Digital Signal Processing

Scientific & Technical

“Digital Signal Processing” No. 3-2012


Topical issue:

- coding and compression;
- motion compensation;
- texture segmentation;
- image fusion;
- quality analysis / improvement;
- subpixel classification;
- image processing using GPUs.

S.V. Umnyashkin, I.O. Sharonov
Using Hexagonal Blocks to Compensate Motion in Videocompression

Motion compensation algorithm based on hexagonal blocks in application to compress video sequences is considered. To compress difference image being obtained after motion compensation SPIHT algorithm based on discrete wavelet transformation (DWT) is used. The experiments show that using hexagonal blocks in motion compensation enables to increase PSNR of processed video sequences by 0.15 on the average.
S.V. Umnyashkin, e-mail:
I.O. Sharonov, e-mail:

Yu.S. Radchenko, S.V. Milyaev
On Spectral Coefficient Correlation for Generalized Fourie Series in Discrete Space

Correlation coefficient for discrete cosine transform, integer cosine transform, Chebyshev discrete transform coefficients within one image block and between neighbor blocks is estimated. Standardized coefficient relations for intra- and interblock correlation of spectral coefficients to correlation distance is analyzed. Chebyshev discrete transform is proved to provide greater coefficient decorrelation as compared to discrete cosine transformation. The results obtained were proved by estimating spectral mode correlation on a real image.
Yu.S. Radchenko, e-mail:
S.V. Milyaev, e-mail:

D.A. Matsypaev, A.G. Bronevich
Fuzzy Parameterized Model for Classifying Blocks of Non-Binary Motion Mask

The problem of detecting moving objects in video stream is solved. The algorithm to detect moving objects using a non-binary motion mask is proposed. Non-binary motion mask being divided into rectangular blocks, the appropriate statistic value averaging the motion measure within the block is estimated for each block. To analyze a distribution function for block statistic values a parameterized mathematical model of moving and stationary blocks is considered. Using the above model, block membership functions are introduced. A criterion for clear discrimination between moving and stationary blocks is proposed. Characteristics to estimate a confidence degree for the results obtained are introduced.
D.A. Matsypaev, e-mail:
A.G. Bronevich, e-mail:

E.I. Minakov, P.S. Seregin
Comparative Analysis for Parallel Image Reconstruction Methods in Magnetic Resonance Tomography

Comparative analysis of simulation-based methods for parallel image reconstruction in magnetic resonance tomography (MRT) is performed. A possibility of regularizing a parallel image reconstruction method in MRT by a compressed sensing method is considered. Special attention is paid to researching SPACE-RIP method regularization. In addition, the paper presents a comparative analysis for parallel reconstruction methods and other regularization methods, SVD-based (singular value decomposition) regularization method in particular.
E.I. Minakov, e-mail:

B.A.Alpatov, V.S. Muraviev, V.V. Strotov, A.B. Feldman
Researching the Effective Use of Image Analysis Algorithms in Unmanned Aircraft Navigation

Research results studying possible use of SIFT and ORB image analysis algorithms in the problem of unmanned aircraft navigation are considered. The research is performed using a dedicated software for simulating onboard video receiving conditions, estimating navigation parameters, calculating effectiveness criteria for image analysis algorithms.
V.S. Muraviev, e-mail:
V.V. Strotov, e-mail:
A.B. Feldman, e-mail:

V.V. Eremeev, A.A. Makarenkov, A.E. Moskvitin, A.A. Yudakov

Improving Object Readability on Hyperspectral Imagery of the Earth’s Surface

Hyperspectral image fusion to form new images with all well-defined objects is considered. Processing results of hyperspectral data are presented.
V.V. Eremeev, e-mail:

N.A. Egoshkin, V.V. Eremeev, A.E. Moskvitin
Image Fusion from Photodetector Lines in Case of Geometric Distortions

Models and algorithms for image fusion from subpixel shifted photodetector lines to increase spatial and radiometric energy resolution in case of geometric distortions for a scene observed.
N.A. Egoshkin, e-mail:
V.V. Eremeev, e-mail:

A.M.Kochergin, A.E.Kuznetsov, V.I. Pobaruev, A.S. Shokol

Technique of Visual Geodesic Orientation for Space Images of the Earth’s Surface

Technique of visual geodesic orientation for space images of the Earth’s surface and a mathematical model it is based on are considered.
A.M.Kochergin, e-mail:
A.S. Shokol, e-mail: ntsomz@

R.V. Tishkin, A.A. Yudakov
Subpixel Classification of Objects on Space Hyperspectral Images

Various approaches to solve the problems of subpixel classification for objects hyperspectral space images are considered. To solve the above problem fuzzy logic-based algorithms are proposed to be used.
R.V. Tishkin, e-mail:
A.A. Yudakov, e-mail:

E.P. Petrov, N.L. Kharina, V.F. Kharyushin

Mathematical Models and Algorithms for Filtering Digital Half-Tone Images Based on Markov Complex Chains

Mathematical model (MM) for digital half-tone image based on Markov complex chain of m-th order is obtained. The model is based on an entropy approach and a principle of superposition for m Markov simple chains. Using MM and a theory for filtering Markov casual processes, algorithms for nonlinear filtering are synthesized.
E.P. Petrov, e-mail:
N.L. Kharina, e-mail:
V.F. Kharyushin, e-mail:

A.M. Tolochko, A.A. Boriskevich
Method of Fast LBP-Feature Computation for Video Image ROI

A method of fast LBP-feature computation for the image based on forming a regular computation structure primitive, defining a mutuality property of threshold and modified threshold functions determining comparison rules for pixels in local neighborhood as well as rules for updating LBP-features is proposed. This method enables to increase a computation rate at about 66 to 72% for rotation invariant LBP-feature with 3x3 primitive as compared to a traditional method due to reducing a specific amount of comparison operations and pixels used for LBP-feature computation.
A.M. Tolochko, e-mail:
A.A. Boriskevich, e-mail:

A.L. Zhiznyakov, D.G. Privezentsev
Using Self-Similarity Distribution Character as Digital Image Sign in Classification Task

Property of digital image self-similarity is described. New image signs characterizing the internal self-similarity distribution and the most similar image areas are proposed. The algorithm to form fractal image signs is described, i.e. characteristic image areas and self-similarity distribution character. Research results for possible use of self-similarity distribution in classifying images are given.
A.L. Zhiznyakov, e-mail:
D.G. Privezentsev, e-mail:

A.Yu. Panteleev
Digital Signal Processing on Modern GPUs

Basic algorithms for digital signal processing (DSP) realized on graphic processor units (GPUs) are considered, i.e. fast Fourie transformation (FFT), convolution, correlation, matrix-vector multiplication. Performance limiting factors are specified for each algorithm. In most cases the performance is limited by memory bandwidth rather than computation capabilities of the units. Examples of using graphic processors in DSP systems are given.
A.Yu. Panteleev, e-mail:

E.V. Medvedeva, E.E. Kurbatova
Method of Texture Image Segmentation Based on Markov Casual Fields

Method of texture segmentation for digital half-tone images (DHI) based on 2D Markov chains is proposed. Estimating the probability of transition between image elements is used as a texture sign. The method is effective in detecting texture areas with various statistical characteristics. It reduces computational resources as well.
E.V. Medvedeva, e-mail:
E.E. Kurbatova, e-mail:

If you have any question please write: