V.E. Dementiev – Ph.D.(Eng.), Associate Professor, Department «Telecommunications», Ulyanovsk State Technical University
E-mail: email@example.com, firstname.lastname@example.org
Among the image processing tasks, the filtering task occupies a special place. This is due not only to its complexity, but also the im-portance for further processing. Indeed, accurate image evaluation is an integral part of most segmentation, detection, and recovery algorithms. Usually for carrying out image filtering using different variations of the well-known linear filters, for example, Kalman or Wiener filter. Their essential disadvantage is the requirement for the stationarity of the processed signal, while real images usually have a pronounced spatial heterogeneity.
In this paper we consider a modification of the double stochastic filter based on the application of sliding filtration. It is shown that such modification can be made for images of any dimension. The basic approaches allowing to obtain recurrent relations for filtering in sliding windows of any shape and size are presented. Tensor relations describing the filtration process and its efficiency are obtained for an important case of two-dimensional image based on the autoregressive stochastic model twice. It is shown that these relations can be used for processing a flat image in at least two important cases. First, for sequential processing of the image when serpentine scan, and secondly with cascade filtration. In the latter case, it is possible to implement a double stochastic filter on high-speed parallel microcontrollers, for example, GPU.
This modification favorably differs from the usual double stochastic filter by significantly higher processing speed and non-causal properties, which allow the filter to adapt faster when crossing borders between objects. The proposed algorithm is compared with the usual two-stochastic filter and the LRU algorithm. The gain in the variance of the error filtering up to 15−20%. The marked properties of the proposed algorithm make it possible to recommend this modification of the double stochastic filter for processing real spatially inhomogeneous signals.
- Gonsales R., Vuds R. Czifrovaya obrabotka izobrazhenij. Izd.3-e. M.: Texnosfera. 2012. 1104 s.
- Vasil’ev K.K., Krasheninnikov V.R. Statisticheskij analiz izobrazhenij. Ul’yanovsk: UlGTU. 2014. 214 s.
- Vasil’ev K.K. Optimal’naya obrabotka signalov v diskretnom vremeni. M.: Radiotexnika. 2016. 282 s.
- Vasil'ev K.K., Dement'ev V.E., Andriyanov N.A. Application of mixed models for solving the problem on restoring and estimating image parameters // Pattern Recognition and Image Analysis (Advances in Mathematical Theory and Applications). 2016. V. 26. № 1. P. 240−247.
- Vasiliev K.K., Dementiev V.E., Andriyanov N.A. Filtration and restoration of satellite images using doubly stochastic random fields // CEUR Workshop Proceedings. 2017. V. 1814. P. 10−20.
- Dement’ev V.E. Primenenie dvazhdy’ stoxasticheskix avtoregressionny’x modelej dlya obrabotki sputnikovy’x izobrazhenij // Radiotexnika. 2017. № 6. S. 18−22.
- Paliy D., Foi A., Bilcu R., Katkovnik V. Denoising and Interpolation of Noisy Bayer Data with Adaptive Cross-Color Filters // SPIE IS&T Electronic Imaging, Visual Communications and Image Processing. January 2008. V. 6822. San Jose. CA.