level of quality of informative abilities of parameters in compare to operating in the field of real numbers and allows to solve some additional tasks of sensitivity and resolution increasing during visual analysis of low-contrast images.
After obtaining this new characteristics space (we interpreted our data in this space as multidimensional image again) we performed calculation of phase characters based on angles between spectral components and absolute value of spectral components.
The main problem of SVD usage was selected type of images. SVD can’t be used for grayscale images processing because it needs multidimensional data. For solving this problem we used automorphic mapping method based on sliding window for getting LxL grayscale images where L is window size (in our experiments L = 3).
We made experimental researches of proposed in this article approach on example of processing various images. In our article we showed the results of processing image of white sheet of paper and image of radiation of fiber optic cable. In the first case our approach allowed to distinguish the structure of paper sheet and in the second case we could obtain the structure of image and highlight low-contrast abnormal areas.
In common our experiments show that applying proposed in this article approach leads to sensitivity of segmentation of low-contrast images increasing and allows to distinguish faint objects of interest.

References

Gonzalez R., Woods R. Digital image processing [transl. from English edited by Chochia P.A.]. – M.: Technosphera, 2006. – 1070 p.

A. Yehorov, L. Akhmetshyna. Optymizatsiya yaskravosti zobrazhenʹ na osnovi neyro-fazzi tekhnolohiy / A. Yehorov, L. Akhmetshyna. Monohrafiya. Vyd. Lambert. -2015. -139 c.

Pearson, K. On lines and planes of closest fit to systems of points in space / K. Pearson // The London, Edinburgh and Dublin Philosophical Magazine and Journal of Sciences. – 1901. – Vol. 6, Issue 2. – P. 559-572.

Hoteling, H. Analysis of complex variables into principal components / H. Hoteling // Journal of Educational Psy- chology. – 1933. – Vol. 24, Issue 6. – P. 417-441.

Yang, J. Two-dimensional PCA: A new approach to ap- pearance-based face representation and recognition / J. Yang, D. Zhang, A.F. Frangi, J.-Y. Yang // IEEE Trans- actions on Pattern Analysis and Machine Intelligence. – 2004. – Vol. 26, Issue 1. – P. 131-137.

Akhmetshina L.G. Increasing the sensitivity of the analysis of x-ray images by the method of automorphic display in the basis of proper images / Akhmetshin AM, Akhmetshina LG // Clinical computer science and telemedicine. - 2008. - T. 4, - No 5. - S. 30-36.

Downloads

Published

2020-03-27